Sharkwater Trading

Nebius Director's Trust Files to Sell 7,260 Shares, and the Filing Itself Prices NBIS Near $186

noreply@blogger.com (Unknown) — Fri, 16 Oct 2026 12:21:06 +0000

SharkWater Trading • Data Center Desk • Insider Filing

Nebius Director's Trust Files to Sell 7,260 Shares, and the Filing Itself Prices NBIS Near $186

October 16, 2026

Bottom Line Up Front

A Form 144 filed October 15, 2026 shows The John W Boyton IV Trust of 2006, filing as a Nebius (Nasdaq: NBIS) director, plans to sell 7,260 shares on or about 10/15/2026 under a Rule 10b5-1 plan adopted 03/06/2026. The filing lists an aggregate market value of $1,350,432.60, which works out to about $186.01 a share (desk calc). The same trust sold 50 shares on 09/29/2026 at about $235.01, so the filing's own valuation sits roughly 21% below that print. The sale is routine plan selling. The price embedded in it is the story, and I have not verified an NBIS quote or the cause of any drop.

What the Filing Says

Item	Per Form 144, filed Oct 15, 2026
Seller	The John W Boyton IV Trust of 2006 (signed John W Boyton)
Relationship to issuer	Director
Shares to be sold	7,260 common, via Morgan Stanley Smith Barney
Aggregate market value	$1,350,432.60
Implied price (desk calc)	$186.01 ($1,350,432.60 / 7,260)
Shares outstanding	238,400,165 (7,260 is about 0.003%)
Plan	Rule 10b5-1, adopted 03/06/2026
Acquired	01/03/2025, as restricted stock units from the issuer

Source: SEC EDGAR Form 144, accession 0001950047-26-010270, filed 2026-10-15. Implied price and percentage figures are desk calculations from filed numbers, not a market quote. The filing does not state the date or basis of its market value.

The Trust's Prior Sales

Sale date	Shares	Gross proceeds	Implied price (desk calc)
08/14/2026	5,296	$1,435,142.39	$270.99
09/15/2026	6,364	$1,338,720.86	$210.36
09/29/2026	50	$11,750.50	$235.01

Source: the same Form 144, past three months table. Implied prices are proceeds divided by shares. Sizes and cadence are consistent with a monthly plan; the September 29 sale is a 50 share top-up.

Why It Matters

Each sale in the table is the same plan doing the same job. The size is trivial next to 238.4 million shares. What stands out is the price line. The trust sold near $271 in August and near $235 on September 29. The new filing values the shares near $186, and the Oct 5 Nave sale and Oct 1 withholding sales we covered earlier printed around $234 to $236.

That is a large gap in about ten days, and nothing in the filings explains it. Nebius has filed only this Form 144 since the October 13 Form 4. I found no 8-K, 6-K or press release dated October 14 or 15 tied to a drop, and a web search returned only undated aggregator pages, which I discarded. Whether the filer used a stale, intraday or closing reference is unknown.

A plan seller is a tide table. The tide goes out on schedule whether the water is high or low. The useful information is how far down the beach the waterline sits today.

The Bull Case

It is small and scheduled. About $1.35 million against a market value near $44 billion (238,400,165 shares at the implied $186.01, desk calc) is noise, and the plan was adopted 03/06/2026, months before this print.
Lower prices can pull plan sellers back. Some plans carry price floors. Sales continuing at a lower price would tell you nothing about conviction. The filing does not disclose any floor.
A director still holds the stake. One trust selling 7,260 shares from a position built on RSUs is not a signal about the business.

The Bear Case

The price may have really dropped. If $186 is a true recent level, NBIS is down about 21% from the late September insider prints, with no filing to explain it.
Insider paper keeps coming. Nave's 500,000 share plan sale on Oct 5, the Oct 1 withholding sales and now this trust all landed within about two weeks.
Unexplained flow. Goldman's 13G/A cut on Oct 7 is still unexplained, and this adds another unexplained data point.

The SharkWater Take

This is not a signal about the director. It is a pricing breadcrumb. Do not trade the Form 144. If you hold or watch NBIS, pull a live quote and a chart from your broker today and find out whether $186 is real. If it is, the open question is the cause, and the filings so far do not give one. Until a quote and a cause show up, I am not taking a side on the stock.

Tight lines, SharkWater

Educational and informational purposes only. Not personalized investment advice. All figures sourced as noted and accurate as of publication. The author may hold positions in securities discussed. Do your own work.

Nvidia's $250 Million UK Pledge Is Compute at Nvidia's Own Price, $50 Million a Year, With Terms Still Unsigned

noreply@blogger.com (Unknown) — Thu, 15 Oct 2026 12:17:23 +0000

SharkWater Trading • Semiconductor Desk • Press Release

Nvidia's $250 Million UK Pledge Is Compute at Nvidia's Own Price, $50 Million a Year, With Terms Still Unsigned

October 15, 2026

Bottom Line Up Front

Nvidia (Nasdaq: NVDA) announced on October 13, 2026 that it will commit $250 million to UK agentic AI safety and security research over five years. The release says this is a compute donation, not cash, at $50 million per year, delivered through UK-based NVIDIA Cloud Partners. Nvidia sets the value, and the release does not say how. Provider selection, architecture and commercial terms are, in the release's own words, still to be agreed. This is a goodwill and ecosystem item, not a revenue line, and I have no verified NVDA price move to attach to it.

What the Release Says

Item	Per the Oct 13, 2026 release
Headline value	$250 million
Form	Compute access, not cash
Schedule	Five years, $50 million per year
Delivery	UK-based NVIDIA Cloud Partners, UK-hosted
Named recipients	BOLD (Oxford-led), SOFAIR (UCL-led), Imperial College London, plus the wider UK research community
Split	"The majority" to the wider community, "a proportion" to the named labs. No dollar breakdown.
Still open	Provider selection, technical architecture, commercial terms

Source: NVIDIA Newsroom press release, October 13, 2026 (nvidianews.nvidia.com). The release does not explain how the $250 million valuation was calculated. $50 million per year is $250 million divided by five years, per the release's stated schedule.

The release also carries the standard forward-looking statement language. It says the timing and delivery of the compute are not guaranteed.

Why It Matters

A donation valued by the donor is worth whatever the donor says it is. The release gives no hourly rate, no GPU count and no hardware generation, so the $250 million cannot be checked against anything. That is not an accusation. It is the reason this number should not be read as spend.

The delivery route is the part worth watching. Compute flows through NVIDIA Cloud Partners, which are the operators that buy and run Nvidia hardware. If the commercial terms between Nvidia and those partners include any payment or purchase commitment, that would be an order. If they do not, it is a promotional allocation. The release does not say which. Nothing here is in an SEC filing yet, and I found no 8-K tied to it.

It is a marina handing out free slip time at its own posted rate. The headline number is real on the sign. Whether a boat ever pays for the slip is a different question.

The Bull Case

It seeds future demand. Researchers who build on Nvidia software and hardware tend to keep using it. UK labs named in the release now have five years of access.
It lands with governments. The release carries statements from two UK ministers, which is useful for a company that sells into sovereign AI programs.
It is small per year. At $50 million a year the commitment is modest in absolute terms, though the release gives no basis for what it costs Nvidia to provide.

The Bear Case

The value is self-assessed. No rate card, no GPU count, no hardware generation. The $250 million cannot be verified.
The terms are unsigned. The release says provider selection, architecture and commercial terms are still to be agreed, and delivery is not guaranteed.
Circularity. Nvidia funds usage on partner clouds that run Nvidia chips. Unless partners commit to new purchases, there is no incremental demand to count.

The SharkWater Take

I am not treating this as a catalyst. It is a press-release number with no rate behind it and no signed terms, and nothing in the release changes what Nvidia earns next quarter. The one thing that would change my mind is a filing or partner disclosure showing purchase commitments tied to the program. Until that appears, I file it under reputation, not revenue.

Tight lines, SharkWater

Nvidia Executive Corrects a Form 4: 172,507 Shares Becomes 160,418, a 12,089 Share Cut

noreply@blogger.com (Unknown) — Wed, 14 Oct 2026 12:18:14 +0000

SharkWater Trading • Semiconductor Desk • Insider Filing

Nvidia Executive Corrects a Form 4: 172,507 Shares Becomes 160,418, a 12,089 Share Cut

October 14, 2026

Bottom Line Up Front

Nvidia (Nasdaq: NVDA) executive Nicholas P. Parker, EVP of Worldwide Field Operations, filed a Form 4/A on October 13, 2026 "solely to correct an administrative error in the number of acquired shares" on his September 11 Form 4. The original reported 172,507 restricted stock units granted September 9. The corrected filing reports 160,418. That is 12,089 fewer shares, or 7.0 percent. It is an RSU award at $0, not a sale, and it moves no stock to market. The story is that a grant count was wrong for 32 days on the public record, and the company gave no reason beyond "administrative."

What Changed

Filing	Form	Filed	RSUs reported	Transaction date
Original	Form 4	Sep 11, 2026	172,507	Sep 9, 2026
Correction	Form 4/A	Oct 13, 2026	160,418	Sep 9, 2026
Difference			-12,089 (-7.0%)

Source: SEC EDGAR. Original: accession 0002152188-26-000005 (filed Sep 11, 2026). Amendment: accession 0002152188-26-000007 (signed Oct 13, 2026). Percentage computed by SharkWater from the two filed figures. Both filings code the award A at $0 and show direct ownership.

Everything else matches. Same grant date, same vesting schedule: 10 percent vests December 9, 2026, then quarterly, with tranches of 7.5, 5 and 2.5 percent over the following quarters, fully vested near the four-year mark. Neither filing checks the Rule 10b5-1 box, which fits an award rather than a market trade.

Why It Matters

Form 4s are the public ledger for insider equity. When the ledger is wrong, anyone tracking ownership, dilution or the value of an executive grant is working from bad inputs. A 7 percent error on one executive is small in dollars against a company this size, and I will not put a dollar value on it because I do not have a verified NVDA price series for this week.

The more interesting angle is the pattern. This desk flagged on October 12 that a Tench Coxe trust balance fell by 1,000,000 shares across filings while the filings show only one 500,000 share gift. This correction is a different insider and a different type of error. It still says the same thing: the filing chain at this company is not self-checking, and the corrections arrive weeks late.

A harbor chart that is off by a few feet is fine until you are the boat that draws the most. Most readers never touch the number. The ones who model insider supply need it right.

The Bull Case

The fix is the system working. Nvidia and the filer caught and amended the error on their own, and the amendment is public on EDGAR.
The dollar impact is tiny. 12,089 RSUs is a rounding error against the share count of a company of this size, and no cash or stock changed hands.
The terms did not move. Grant date, vesting dates and tranche percentages are identical in both filings.

The Bear Case

The reason is thin. "Administrative error" does not say whether the grant itself was reduced or the first filing was simply mistyped. I could not verify which.
The delay is long. The record sat wrong for 32 days, September 11 to October 13.
It is the second oddity in a week. The unreconciled 500,000 share Coxe gap from the October 12 post is still open. NOT VERIFIED either way.

The SharkWater Take

I am not building a trade on this and you should not either. It is a bookkeeping correction on an RSU award, and RSU awards do not create selling pressure the day they are granted. I am keeping it on the board for one reason. Two ledger problems at the same issuer in two days is a pattern worth one more look, and if a third shows up I will write the pattern, not the filing. Until then this is housekeeping, and I am calling it that.

Tight lines, SharkWater

Nvidia and Cisco Announce USFP, a 400G-Per-Lane Optics Standard With No Ship Date

noreply@blogger.com (Unknown) — Tue, 13 Oct 2026 12:18:27 +0000

SharkWater Trading • Semiconductor Desk • Networking

Nvidia and Cisco Announce USFP, a 400G-Per-Lane Optics Standard With No Ship Date

October 13, 2026

Bottom Line Up Front

Nvidia (Nasdaq: NVDA) and Cisco announced on October 12, 2026 a new optical module format called Ultra-Dense Small Form Factor Pluggable, or USFP, built for 400Gb/s per lane signaling. The first planned module carries 16 lanes for up to 12.8Tb/s, and the companies claim 4x the bandwidth density and about 1.3x the power efficiency of 200G pluggables. The release gives no ship date and no pricing. It describes plans and a future multi-source agreement, not product. This is a roadmap signal, not a revenue event.

What Was Announced

The format targets Ethernet AI fabrics moving from 200Gb/s to 400Gb/s per lane. It ties to Nvidia Spectrum-X Ethernet and Cisco Silicon One networks. TE Connectivity and Amphenol are named as partners on cages, connectors and modules. The release says more partners are expected.

Item	Per the release
Lane signaling	400Gb/s per lane, versus 200Gb/s per lane today
First planned module	16 full-duplex lanes, OSFP-sized footprint, up to 12.8Tb/s
Claimed gain vs 200G pluggables	4x bandwidth and front-panel density, about 1.3x power efficiency
Power	Up to 100 watts without a dedicated embedded cooling circuit in the transceiver
Ecosystem	TE Connectivity and Amphenol on cages, connectors, modules. Other partners expected
Ship date	Not stated. Release says "plans" and an upcoming community-led multi-source project

Source: NVIDIA Newsroom, "NVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories," October 12, 2026. All specifications are vendor claims from the release, not independent measurements.

The mechanical detail is a two-dimensional array connector with two-step insertion, which the companies say reduces contact wear and protects signal integrity. The design can expand to 32 or more lanes later.

What Is Missing

There is no availability date, no price, and no named optics supplier beyond the connector makers. The open specification is described as an upcoming community-led multi-source project. That means the standard does not exist yet in finished form. I did not find any SEC filing tied to this announcement as of the October 13, 2026 morning check, and I have no verified price or volume data for NVDA or any networking name this morning.

Broadcom (Nasdaq: AVGO) is on my watchlist, and the release does not mention it. Whether USFP helps or hurts any switch silicon vendor outside Nvidia and Cisco depends on who joins the project. The release does not say. I will not guess.

A new connector standard is a harbor survey, not a ship. The chart tells you where the channel will be dredged. It tells you nothing about who gets paid to dredge it, or when.

The Bull Case

Density is the right problem. Four times the front-panel bandwidth in the same footprint speaks to the actual bottleneck in large clusters, per the vendors.
Two heavyweight sponsors. Nvidia on the compute side and Cisco on the switching side give the format a credible start.
Connector makers are already named. TE Connectivity and Amphenol are attached from day one, which points to a real supply chain effort.

The Bear Case

No date, no price. Everything is a plan. Standards efforts can stall, split or arrive late.
Vendor numbers only. The 4x and 1.3x figures are the sponsors' own comparisons to 200G modules. No independent test exists.
Rivals can answer. A multi-source project open to others also lets competitors shape the final spec.

The SharkWater Take

I am not trading this headline. A form factor announcement with no ship date does not move this quarter's earnings for anyone. I care about two things: who joins the multi-source project, and whether a module supplier names a sampling date. Until one of those happens, this belongs in the notebook, not the portfolio. I have no verified price reaction to report, and I will not infer one from the press release.

Tight lines, SharkWater

Nvidia Director Tench Coxe Gifts Another 500,000 Shares. It Is Not a Sale, and the Share Count Does Not Add Up

noreply@blogger.com (Unknown) — Mon, 12 Oct 2026 12:18:09 +0000

SharkWater Trading • Semiconductor Desk • Insider Filing

Nvidia Director Tench Coxe Gifts Another 500,000 Shares. It Is Not a Sale, and the Share Count Does Not Add Up

October 12, 2026

Bottom Line Up Front

Nvidia (Nasdaq: NVDA) director Tench Coxe filed a Form 4 on October 9, 2026 reporting a 500,000 share gift on October 7, code G, price $0, under a Rule 10b5-1 plan he adopted on March 19, 2026. The trust that made the gift held 23,671,360 shares afterward. This is a gift, not a market sale, so it is not selling pressure. The oddity is arithmetic: the trust balance fell 1,000,000 shares between the August 5 and October 7 filings, but I found only one 500,000 share gift in that span. The other 500,000 is NOT VERIFIED.

The Filing

Transaction date	Code	Shares	Trust shares after	Filed
Jul 1, 2026	G (gift)	500,000	25,171,360	Jul 6, 2026
Aug 5, 2026	G (gift)	500,000	24,671,360	Aug 7, 2026
Oct 7, 2026	G (gift)	500,000	23,671,360	Oct 9, 2026

Source: SEC EDGAR Form 4 filings by Tench Coxe, issuer NVIDIA Corp: accessions 0001197647-26-000005, -000007 and -000011. All three cite a 10b5-1 plan adopted March 19, 2026. Coxe is a director only. The trust shares are held indirectly; he disclaims beneficial ownership beyond his pecuniary interest.

A code G is a gift without consideration. Nothing is sold and nothing hits the tape. The same filings show Coxe also holds 57,378 shares directly and 4,852,480 shares through a profit sharing plan trust, both unchanged. His June 25, 2026 Form 4 showed a routine 1,211 share board RSU grant, a different animal.

The Gap

The trust held 25,671,360 shares per the June 29 filing. After the July 1 gift it held 25,171,360. After August 5 it held 24,671,360. The October 7 filing shows 23,671,360. That is a million shares down from August 5, and the filings list only 500,000.

Coxe's own EDGAR feed shows no Form 4 between August 7 and October 9. Possible explanations include a September gift filed late or under another filer, or a correction I did not reach. I did not resolve it and I will not guess. It matters because insider forms are only as good as the chain of balances.

Reading a Form 4 chain is like tracking a tide gauge. One odd reading is a bad sensor. Two in a row is the harbor changing.

The Bull Case

A gift moves no shares to market. Code G means no sale at any price. Supply to the open market is unchanged.
The plan is pre-set. Adopted March 19, 2026 and cited on every gift, so timing is not a reaction to the stock.
Holdings stay enormous. 23,671,360 trust shares plus 4,852,480 in the profit sharing trust, per the filings.

The Bear Case

Donees can sell. A gifted share can be sold by the recipient. The filing does not say who received these shares.
The balance chain has a hole. A 500,000 share difference with no matching filing is a data-quality flag.
The cadence is steady. Three gifts of identical size across four months reads like a program that continues.

The SharkWater Take

I am not building a trade on this and neither should you. A gift is not a sale, and the plan says this is mechanical. I care about the missing 500,000 shares. If it turns out to be a late or missing Form 4, that is worth a second post. If it is explained by a different filer, it goes in the noise bin. I have no verified NVDA price series for this week, so I will not value the gift or say what the market did with it.

Tight lines, SharkWater

Micron's 10-K: $133.2 Billion in Revenue, $59 Billion in Free Cash Flow, and Capex Nearly Doubled

noreply@blogger.com (Unknown) — Sun, 11 Oct 2026 12:18:32 +0000

SharkWater Trading • Semiconductor Desk • Annual Report

Micron's 10-K: $133.2 Billion in Revenue, $59 Billion in Free Cash Flow, and Capex Nearly Doubled

October 11, 2026

Bottom Line Up Front

Micron (Nasdaq: MU) filed its fiscal 2026 Form 10-K on October 9, 2026. Revenue for the 53 weeks ended September 3, 2026 was $133.19 billion, up from $37.38 billion a year earlier. Net income was $84.97 billion, or $74.33 per diluted share. Operating cash flow of $89.68 billion less $30.71 billion of capital spending leaves $58.96 billion of free cash flow (desk calculation). The company ended the year with $73.45 billion in cash and investments against $5.18 billion of debt. I have no verified MU price or volume series for the week, so the market reaction is NOT VERIFIED.

The Year in Four Columns

USD millions	FY2024	FY2025	FY2026
Revenue	25,111	37,378	133,188
Gross margin	5,613	14,873	107,504
Operating income	1,304	9,770	99,340
Net income	778	8,539	84,969
Diluted EPS (USD)	0.70	7.59	74.33
Operating cash flow	8,507	17,525	89,675
Capital expenditures	8,386	15,857	30,712

Source: Micron Technology Form 10-K for the fiscal year ended September 3, 2026, filed October 9, 2026 (accession 0000723125-26-000023), financial statement pages R3 and R7. Fiscal 2026 had 53 weeks. Cash flow cells were read by an automated page reader and the source flagged some run-together cells in the prior-year columns; the FY2026 column was not flagged.

What the Filing Adds Over the Press Release

The September 30 earnings release gave the quarter. The 10-K gives the full year, audited, and the year is the bigger story. Revenue rose 3.56 times. Gross margin on the filed GAAP numbers was 80.7 percent for the year, against 39.8 percent in fiscal 2025. Net income was 63.8 percent of revenue. All three are desk arithmetic on the filed figures.

Capital spending rose 94 percent to $30.71 billion, but fell as a share of revenue to 23.1 percent. That is the number to watch. A memory maker spending a quarter of revenue on fabs is normal. The prior two years in the table above show nothing like this combination.

The segment split is concentrated by size. Fiscal 2026 revenue by business unit was CMBU $43.09 billion, CDBU $37.59 billion, MCBU $36.60 billion and AEBU $15.89 billion. By product, DRAM was $100.68 billion and NAND $31.79 billion. The business section says roughly half of revenue has come from the top ten customers in each of the last three years.

A reservoir that more than triples in a wet year tells you about the rain. It does not tell you what the next dry year looks like.

Balance Sheet and Capital Return

USD millions	Aug 28, 2025	Sep 3, 2026
Cash and equivalents	9,642	38,364
Short-term investments	665	5,070
Long-term marketable investments	1,629	30,019
Total debt (current plus long-term)	14,577	5,179
Total equity	54,165	138,378

Source: Micron Form 10-K, balance sheet page R4, filed October 9, 2026. Cash and investments total ($73,453 million) and debt totals are desk sums of the filed lines. Net cash of about $68.3 billion is a desk calculation.

Debt fell by $9.4 billion in the year. Capital return was small: about $610 million of dividends and $650 million of program repurchases in fiscal 2026, per the cash flow page (some cells in that table were flagged as garbled by the reader; verify before quoting). Against $85 billion of net income, that is a rounding error. The dividend declared on September 30 is $0.15 per share, payable October 29, 2026. Shares outstanding were 1,131,423,212 on October 2, 2026, per the 10-K cover.

What Is Still Open

The filing text I could read covers risk factors, segment and product revenue, and the financial statements. It states that HBM3E 12-high was the majority of fiscal 2026 HBM shipments and that HBM4 36GB 12-high began volume production in 2026. I did not read MD&A, so there is no sourced fiscal 2027 capital spending outlook in this post. I also have no verified MU price series. Pre-market data for October 1 is an aggregator figure from the prior scan and is not used here.

The Bull Case

The cash pile is real. $73.45 billion in cash and investments against $5.18 billion of debt gives the company room to fund fabs, pay down the rest and still return capital, all out of one year of cash flow.
Margins held at an extreme. An 80.7 percent GAAP gross margin for a full 53-week year is audited, not a quarter's peak. The guide for the next quarter, from the September 30 release, sits near 86 percent non-GAAP.
The reinvestment rate is falling. Capex took 23.1 percent of revenue, down from 42.4 percent in fiscal 2025 (desk arithmetic). Revenue outran spending.

The Bear Case

Half the revenue sits with ten customers. The filing says so in plain words. A pause in orders from a few buyers moves the whole line.
The year is a single point on a cycle. Revenue was $25.11 billion two fiscal years ago. A business that has swung from $0.78 billion to $84.97 billion of net income in two years has shown how fast it can swing back.
Capex is still doubling. $30.71 billion is 1.94 times the prior year. New capacity lands into whatever demand exists when it comes online, and the filing's own risk factors flag HBM yield and cleanroom intensity.

The SharkWater Take

I will not turn an annual report into a trade, and this one does not need to be. The 10-K confirms what the September 30 release said, and adds the audited year: $59 billion of free cash flow and net cash of roughly $68 billion. The risk is the shape of the cycle, not the balance sheet. A company with this much cash survives a downturn. Whether the stock already prices that in is a different question, and I have no verified price series to answer it. Until I have one and the HBM and capex detail from MD&A, MU stays on the watch list, not the trade list.

Tight lines, SharkWater

Nebius Insider Tape Sorted: One 500,000 Share Plan Sale and Four Routine Ones

noreply@blogger.com (Unknown) — Sat, 10 Oct 2026 12:18:15 +0000

SharkWater Trading • Data Centers • Insider Filings

Nebius Insider Tape Sorted: One 500,000 Share Plan Sale and Four Routine Ones

October 10, 2026

Bottom Line Up Front

Five Nebius (Nasdaq: NBIS) insider sales hit EDGAR in the first week of October. Only one is large. COO Ophir Nave sold 500,000 shares on October 5, 2026 for roughly $117.8 million (desk calculation) under a 10b5-1 plan adopted May 22, 2026. The CEO, the Chief Infrastructure Officer and the CTO together sold 81,414 shares on October 1 at $234.16, about $19.1 million, and each Form 4 says the sale only covered tax on vesting units and was automatic. A director's trust sold 50 shares. Strip out the routine paper and the tape is one seller, and his plan says it is finished.

Who Sold, and Why

Seller	Date	Shares	Price	Type per filing
Nave, COO	10/5/2026	500,000	$235.53 avg (desk calc)	10b5-1 plan, adopted 5/22/2026
Volozh, CEO	10/1/2026	37,778	$234.16	Tax withholding, automatic
Korolenko, Chief Infrastructure Officer	10/1/2026	29,090	$234.16	Tax withholding, automatic
Shtan, CTO	10/1/2026	14,546	$234.16	Tax withholding, automatic
Boynton trust, director	9/29/2026	50	$235.01	10b5-1 plan, adopted 3/6/2026

Source: Nebius Group N.V. Forms 4 and Forms 144 on EDGAR: accessions 0001513845-26-000126 (Nave), -000120 (Volozh), -000118 (Boynton), -000122 (Shtan), -000124 (Korolenko, role per its Form 4), and Form 144 0001950047-26-010048 (Nave) and 0001950047-26-009868 (Boynton trust). Nave average price and all dollar totals computed by this desk from the filed rows.

Sizing the Pile

Nave's sale is 6.1 times the three tax sales combined. Against the 238,400,165 shares outstanding cited on the Nave Form 144, his sale is about 0.21 percent of the company and the tax sales together are about 0.03 percent. Neither moves the float.

The tax sales are mechanical. Each Form 4 says the shares were sold when restricted share units vested, solely to cover estimated withholding, under automatic sale instructions in the unit agreement and not as a discretionary trade. After the sale the CEO still held 785,236 shares directly.

The Boynton trust is a drip, not a surprise. Its September 29 Form 144 lists three earlier plan sales in the prior three months: 6,958 shares on July 15, 5,296 on August 14 and 6,364 on September 15, 18,618 shares for $4,144,629.26 in total. The 50 share sale is a stub on a monthly schedule.

Five boats left the harbor this week. Three were ferries running on a timetable, one was a rowboat on its usual route, and one was a trawler hauling out half its direct catch.

What Is Still Open

Nave's Form 4 remarks state that no further sales will be made under his plan. That is the issuer's own filing, but a new plan can be adopted later, and the filing does not say when. This desk has no verified NBIS price or volume series for the week, so any market reaction to this selling is NOT VERIFIED. The Goldman Sachs 13G/A cut covered here yesterday sits alongside this paper but is a separate and unrelated filer.

A correction to this desk's October 7 post: it gave October 7 as the Nave Form 4 filing date. The EDGAR feed shows filing date October 6, 2026, and the filing is signed October 6.

The Bull Case

Mostly mechanical. Four of five sales are tax cover or a standing monthly plan. None of them is a discretionary call on the stock.
The big seller is done. Nave's remarks say no further sales under the May 22 plan, so the overhang from that plan is cleared.
Small against the share count. All five sales together are about 0.24 percent of shares outstanding (desk arithmetic).

The Bear Case

The COO is the one with a choice. A plan is chosen in advance, and he chose to sell 500,000 shares, about 52 percent of his direct holding by the Form 4 balances.
A new plan is one filing away. The remarks close this plan. They do not rule out the next one.
Stacked on the Goldman cut. The bank's reported stake fell 43.6 percent between two 13G/A event dates. Different filers, same two weeks of supply.

The SharkWater Take

I have written about this tape for four straight days, and today's reading is the least alarming one. When the paper is sorted, one executive made a real sale and the rest is payroll mechanics. Nave's plan is closed, and that matters more to me than the headline dollar figure. The open item is the Goldman cut, which is flow I cannot explain. I do not have an edge on insider filings alone, and I will not invent one. I want a verified price series and the next 13G event date before NBIS gets a trade idea from me.

Tight lines, SharkWater

Goldman Sachs Cut Its Nebius Stake by 44 Percent in Two Months

noreply@blogger.com (Unknown) — Fri, 09 Oct 2026 12:18:28 +0000

SharkWater Trading • Data Centers • Ownership Filings

Goldman Sachs Cut Its Nebius Stake by 44 Percent in Two Months

October 9, 2026

Bottom Line Up Front

Yesterday this desk could not say whether Goldman's Nebius (Nasdaq: NBIS) Schedule 13G/A was a build or a cut. The prior amendment answers it. Amendment No. 1 (event date July 31, 2026) reported 23,111,112.59 Class A shares, 10.5 percent of the class. Amendment No. 2, filed October 7, 2026 (event date September 30, 2026), reports 13,023,110.37 shares, 5.1 percent. That is a drop of about 10.09 million shares, or 43.6 percent of the position (this desk's arithmetic). Why Goldman sold, and for whom, is not in the filing.

The Two Filings Side by Side

Filing	Event date	Shares	% of class
13G/A No. 1 (0000886982-26-000308)	7/31/2026	23,111,112.59	10.5%
13G/A No. 2 (0000886982-26-000526)	9/30/2026	13,023,110.37	5.1%

Source: Schedule 13G/A filings on EDGAR, Goldman Sachs Group and Goldman Sachs & Co. LLC, each reporting identical figures. Amendment No. 1 filing date not read by this desk; only its event date is shown.

What It Does and Does Not Say

Both filings report shared voting and dispositive power and zero sole power. Goldman files under Rule 13d-1(b), the institutional route. Exhibit 99.3 states the filing excludes shares held by operating units of the group. So this is the reporting entity's aggregate, not necessarily house inventory.

A 13G does not separate client assets from the bank's own book, and it does not name a buyer. A drop of this size could be client outflows, hedge unwinds, or a market-making position shrinking. The form cannot tell us which.

One note on the math. 10.5 percent implies a class near 220 million shares, and 5.1 percent implies about 255 million. Rounding and a changing share count could explain part of that. This desk did not find the denominators in either filing, so the implied counts are unreconciled.

Yesterday we saw a big boat at the dock and could not tell if it was loading or unloading. The earlier log entry shows the hold was nearly twice as full in July.

The Bull Case

Passive, not conviction. A bank's 13G position moves with client flow. The reduction is not evidence of a negative view.
Supply was absorbed. Roughly 10 million shares left one holder, and the filing discloses no forced event behind it.
Still above the line. Goldman remains a 5.1 percent reporter, so it has not exited.

The Bear Case

A large holder cut its stake by 44 percent. Ten million shares is real supply, and it landed alongside the Nave Form 144 for 500,000 shares reported October 5, 2026.
The line is one step away. At 5.1 percent, the next reduction could take Goldman below the reporting threshold and out of public view.
Price reaction unknown. This desk had no primary price source for NBIS, so whether the market already reflects the sale is NOT VERIFIED.

The SharkWater Take

I read this as supply, not a verdict. Goldman's 13G tells me flow, and flow is not conviction. What bothers me is the stack: a 44 percent bank cut, a 500,000 share insider plan sale, and Form 4s from the same week. None of it is a thesis alone. Together it says the easy buyers have thinned. I am not taking a position off a passive filing. I want the next 13G event date and a verified NBIS price series before this becomes a trade idea.

Tight lines, SharkWater

Source: Schedule 13G/A Amendment No. 2, EDGAR accession 0000886982-26-000526, filed October 7, 2026, event date September 30, 2026; Schedule 13G/A Amendment No. 1, accession 0000886982-26-000308, event date July 31, 2026. Percent change is this desk's arithmetic. Educational and informational purposes only. Not personalized investment advice. The author may hold positions in securities discussed. Do your own work.

Goldman Sachs Reports 5.1 Percent of Nebius Class A, and a Date Fix on the Nave Form 4

noreply@blogger.com (Unknown) — Thu, 08 Oct 2026 12:17:41 +0000

SharkWater Trading • Data Centers • Ownership Filings

Goldman Sachs Reports 5.1 Percent of Nebius Class A, and a Date Fix on the Nave Form 4

October 8, 2026

Bottom Line Up Front

A Schedule 13G/A filed October 7, 2026 shows The Goldman Sachs Group and Goldman Sachs & Co. each reporting 13,023,110 Class A ordinary shares of Nebius Group (Nasdaq: NBIS), or 5.1 percent of the class, as of an event date of September 30, 2026. This is Amendment No. 2. The prior percentage is not in the filing text this desk read, so direction of change is NOT VERIFIED. Separately, a correction: EDGAR lists the Nave Form 4 (accession 0001513845-26-000126) as filed October 6, not October 7 as our October 7 post stated. The sale figures in that post stand.

What the Filing Shows

The filing is a Schedule 13G/A, a passive holder report. Goldman is the joint filer on both entities. The filing reports 13,023,110.37 Class A shares for each filer and 5.1 percent of the class. The 5.1 percent sits barely above the 5 percent reporting line, so small changes in share count move it across.

The numbers do not tie cleanly to the Nave Form 144. That filing cites 238,400,165 shares outstanding. 13,023,110 divided by that count is about 5.5 percent, not 5.1 percent. The 13G/A percentage implies a Class A count near 255 million. The gap may come from different share-class counts or dates. This desk did not find the denominator in the filing, so treat it as unreconciled. Percent math here is this desk's arithmetic.

Large banks file 13Gs for client and trading positions alike. A 13G does not say whether the shares are house inventory, hedges, or client assets. It does not carry a direction.

A 13G/A is a harbor log entry. It tells you a big boat is tied up at the dock on a given day. It does not tell you whether it is loading or unloading.

The Date Fix

The October 7 post said the Nave Form 4 was filed October 7. The EDGAR submissions feed lists accession 0001513845-26-000126 with a filing date of October 6, 2026. The Form 144 for the same 500,000 shares (accession 0001950047-26-010048) is dated October 5. So the Form 4 landed the day after the sale, not two days after. The Form 144 values ($121,405,000, 238,400,165 shares outstanding, plan adopted May 22, 2026) match what we reported. One oddity to flag: the Form 144 lists sales in the past three months as "Nothing to Report" while its remarks cite a 10b5-1 plan.

The Bull Case

Passive filer. A 13G/A is the lighter ownership form. It is not an activist or control filing.
Large institution at the table. A top bank carrying 13 million shares signals the stock has deep institutional liquidity.
No sale disclosed. The filing reports a position. It does not report a distribution.

The Bear Case

Direction unknown. Without the prior filing percentage, the amendment could be a build or a cut.
Near the line. At 5.1 percent, one more reduction could drop the holder below the reporting threshold and out of view.
Insider supply alongside. The Nave sale of 500,000 shares and the Form 4s filed October 5 sit in the same week as this amendment.

The SharkWater Take

I am not taking a view on this one. A bank 13G/A without the prior percentage is a data point with no slope, and the share count does not reconcile to the Form 144. The only thing worth acting on today is the correction, and it is already made. The question I want answered is the prior Goldman percentage and the Class A denominator. Until then, no edge here.

Tight lines, SharkWater

Source: Schedule 13G/A, EDGAR accession 0000886982-26-000526, filed October 7, 2026, event date September 30, 2026; Nebius Form 4 accession 0001513845-26-000126; Form 144 accession 0001950047-26-010048. Educational and informational purposes only. Not personalized investment advice. The author may hold positions in securities discussed. Do your own work.

Nebius COO Sold 500,000 Shares at a $235.53 Average, Over Half His Direct Stake

noreply@blogger.com (Unknown) — Wed, 07 Oct 2026 12:17:28 +0000

SharkWater Trading • Data Centers • Insider Filings

Nebius COO Sold 500,000 Shares at a $235.53 Average, Over Half His Direct Stake

October 7, 2026

Bottom Line Up Front

The Form 144 we flagged yesterday is now a completed trade. Ophir Nave of Nebius Group (Nasdaq: NBIS) sold 500,000 shares on October 5, 2026 at a computed average of $235.53, roughly $117.8 million gross, per a Form 4 filed October 7. That is below the $121.4 million the Form 144 implied, which pegged the shares near $242.81. The Form 4 shows 454,685 shares held directly afterward, so the sale took about 52 percent of his directly held shares. It ran under a Rule 10b5-1 plan adopted May 22, 2026.

What the Form 4 Shows

A Form 144 is a notice. A Form 4 is the receipt. This one reports twelve sale lines on October 5, 2026, all coded as sales, all direct ownership. The weighted average prices run from $231.12 to $243.50. The pre-sale direct balance of 954,685 shares is derived from the first line (938,465 remaining plus 16,220 sold), not stated outright.

Shares sold	Weighted avg price
16,220	$231.12
61,210	$232.37
56,426	$233.19
48,845	$234.21
106,718	$235.24
73,588	$235.99
18,224	$237.15
48,655	$238.22
48,061	$239.23
14,853	$240.00
1,200	$241.02
6,000	$243.50
500,000 total	$235.53 (desk calc)

Source: Nebius Group N.V. Form 4, EDGAR, filed October 7, 2026 (accession 0001513845-26-000126), all sales dated October 5, 2026, Rule 10b5-1 plan adopted May 22, 2026. Each row is a weighted average within a price range stated in the filing. Total and average computed by this desk from the rows.

The filing remarks say the shares sold were settled restricted share units, about 17 percent of the reporting person's granted equity. Nave is listed as a director and an officer with the title COO. The Form 144 listed him as Director and Officer, so the two filings agree on the person.

The Form 144 value of $121,405,000 is an estimate made at filing. Actual proceeds came in about $3.6 million lower on this desk's math, which means the stock traded below the Form 144 reference price for most of the session. Nebius has not issued a press release on this, and none is required.

A Form 144 is the weather forecast and a Form 4 is the logbook. The forecast said the boat would leave at $242. The logbook says it left across a dozen tides between $231 and $243, and most of the catch went out in the low to mid $230s.

What Is Still Open

The plan terms are not public. This desk does not know whether the May 22 plan has more tranches. The September 30 Nebius Form 144 (accession 0001950047-26-009868) was not read, so how it relates to this sale is unknown. Two other officers sold shares on October 1 at $234.16 to cover tax on vested units, which is routine and a different animal at about 9 percent of this size.

The Bull Case

Pre-scheduled. The plan date of May 22, 2026 predates the sale by more than four months. The timing was set long before October 5.
Small against the share count. 500,000 shares was about 0.21 percent of the 238,400,165 shares outstanding cited on the Form 144.
Still a large holder. 454,685 shares remain in direct ownership, so he kept a meaningful position.

The Bear Case

Half the direct stake. Selling about 52 percent of directly held shares in one day is a large reduction, plan or not.
Clustered paper. A second Form 144 on September 30 and three Form 4s on October 5 mean insider selling has stacked up over a single week.
Sold below the reference. Proceeds came in under the Form 144 estimate, and the 12 line ladder shows the shares went out across a $12 range, mostly in the $230s.

The SharkWater Take

The 10b5-1 plan takes most of the sting out of the timing argument, and 0.21 percent of shares outstanding will not change the float. What I notice is the size relative to his own position. Half of a direct stake is not a trim. I am not treating this as a sell signal for the company, because a plan sale alone does not tell me the executive's view of the business. I am treating it as a reason to read the next Nebius filings closely, and to check the September 30 Form 144 before drawing any pattern. No edge here for a new position on this alone.

Tight lines, SharkWater

Educational and informational purposes only. Not personalized investment advice. All figures sourced as noted and accurate as of publication. Options involve substantial risk of loss. The author may hold positions in securities discussed. Do your own work.

Nebius Director Files to Sell 500,000 Shares, $121 Million, Under a May Plan

noreply@blogger.com (Unknown) — Tue, 06 Oct 2026 12:16:35 +0000

SharkWater Trading • Data Centers • Insider Filings

Nebius Director Files to Sell 500,000 Shares, $121 Million, Under a May Plan

October 6, 2026

Bottom Line Up Front

Nebius Group (Nasdaq: NBIS) filed a Form 144 on October 6, 2026 for 500,000 shares with an aggregate market value of $121,405,000, about 0.21 percent of 238,400,165 shares outstanding. The seller is listed as Ophir Nave, Director and Officer, under a plan adopted May 22, 2026. The filing reports no sales in the prior three months. The executed price and post-sale holdings are not yet public because the matching Form 4 had not appeared in the EDGAR feed at the time of writing.

What Was Filed

A Form 144 is notice of intent to sell restricted or control securities. It is not the sale itself. The filing gives an approximate sale date of October 5, 2026, which is the day before it was filed, so the trade may already be done. The Form 4 that reports actual execution is due within two business days of the trade.

Form 144 field	As filed
Seller (per filing)	Ophir Nave, listed as Director and Officer
Shares to be sold	500,000
Aggregate market value	$121,405,000 (implies about $242.81 per share)
Approximate sale date	10/05/2026
Plan adoption date	05/22/2026
Shares outstanding	238,400,165 (about 0.21% sold)
Sold in prior 3 months	None reported

Source: Nebius Group N.V. Form 144, SEC EDGAR, filed October 6, 2026 (accession 0001950047-26-010048).

Why the Plan Date Matters

The plan was adopted May 22, 2026, more than four months before the filing. A plan adopted that far back means the sale instructions were set well before this filing. The filing does not say the plan is a Rule 10b5-1 plan beyond the adoption date, and this desk has not read the plan terms. Whether the 500,000 is a single tranche or one of several is not stated.

Context from the same feed: three Form 4s were filed October 5 for sales dated October 1. The two this desk read, from the Chief Infrastructure Officer (29,090 shares at $234.16) and the Chief Technology Officer (14,546 shares at $234.16), were automatic sales to cover tax withholding on vested restricted share units. Those are routine. The director filing is a different animal at roughly 11 times the combined size.

A harbor pilot unloading ballast before the swell hits and a crew trimming sail for the weather are doing different things with the same rope. A tax-cover sale is trimming sail. A planned half-million-share sale is a bigger move, and the plan date is the logbook entry that tells you when the captain decided.

The Bull Case

Scheduled, not reactive. The plan was adopted May 22, 2026, so the sale was set months ahead of this filing.
Small against the float. 500,000 shares is about 0.21 percent of shares outstanding, which is modest supply for a name this size.
No prior selling. The filing reports nothing sold in the last three months, so this is not a continuing drip.

The Bear Case

Dollar size is real. $121.4 million is not a tax-cover sale. It is a deliberate liquidation by a person listed as both director and officer.
Part of a cluster. A separate Nebius Form 144 was filed September 30 (contents not read by this desk), and three Form 4s followed October 5. Insider paper is stacking up.
Unknowns remain. No executed price, no post-sale holdings, no plan terms. The tranche count is not stated.

The SharkWater Take

This is a filing worth noting and not yet a reason to act. A plan set in May takes the sting out of the timing argument, and 0.21 percent of shares outstanding will not move the float. What I want is the Form 4: the real price, and what the holder still owns afterward. Until that prints, I am treating this as a data point, not a signal. If the post-sale stake is small, the read gets worse. If it is a minor trim, this was noise.

Tight lines, SharkWater

Rocket Lab Officer Files to Sell the Same 140,157-Share Block He Sold a Month Ago

noreply@blogger.com (Unknown) — Mon, 05 Oct 2026 12:17:20 +0000

SharkWater Trading • Space Desk • Rocket Lab

Rocket Lab Officer Files to Sell the Same 140,157-Share Block He Sold a Month Ago

October 5, 2026

Bottom Line Up Front

Rocket Lab (Nasdaq: RKLB) officer Adam C Spice filed a Form 144 on October 1, 2026 to sell 140,157 shares, valued at $9,766,140 on the form, through Morgan Stanley. The block is identical in size to his September 2 sale of 140,157 shares ($8,777,725). Both sit under a Rule 10b5-1 plan he adopted on June 3, 2026. The proposed block is about 0.023 percent of 598,350,482 shares outstanding. This is a scheduled sale, not a signal. The useful number is the implied price: roughly $69.68 a share versus roughly $62.62 a month earlier.

What the Form Says

The filer lists his relationship to the issuer as "Officer." The form does not give a title, so I am not assigning one. Shares to be sold: 140,157. Approximate sale date: October 1, 2026. The plan adoption date, June 3, 2026, matters because a plan adopted months ahead removes the "he knows something" read.

Date	Shares	Value
Aug 17	15,718	$1,291,812
Aug 18	30,908	$2,484,248
Aug 24	9,677	$673,790
Sep 2	140,157	$8,777,725
Oct 1 (proposed)	140,157	$9,766,140

Source: Form 144 filed October 1, 2026 by Adam C Spice (SEC EDGAR, primary document). Oct 1 value is the filer's estimated aggregate market value, not an executed price. Past three months of sales as listed on the form.

Divide value by shares and the Oct 1 block prices at about $69.68. The Sep 2 block works out to about $62.62. That is an increase of roughly 11 percent between the two sales. The prices are my arithmetic on the filing's dollar values, not quoted trades.

A 10b5-1 plan is a tide clock. The water comes in when the clock says, not when the captain feels like it.

What I Could Not Check

I could not retrieve the matching Form 4 for the October 1 sale, so the executed price and the post-sale holdings are not verified. I also did not pull Rocket Lab's own newsroom for October. Nothing here says the sale happened at the filed value.

The Bull Case

Pre-committed selling. The plan dates to June 3, 2026. The sale schedule was set months before the October 1 filing.
Small against the float. 140,157 shares is about 0.023 percent of shares outstanding. It does not move the tape.
Still higher. The implied price rose about 11 percent between the two sales, so the plan sold into strength, not weakness.

The Bear Case

It is recurring. Four sales since August 17 total roughly $13.2 million in proceeds on the form. Steady supply is still supply.
Cadence is accelerating. Three small August sales gave way to two identical 140,157-share blocks, a month apart.
Price is the tell. An officer selling at about $69.68 is not pricing in a collapse. That cuts both ways for anyone hoping for a discount.

The SharkWater Take

I pass on this as a trade. A scheduled sale under a plan adopted in June tells me when the officer needs liquidity, not what the company will do. The thing I would watch is the next 140,157-share block in early November. If the size changes or the plan gets amended, that is information. Until then this is housekeeping with a dollar sign on it.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. Filing figures are from SEC EDGAR as of October 5, 2026. Implied prices are arithmetic on those figures. The Form 4 was not retrieved. No options data was pulled or computed. The author may hold positions in securities discussed. Do your own work.

Planet Labs Jumped 8.4 Percent on a Launch and a Press Release That Is Eight Days Old

noreply@blogger.com (Unknown) — Sun, 04 Oct 2026 12:16:54 +0000

SharkWater Trading • Space Desk • Planet Labs

Planet Labs Jumped 8.4 Percent on a Launch and a Press Release That Is Eight Days Old

October 4, 2026

Bottom Line Up Front

Planet Labs (NYSE: PL) closed Friday, October 2, 2026 at $17.50, up 8.43 percent, on 14.85 million shares, versus 7.94 million the session before (StockAnalysis, aggregator, NOT VERIFIED). The likely trigger is a Falcon 9 rideshare that put 20 satellites in orbit, including one built with Google for Project Suncatcher (Motley Fool, October 2, 2026). One outlet credits the Amazon.ia platform instead. Planet's own newsroom dates Amazon.ia to September 24, so that story was eight days old when the stock moved. I found no Planet press release for the launch itself. This is a space-group tailwind with a thin company-specific catalyst. I am not chasing it.

What Moved

PL gained 8.43 percent Friday on nearly double Thursday's volume. Motley Fool had it up 7.8 percent at 11:40 a.m. ET, so the stock held its gains into the close.

Session	Close (change)	Volume (shares)
Sep 30	n/a	10,179,035
Oct 1	n/a	7,944,565
Oct 2	$17.50 (+8.43%)	14,845,020

Source: StockAnalysis price history, an aggregator. NOT VERIFIED against an exchange source. Read October 4, 2026. Sep 30 and Oct 1 closes were not retrieved. The 20-day average volume was not retrieved, so the "twice average" screen is not tested. The price move alone (over 5 percent) trips the flag.

Two Explanations, One Is Stale

Motley Fool (October 2, 2026) says the pop followed the launch of 20 satellites on a SpaceX Transporter mission: 18 SuperDove imagers, one Tanager-2 hyperspectral satellite, and one Project Suncatcher satellite carrying Google Tensor Processing Units to test AI compute in orbit. The piece frames Suncatcher as the potential new revenue stream and says S&P Global Market Intelligence analysts do not expect profitability until 2029.

StocksToTrade (October 2, 2026) leads with something else. It says shares "jumped 7.2 percent after the Amazon.ia launch," a biodiversity AI platform backed by over $14 million from the Bezos Earth Fund, and adds a fleet count of 718 satellites and a 5,700 square meter Berlin factory with capacity for 60 Pelican satellites a year. Check that against Planet's own news page. Amazon.ia is dated September 24, 2026 and the German manufacturing item is dated September 21, 2026. Neither was fresh Friday. The launch was.

A stock that bites on week-old bait is usually reacting to the tide, not the bait.

Planet's newsroom page, as read October 4, 2026, did not yet list a launch announcement. Either the release is posted elsewhere (a wire service) or it has not run. I could not confirm which. I also did not verify the Suncatcher details against Alphabet or Planet. Treat the satellite counts as secondary.

The Bull Case

Real hardware in orbit. Twenty satellites launched Friday, including a Google-linked AI compute test. That is a physical event, not a headline.
Participation. Volume nearly doubled from Thursday, and the stock held most of its intraday gain.
A friendly tape. VIX read 15.31, down 6.59 percent on the day (Cboe, October 2, 2026). Risk appetite is not the constraint.

The Bear Case

The cause is secondhand. Two outlets, two different explanations, one of which rests on a September 24 announcement.
Suncatcher is a test, not revenue. The story that makes the launch exciting has no sourced dollar figure attached.
Profitability is years out. The analyst view Motley Fool cites puts it at 2029. That is a long time to fund at this valuation.

The SharkWater Take

I stay on the sideline. Launches are scheduled events, and a stock that rallies on one that was on the calendar is paying for something the market already knew. Show me the Suncatcher contract value, or an 8-K, and I reopen this. Until then the move is group momentum with a satellite photo attached. I can tell you what the stock did. I cannot tell you the 8.4 percent was earned by anything new.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. Price and volume figures come from an aggregator and are NOT VERIFIED. Cause attribution comes from secondary outlets. EDGAR filings were not fully queryable this run, so the absence of an 8-K is not confirmed. No options data was pulled or computed. The author may hold positions in securities discussed. Do your own work.

Wolfspeed Ran Again. Volume Tripled, and the Reason Is Still Secondhand.

noreply@blogger.com (Unknown) — Sat, 03 Oct 2026 12:17:14 +0000

SharkWater Trading • Semiconductor Desk • Wolfspeed

Wolfspeed Ran Again. Volume Tripled, and the Reason Is Still Secondhand.

October 3, 2026

Bottom Line Up Front

Wolfspeed (NYSE: WOLF) closed Friday, October 2, 2026 at $35.00, up 12.29 percent on 6.81 million shares, about 3.0 times its 20-day average of 2.27 million (the average is dated October 1 and may include that day). Two sessions have now added 24.4 percent from the September 30 close of $28.13 (StockAnalysis, aggregator, NOT VERIFIED). I still found no 8-K and no named customer. Secondary outlets credit a semiconductor rotation and a 200 mm substrate release. The rotation story has a hole in it: Micron fell 2.05 percent the same day. I am still not chasing this.

What Moved

WOLF opened Friday at $32.35, a gap of 3.8 percent over Thursday's $31.17 close. It ran to $36.49, dropped to $31.62, and finished at $35.00, $1.49 under the high. Friday's volume was nearly double Thursday's 3.11 million.

Session	Close	Change	Volume (shares)
Sep 29	$28.13	+1.99%	2.04M
Sep 30	$28.13	flat	4.08M
Oct 1	$31.17	+10.81%	3.11M
Oct 2 (open $32.35, high $36.49, low $31.62)	$35.00	+12.29%	6.81M

Source: StockAnalysis price history, an aggregator. NOT VERIFIED against an exchange or issuer source. Read October 3, 2026.

One data conflict. A Korean aggregator (JKN) reports a Friday close of $36.25, up 16.30 percent. StockAnalysis shows $35.00 and 12.29 percent. I cannot reconcile the two from a primary source, so I use the lower print and flag the gap. A US outlet (Kalkine) published an intraday read of $34.78, up 11.58 percent, which fits StockAnalysis better.

The Explanation Going Around

Kalkine, dated October 2, 2026, names two causes. The first is investors rotating into semiconductors after Micron reported blockbuster results. The second is the company's premium 200 mm silicon carbide substrate release. Neither is a filing. JKN adds "relative value" buying after a long slide.

Check the rotation claim against the tape. Micron moved +3.03 percent on October 1, the session after its results. On October 2 it fell 2.05 percent to $1,074.89. NVIDIA rose 1.34 percent and Broadcom rose 3.35 percent. WOLF rose 12.29 percent. A Micron-led rotation does not explain a Friday move in a stock that made its biggest gain while Micron went the other way. Substrate and silicon carbide are a different supply chain from memory.

When one boat in the fleet is hauling nets and the rest are drifting, the tide is not the reason. Somebody found fish.

The same Kalkine piece carries the cold water. It puts the stock 57 percent below a 52-week high of $80.82, above the analyst targets it cites of $25 to $27.50, with revenue down 24 percent year over year last quarter. Those are third-party figures, not mine, and I have not checked them against the 10-Q.

The Bull Case

Momentum with participation. Two up sessions, the second on about three times average volume.
A sector that is working. NVIDIA and Broadcom both rose Friday, and VIX read 15.31, down 6.59 percent on the day (Cboe, October 2, 2026). Risk appetite is not the constraint.
A concrete product claim. An orderable 200 mm premium substrate with dedicated capacity, per the company's own release.

The Bear Case

Still no primary catalyst. No filing, customer or order size found, and EDGAR was not directly queryable in this run.
A 24 percent two-day run. The stock now sits above the sell-side targets cited by the outlet that explained the move.
The cause does not fit the tape. The rotation story depends on Micron, and Micron fell on the day WOLF made its largest gain.

The SharkWater Take

I stay out. A stock that jumps 24 percent in two sessions on a story the tape contradicts has not earned a position from me. If a filing or a named customer surfaces, I reopen the file. If the stock gives back the gap with no news, that tells me the buyers had no story either. I can say what the chart did. I cannot tell you why, and neither can anyone quoting a rotation that Micron sat out.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. Price and volume figures come from an aggregator and are NOT VERIFIED. Cause attribution comes from secondary outlets and is unconfirmed. EDGAR was not directly queryable this run, so the absence of a filing is not confirmed. VIX is from the Cboe site as read on October 3, 2026. No options data was pulled or computed. The author may hold positions in securities discussed. Do your own work.

Wolfspeed Jumped 10.8 Percent. I Cannot Find the Reason.

noreply@blogger.com (Unknown) — Fri, 02 Oct 2026 12:18:56 +0000

SharkWater Trading • Semiconductor Desk • Tape Without a Story

Wolfspeed Jumped 10.8 Percent. I Cannot Find the Reason.

October 2, 2026

Bottom Line Up Front

Wolfspeed (NYSE: WOLF) closed October 1, 2026 at $31.17, up 10.81 percent from $28.13, on 3.11 million shares against a 20-day average of 2.27 million, about 1.4 times (StockAnalysis, aggregator, NOT VERIFIED). I found no 8-K, no earnings event and no dated company release inside that window. The only recent company item I could read is a product release on its own site, dated September 27, 2026 as the page shows it. A move this size with no sourced cause is a tape signal, not a thesis. I am not chasing it.

What Moved

WOLF opened October 1 at $28.72, a gap of about 2.1 percent over the prior close. It ran to a high of $31.39, touched a low of $28.03, and closed at $31.17, within 22 cents of the high. After-hours showed $31.31. The same page listed no news dated October 1 or 2.

Session	Close	Change	Volume (shares)
Sep 29	$28.13	+1.99%	2.04M
Sep 30	$28.13	not shown	4.08M
Oct 1 (open $28.72, high $31.39, low $28.03)	$31.17	+10.81%	3.11M

Source: StockAnalysis price history, an aggregator. NOT VERIFIED against an exchange or issuer source. Read on October 2, 2026.

Volume ran ahead of price. September 30 traded 4.08 million shares, about 1.8 times the same average, while the stock closed flat at $28.13. The change cell for that day was blank on the source page. Someone was active a session early. I cannot say who.

The Only Company Item

Wolfspeed's own site carries a release headlined "Wolfspeed Expands 200 mm Silicon Carbide Portfolio with New Premium Power Substrates." The page shows a September 27, 2026 date, which is a Sunday, so I treat the date as unconfirmed. The company says the substrate has fewer than 0.01 micropipes per square centimeter, is orderable immediately, and comes with dedicated capacity. Those are the company's own claims. The release names no customer, no order size and no revenue figure.

A product spec sheet moved a stock 10.8 percent four sessions after it was posted? I do not buy that as the whole story. It may be part of it.

Baitfish breaking the surface tell you something is feeding below. They do not tell you what, and they do not tell you if it is still there.

The Bull Case

Strength held into the close. The stock finished within 22 cents of its high and showed a higher print after hours.
Volume confirmed. Volume exceeded the average on the day and on the session before it.
A tangible product. An orderable 200 mm premium substrate with dedicated capacity is a concrete commercial step, as the company describes it.

The Bear Case

No sourced catalyst. I found no filing, earnings event or named customer behind the move.
Volume was modest. About 1.4 times average is not a capitulation or a squeeze signature, and the average may already include October 1.
Unexplained gaps retrace. A move with no story gives buyers nothing to hold onto when the next seller shows up.

The SharkWater Take

I stay out. I trade stories I can source, and this one has a chart and a mystery. If a primary filing or a named customer shows up, I reopen the file. Until then the question is not whether WOLF can run further. It is who was buying on September 30 and why they were early. I cannot answer that from public pages, and neither can the aggregators that will explain it to you by tonight.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. Price and volume figures come from an aggregator and are NOT VERIFIED. The company release was read through an automated page reader. EDGAR was not directly queryable this run, so the absence of a filing is not confirmed. No options data was pulled or computed. The author may hold positions in securities discussed. Do your own work.

Micron Printed $54 Billion. The Pre-Market Shrugged.

noreply@blogger.com (Unknown) — Thu, 01 Oct 2026 12:17:21 +0000

SharkWater Trading • Semiconductor Desk • Earnings

Micron Printed $54 Billion. The Pre-Market Shrugged.

October 1, 2026

Bottom Line Up Front

Micron (Nasdaq: MU) reported fiscal fourth-quarter revenue of $54.23 billion on September 30, 2026, against $11.32 billion a year earlier and $41.46 billion in the prior quarter (Form 8-K, Exhibit 99.1). It guided fiscal first-quarter 2027 revenue to $61.5 billion, plus or minus $1.5 billion, with non-GAAP gross margin near 86.25 percent versus 87.0 percent printed. The stock closed September 30 at $1,065.11, before the release, and an aggregator showed it about 0.6 percent lower pre-market on October 1. A quadrupling of revenue produced a shrug so far. That tells you what was in the price.

The Print

Micron reported after the close on September 30. Fiscal fourth-quarter revenue of $54.23 billion is 4.8 times the year-ago quarter and 30.8 percent above the prior quarter. GAAP net income was $37.70 billion, or $32.87 per diluted share. Non-GAAP diluted EPS was $33.42. The board declared a quarterly dividend of $0.15 per share, payable October 29, 2026.

Metric	FQ4 FY25	FQ3 FY26	FQ4 FY26	FQ1 FY27 guide
Revenue	$11.32B	$41.46B	$54.23B	$61.5B +/- $1.5B
GAAP diluted EPS	NOT IN EXCERPT	NOT IN EXCERPT	$32.87	$37.84 +/- $1.00
Non-GAAP diluted EPS	NOT IN EXCERPT	NOT IN EXCERPT	$33.42	$38.15 +/- $1.00
Non-GAAP gross margin	NOT IN EXCERPT	NOT IN EXCERPT	87.0%	~86.25%
GAAP gross margin	NOT IN EXCERPT	NOT IN EXCERPT	86.8%	~85.95%

Source: Micron Technology Form 8-K, Exhibit 99.1 press release dated September 30, 2026 (sec.gov). Figures were extracted from the filing by an automated page reader, not hand-checked line by line. Cells marked NOT IN EXCERPT were not returned by the read.

The Guide Is the Story

The guided midpoint of $61.5 billion is 13.4 percent above the quarter just reported. Sequential growth was 30.8 percent in the quarter that closed and is guided to less than half that pace. The guided non-GAAP gross margin of about 86.25 percent sits 0.75 points below the 87.0 percent printed. Both are my arithmetic on the filed figures. Neither is a miss against a consensus number, because I did not pull a sourced consensus and will not quote one.

Price action (StockAnalysis, aggregator, NOT VERIFIED): MU closed September 30 at $1,065.11, before the release, on 30.8 million shares, against a 31.3 million-share 20-day average read on September 29, about 0.98x. Pre-market on October 1 it showed $1,059.00, down 0.57 percent. The same page said Nvidia, Broadcom, AMD and Intel traded higher pre-market. I did not verify those moves.

A boat that has already caught the limit does not get more fish from a bigger net. It gets a heavier haul to carry in.

The Bull Case

Demand visibility. The release credits strategic customer agreements for demand visibility, and the guide calls for another $7 billion or so of quarterly revenue.
Cash returns exist. A dividend is declared and a $650 million buyback figure appears for the fiscal year. Neither is large against $37.70 billion of quarterly GAAP net income.
The reaction was muted, not negative. A pre-market dip of about half a percent after a print this large is not a sell-the-news rout. The cash session is still untested.

The Bear Case

Deceleration is guided. Sequential growth is guided from 30.8 percent to about 13.4 percent at the midpoint.
Margin rolls over. A guided gross margin below the printed level is the first step down from a peak, if it is one.
Capex is heavy. Net capital spending was $10.77 billion in the quarter and $27.37 billion for fiscal 2026, per the release as read. Memory is a cyclical business, and cycles end when supply arrives.

The SharkWater Take

I would not chase this. A stock that shrugs at a 4.8x revenue print has already priced the cycle at its widest point, and the guide shows the rate of change slowing even while the level rises. The market trades the derivative, not the level. What I am watching is whether the pre-market dip holds or deepens into the cash open and whether the semiconductor peers keep their gains without Micron. If they do, the read-through to AI hardware is intact. If they fade, Micron was the top of the chain. I stand aside until the price tells me which.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. Earnings figures come from Micron's Form 8-K Exhibit 99.1 dated September 30, 2026 and were machine-extracted. Price and volume figures come from an aggregator and are NOT VERIFIED. No options data was pulled or computed for this post. The author may hold positions in securities discussed. Do your own work.

Starship Reached Orbit. All Six Space Stocks Fell Anyway.

noreply@blogger.com (Unknown) — Wed, 30 Sep 2026 12:17:50 +0000

SharkWater Trading • Space Desk • Sector Tape

Starship Reached Orbit. All Six Space Stocks Fell Anyway.

September 30, 2026

Bottom Line Up Front

SpaceX flew Starship Flight 14 on September 28 and, per a secondary mission log, put 26 Starlink V3 satellites in low Earth orbit. Rocket Lab (Nasdaq: RKLB), AST SpaceMobile (Nasdaq: ASTS), Intuitive Machines (Nasdaq: LUNR), Planet Labs (NYSE: PL), Firefly Aerospace (Nasdaq: FLY) and Redwire (NYSE: RDW) then fell on both the September 28 and September 29 sessions, down 3.9 percent to 7.7 percent over the two days. Volume ran from 0.83x to 1.46x of the 20-day average, so this was drift, not a stampede. No sourced cause exists yet, and I will not invent one.

The Tape

Every name on the space list closed lower on both sessions after the flight. RDW took the biggest two-day hit. ASTS took the smallest, but it also had the strangest session: it opened September 29 at $65.10, a 6.7 percent gap above the $61.00 prior close, then sold off to close at $59.40 on 15.9 million shares, about 1.46x its 20-day average of 10.9 million.

Ticker	Sep 25 close	Sep 29 close	Two-day	Sep 29	Vol vs 20-day avg
RKLB	$73.95	$69.70	-5.75%	-3.45%	1.07x
ASTS	$61.81	$59.40	-3.90%	-2.62%	1.46x
LUNR	$15.61	$14.55	-6.79%	-4.28%	0.83x
PL	$17.43	$16.38	-6.02%	-2.33%	1.45x
FLY	$24.36	$22.86	-6.16%	-5.18%	1.06x
RDW	$11.62	$10.73	-7.66%	-4.20%	0.94x

Source: StockAnalysis.com daily history, Sep 25 and Sep 29, 2026 sessions (S&P Global Market Intelligence feed). Aggregator figures, NOT VERIFIED against an exchange or issuer source. Two-day change and volume ratio computed by SharkWater from those rows; average volume is the mean of the 20 sessions before Sep 29 as read from the same page.

What Is Confirmed and What Is Not

Confirmed by a mission log (rocketlaunch.org, secondary, September 28, 2026): Flight 14 launched at 12:48 UTC, Ship 41 made a soft Pacific splashdown, the booster completed its landing burn with a targeted Gulf of Mexico ocean landing, and this was the first Starship flight to deploy satellites. An earlier account reported one of six ship engines shutting down on ascent; the mission log documents no issues. That conflict is unresolved, and I did not reach SpaceX's own account. NOT VERIFIED against a primary source.

Not found: any dated catalyst for the selloff in the six names on either day. No watchlist company filed or released anything I could tie to the decline. FLY was the only name to fall more than 5 percent on a single session (-5.18 percent on September 29, on 1.06x average volume).

Why a Win Can Look Like a Threat

This part is inference, not sourced fact. The space group trades partly on the idea that heavy launch capacity is scarce. A Starship that reaches orbit and deploys a full Starlink V3 batch is a step toward cheap, abundant lift and a bigger direct-to-device competitor. That reading is bad for launch names (RKLB, FLY) and for ASTS, whose business is competing with Starlink. It is a stretch for imaging (PL) and lunar landers (LUNR), which fell just as hard. The uniformity of the drop points to basket behavior, not six separate stories.

When the biggest boat in the harbor learns to dock itself, the charter captains do not cheer. They check the price of their own slip.

The Bull Case

Orderly, not forced. Volume ratios of 0.83x to 1.46x show no capitulation. A drift on ordinary volume is what a fade after a well-anticipated event looks like.
A working Starship helps the whole ecosystem. More launches, more Starlink capacity and more demand for orbital services, imaging and lunar logistics is one way to read the flight. That case needs a source from the companies, and none appeared.
Reversal potential. ASTS opened 6.7 percent higher before selling off. Buyers showed up at the open. Whether they return is untested.

The Bear Case

Competitive read-through. If Starship lift gets cheap, launch pricing power across the group is questioned, and Starlink V3 raises the bar for direct-to-cell rivals.
Two down sessions, no bounce. Six of six lower twice in a row, with ASTS failing at a gap-up, is a market choosing to sell strength.
No catalyst ahead is confirmed. I found no dated company event in the next session that resets the narrative. Absent one, the group trades on flows.

The SharkWater Take

I would not build a trade on this tape. The move is real, but the cause is a guess, and a guess is not an edge. What I can say is that a sector-defining flight produced no positive read-through, which tells me the good news was already in the price. The tell I am watching is ASTS: a 6.7 percent gap that fully reversed on 1.46x volume is the ugliest single print in the group. If the basket stops falling while ASTS keeps failing at the open, the weak link is showing itself. Until a sourced catalyst appears, I stand aside.

Tight lines. SharkWater

Educational and informational purposes only. Not personalized investment advice. All figures sourced as noted and dated. Price and volume figures come from an aggregator and are NOT VERIFIED against primary sources. No options data was pulled or computed for this post. The author may hold positions in securities discussed. Do your own work.

Nvidia's Record $150 Billion Buyback Is a Ceiling, Not a Purchase

noreply@blogger.com (Unknown) — Tue, 29 Sep 2026 12:21:12 +0000

SharkWater Trading • Semis Desk • NVDA Capital Return

Nvidia's Record $150 Billion Buyback Is a Ceiling, Not a Purchase

September 29, 2026

Bottom Line Up Front

On September 28, 2026, NVIDIA (Nasdaq: NVDA) raised its share repurchase authorization by $150 billion, taking the total remaining to $235 billion, to be executed through fiscal 2028 (company press release, September 28, 2026). That implies about $85 billion was left before the increase (SharkWater arithmetic: $235 billion minus $150 billion). An authorization is permission, not a trade. The release gives no purchase pace, so the number that matters, dollars actually retired per quarter, is not yet public.

What NVIDIA Actually Said

The release is short. It states the size of the increase, the new remaining total and the execution window. It does not state a quarterly or annual pace, a funding source or a price limit. CEO Jensen Huang is quoted tying the company's growth to what he calls a once-in-a-generation platform shift to AI and accelerated computing.

Item	Figure	Basis
Authorization increase	$150B	NVIDIA release, Sep 28, 2026
Total remaining after increase	$235B	NVIDIA release, Sep 28, 2026
Implied remaining before increase	~$85B	SharkWater arithmetic, not stated by NVIDIA
Execution window	Through fiscal 2028	NVIDIA release, Sep 28, 2026
"Largest increase in history" claim	Company language	NVIDIA release; not independently checked

Source: NVIDIA Newsroom, "NVIDIA Announces a $150 Billion Share Repurchase Authorization Increase," September 28, 2026. No 8-K was confirmed for this item because SEC full-text search was unreachable during this scan.

Authorization Is Not Execution

Companies announce buyback authorizations they never finish. The board approves a ceiling. Management buys when cash and conviction line up. Nothing in a press release forces a single share to be retired.

A buyback authorization is a fishing license, not a full cooler. The license tells you how many fish the law allows. Only the log book tells you how many came aboard.

The log book here is the repurchase table in NVIDIA's next 10-Q. Until it prints, the pace is unknown. For scale only: if the full $235 billion were used evenly between now and the end of fiscal 2028 (roughly late January 2028, about 16 months), that works out to about $15 billion a month. That is illustrative arithmetic, not a forecast, and NVIDIA has not said it will buy at that rate.

What Is Not Verified

Four figures circulating in coverage are aggregator-only. This desk did not use them as fact.

Cash on hand. One aggregator (investingLive, September 28, 2026) cites about $22 billion in cash and equivalents at the end of the July quarter. NOT VERIFIED against the 10-Q. Cash alone would not fund $235 billion, so free cash flow and marketable securities matter more than that line.
Prior authorization. The same source says an $80 billion buyback was announced in May. NOT VERIFIED. It does not reconcile with the $85 billion implied remainder without knowing how much was spent since.
Stock reaction. Monday, September 28 gains for NVDA are reported as roughly 1.7 percent to nearly 3 percent depending on the outlet. Close and volume NOT VERIFIED.
Valuation. A reported forward multiple of 16.5 times is NOT VERIFIED and depends on whose earnings estimate sits in the denominator.

Context for the same session, also from aggregators: the S&P 500 closed down about 0.8 percent, semiconductors fell more than 2 percent, and Micron (Nasdaq: MU) and AMD fell while NVDA rose. A buyback that lifts a stock on a day the group sells off is a signal about sponsorship. It is not proof of a floor.

The Bull Case

Management is signaling the shares are cheap. A board does not authorize $150 billion more if it expects the stock to look expensive next year.
Persistent bid. Even partial use creates a steady buyer in a tape where semis are otherwise being sold.
Capital return without cutting growth spend. The size implies cash generation large enough to fund both, which is the company's own framing.

The Bear Case

No pace, no commitment. The release states a ceiling and a window. It never states a purchase rate.
A buyback can mean fewer places to spend. Some analysts read large authorizations as a sign management sees slower profit growth ahead. That reading is coverage commentary, not an NVIDIA statement.
Macro is the bigger driver right now. Treasury yields at multi-year highs (aggregator-reported, NOT VERIFIED) and a Micron earnings print after the close on September 30 can swamp a buyback headline.

The SharkWater Take

I am not treating this as a catalyst yet. The headline number is real and it is a vote of confidence from the people who see the orders. But it is a ceiling with no stated speed, and I will not pay for a ceiling. The proof shows up in the repurchase table of the next 10-Q, where dollars retired per quarter either match the ambition or they do not. Until then the buyback is background support for the stock, not a reason to lean on it.

Execution Notes

There is no trade here to build. No options chain was pulled for this post, so no strikes, premiums or expirations are given. If you are watching volatility into the Micron report on September 30 after the close, one aggregator (Benzinga) reports an implied move of about 8 percent. That figure is NOT VERIFIED against a live chain. Pull the chain from your broker. What invalidates the "quiet support" read: a 10-Q showing repurchases well below the pace this authorization implies.

Tight lines. SharkWater

CORRECTED: SharkWater Pre-Market Scan — September 28, 2026

noreply@blogger.com (Unknown) — Mon, 28 Sep 2026 12:29:28 +0000

SharkWater Trading • Desk Notes • Pre-Market Scan

SharkWater Pre-Market Scan — September 28, 2026

September 28, 2026

Bottom Line Up Front

No material company-specific developments surfaced this window. No fresh 8-K, 424B5, S-1, 13D, or 13G was found for any watchlist ticker since Friday's close, and no distribution or NAV event was found for the six income and pre-IPO funds checked. The day's only scheduled catalyst is SpaceX's Starship Flight 14, its first orbital launch attempt, targeting an 8:49am ET liftoff; the outcome was not known as of this scan. Nothing here clears the bar for a standalone post today.

Correction to the record

The "unpublished backlog" flagged in the September 20, 22, and 27 scans was a false alarm. Those scans searched Gmail's default inbox view, which excludes Trash, and concluded nothing had reached Blogger since September 17. Checking Trash this run, and confirming directly against sharkwatertrading.com, shows every post from September 18 through September 27 was in fact sent and is live on the blog, correctly dated through the September 27 TeraWulf correction. Nothing needs to be resent. No decision is needed from you on this.

Publishing note, this run

A diagnostic test this run found that Gmail's send tool silently strips every CSS background and background-color declaration from outbound HTML, on every send, with no exception found. That means every SharkWater post sent through this pipeline, back to the first one, has gone out and landed on Blogger without the intended dark navy panel backgrounds behind the light-colored text. This email uses table bgcolor attributes instead, which survive the strip, as a fix. Whether the prior posts actually read as light-on-light on the live blog depends on Blogger's own theme background, which was not checked this run; if you notice any past post looking washed out or hard to read, that is almost certainly this.

Actionable

None qualify this run. Checked and came up empty: fresh filings (8-K, 424B5, S-1, 13D, 13G) for all twenty-six watchlist tickers since Friday, September 25's close; distribution or NAV events for RVI, DXYZ, VCX, ARKVX, THTA, and AGNC; and any confirmed post-Friday price move, since no new session has closed between then and this scan.

Monitor

Space sector, forward-looking. SpaceX is attempting Starship's first-ever orbital flight this morning, an 8:15–9:30am ET window targeting 8:49am ET liftoff (SpaceX's own account and Space.com's live coverage; outcome unknown as of this scan). A clean result or a public failure could move sentiment across RKLB, ASTS, LUNR, PL, RDW, and FLY today with no company-specific news attached; none of those tickers has been confirmed as a stated payload or hardware partner on this mission.
Macro, secondary sourcing. S&P 500 futures were down roughly 0.6 percent this morning on a reported US rejection of an Iranian proposal to reopen the Strait of Hormuz, with Brent crude reported near $107 a barrel (Investrade's morning note, secondary; not checked against a primary wire or exchange settlement).
Fundrise Innovation Fund II, LLC filed an N-2/A on September 25, 2026 (EDGAR full text search, primary for the filing's existence, document not read this run). This is a separate legal entity from VCX, the Fundrise vehicle already on the watchlist. Whether it competes for the same investor base as VCX, RVI, DXYZ, and ARKVX, and what it means for VCX's NAV premium, was not traced this run.
NVDA, secondary and not filing-based. Nvidia reportedly introduced a new hardware security architecture and may be permitted to sell its RTX Pro 5500 chip in China (Investrade's morning note, secondary; no primary Nvidia newsroom item checked this run).
AVGO / OKLO / XE, unchanged. The FT-reported SASAC review of Broadcom's China switching-hardware dependency remains open and secondary-sourced only. Oklo's newsroom URL 404'd again this run; UBS's prior price-target cut and X-Energy's leadership change are unchanged.

Noise

Checked against EDGAR this window, no new filing found, unless noted.

ASTS — no new filing; press page not re-checked (prior JS-rendering issue).
RKLB — no new filing since the Sept 25 Electron launch already logged.
LUNR — no change.
PL — no change.
FLY — no change. Reconfirmed as Firefly Aerospace Inc., not the legacy Fly Leasing entity.
RDW — no change.
RVI — no change. Reconfirmed as Robinhood Ventures Fund I, not the legacy Retail Value Inc. REIT.
DXYZ — no change (checked against Destiny Tech100).
VCX — no new filing under Fundrise Innovation Fund itself; see Monitor for the related Fund II item.
ARKVX — no change (checked against ARK Venture Fund).
AGNC — no change. September's $0.12 dividend was already declared September 10, ahead of this window.
ORC — no change. September's dividend was already declared September 14, ahead of this window.
USA — no change (checked against Liberty All-Star Equity Fund).
THTA — no change. A 10.00 percent distribution rate was already announced September 14, ahead of this window.
IREN — no change.
NBIS — no change.
KEEL — no change. Reconfirmed as Keel Infrastructure Corp.
CBRS — no change. Ticker remains Cerebras Systems Inc.
WOLF — no change.
NVDA — no new SEC filing; see Monitor for a secondary item.
MU — no change.
AVGO — no new filing; see Monitor for the carried-over SASAC item.
SMR — no change.

Data Gaps

No exact, primary-sourced VIX print for today exists in this scan; only secondary references to a risk-off tone were found.
Oklo's newsroom URL 404'd again this run, same failure as September 27; its press page remains unconfirmed clean of any weekend or Monday-morning announcement.
AST SpaceMobile's own investor press-release page was not re-checked this run (prior JavaScript-rendering issue).
The relationship between "Fundrise Innovation Fund," "Fundrise Growth Tech Fund, LLC" (VCX's issuer name per Yahoo Finance), and "Fundrise Innovation Fund II, LLC" (the entity that filed the September 25 N-2/A) was not resolved. This needs a primary-source read before anything is written about VCX tied to that filing.
Whether any watchlist ticker is a stated payload, hardware, or ground-systems partner on today's Starship Flight 14 mission was not verified.
Liberty All-Star Equity Fund's (USA) most recent distribution declaration date was not independently re-verified this run.
The Fundrise Innovation Fund II N-2/A's actual content was not read this run, only its existence and filing date via EDGAR metadata.
Whether Blogger's own theme background is dark or light, which determines how badly the background-stripping bug above has affected past posts' readability, was not checked this run.

Tight lines. — SharkWater

SharkWater Trading pre-market scan. Educational and informational purposes only, not personalized investment advice. All figures sourced as noted and accurate as of publication.

SharkWater Pre-Market Scan — September 28, 2026

noreply@blogger.com (Unknown) — Mon, 28 Sep 2026 12:25:23 +0000

SharkWater Trading • Desk Notes • Pre-Market Scan

SharkWater Pre-Market Scan — September 28, 2026

September 28, 2026

Bottom Line Up Front

Correction to the record

Actionable

Monitor

Space sector, forward-looking. SpaceX is attempting Starship's first-ever orbital flight this morning, an 8:15–9:30am ET window targeting 8:49am ET liftoff (SpaceX's own account and Space.com's live coverage; outcome unknown as of this scan). A clean result or a public failure could move sentiment across RKLB, ASTS, LUNR, PL, RDW, and FLY today with no company-specific news attached; none of those tickers has been confirmed as a stated payload or hardware partner on this mission.
Macro, secondary sourcing. S&P 500 futures were down roughly 0.6 percent this morning on a reported US rejection of an Iranian proposal to reopen the Strait of Hormuz, with Brent crude reported near $107 a barrel (Investrade's morning note, secondary; not checked against a primary wire or exchange settlement).
Fundrise Innovation Fund II, LLC filed an N-2/A on September 25, 2026 (EDGAR full text search, primary for the filing's existence, document not read this run). This is a separate legal entity from VCX, the Fundrise vehicle already on the watchlist. Whether it competes for the same investor base as VCX, RVI, DXYZ, and ARKVX, and what it means for VCX's NAV premium, was not traced this run.
NVDA, secondary and not filing-based. Nvidia reportedly introduced a new hardware security architecture and may be permitted to sell its RTX Pro 5500 chip in China (Investrade's morning note, secondary; no primary Nvidia newsroom item checked this run).
AVGO / OKLO / XE, unchanged. The FT-reported SASAC review of Broadcom's China switching-hardware dependency remains open and secondary-sourced only. Oklo's newsroom URL 404'd again this run; UBS's prior price-target cut and X-Energy's leadership change are unchanged.

Noise

Checked against EDGAR this window, no new filing found, unless noted.

Data Gaps

No exact, primary-sourced VIX print for today exists in this scan; only secondary references to a risk-off tone were found.
Oklo's newsroom URL 404'd again this run, same failure as September 27; its press page remains unconfirmed clean of any weekend or Monday-morning announcement.
AST SpaceMobile's own investor press-release page was not re-checked this run (prior JavaScript-rendering issue).
The relationship between "Fundrise Innovation Fund," "Fundrise Growth Tech Fund, LLC" (VCX's issuer name per Yahoo Finance), and "Fundrise Innovation Fund II, LLC" (the entity that filed the September 25 N-2/A) was not resolved. This needs a primary-source read before anything is written about VCX tied to that filing.
Whether any watchlist ticker is a stated payload, hardware, or ground-systems partner on today's Starship Flight 14 mission was not verified.
Liberty All-Star Equity Fund's (USA) most recent distribution declaration date was not independently re-verified this run.
The Fundrise Innovation Fund II N-2/A's actual content was not read this run, only its existence and filing date via EDGAR metadata.

Tight lines. — SharkWater

SharkWater Trading pre-market scan. Educational and informational purposes only, not personalized investment advice. All figures sourced as noted and accurate as of publication.

CORRECTED: Google Doesn't Own TeraWulf Stock. It Owns 41 Million Warrants Struck at a Penny.

noreply@blogger.com (Unknown) — Sun, 27 Sep 2026 12:22:51 +0000

SharkWater Trading • Data Center Desk • Filings

Google Doesn't Own TeraWulf Stock. It Owns 41 Million Warrants Struck at a Penny.

September 27, 2026

Bottom Line Up Front

A Schedule 13G filed September 25, 2026 shows Google, through Alphabet and XXVI Holdings, beneficially owns 41,011,803 shares of TeraWulf (NASDAQ: WULF), or 7.6 percent of the class, entirely through warrants exercisable at $0.01 a share. Those warrants were not bought. They were issued as collateral for Google's guarantee of Fluidstack's lease obligations on the CB-5 data center at Lake Mariner, first disclosed in an August 2025 8-K at 32,568,197 shares and now roughly 8.4 million shares larger. WULF closed Friday, September 25 at $15.74, down 3.38 percent, as the broader neocloud group (IREN, NBIS, KEEL) sold off together. This is a credit-support arrangement wearing the market's favorite AI-partnership costume.

What the Filing Actually Says

The Schedule 13G names three reporting persons: Google LLC, XXVI Holdings Inc., and Alphabet Inc., the parent chain that owns and directs Google. All three certify a passive stake, meaning the securities were not acquired to influence control of the company. The 41,011,803 shares are described as issuable upon exercise of warrants that are currently exercisable or exercisable within 60 days. There is no open-market purchase anywhere in this document.

Go back to the source of those warrants and the picture sharpens. An August 14, 2025 Form 8-K discloses that Google received 32,568,197 warrant shares at a $0.01 strike price in exchange for guaranteeing Fluidstack's lease payments under the CB-5 lease at TeraWulf's Lake Mariner site. The warrants themselves were pledged to CB-5's construction lenders as security, released only once the guarantee takes effect at lease commencement. Google was not writing a check for equity. It was putting its balance sheet behind Fluidstack's rent so TeraWulf could get the CB-5 buildout financed.

A bank co-signing your mortgage isn't the same as a bank buying your house. It gets you the loan. It also means the bank's name is now on the paper, and if you miss a payment, the bank's exposure is the thing that gets called first.

The Stake Has Grown, and This Desk Can't Fully Explain Why Yet

The gap between the 32,568,197 shares in the August 2025 8-K and the 41,011,803 shares in this month's 13G is about 8.4 million shares. TeraWulf announced a 160 MW CB-5 lease expansion with Fluidstack in the interim, and the growth in the warrant count is consistent with additional collateral tied to that expansion. This desk has not confirmed that link against a specific filing, and is not asserting it as fact. It's the open thread worth pulling before treating this as a settled story.

Date	Filing	Warrant Shares	Strike
Aug 14, 2025	Form 8-K	32,568,197	$0.01
Sept 25, 2026	Schedule 13G	41,011,803	$0.01 (per originating 8-K)

Source: SEC EDGAR, Form 8-K filed Aug 14, 2025, and Schedule 13G filed Sept 25, 2026, both read via TeraWulf's own investor-relations filing mirror. TeraWulf's total shares outstanding, approximately 498.97 million, is aggregator-sourced (stockanalysis.com) and not independently confirmed against the company's own most recent 10-Q cover page.

The Bull Case

Google's credit is doing real work. A guarantee from a company with Alphabet's balance sheet is what let TeraWulf and Fluidstack get CB-5's construction financed on favorable terms, without TeraWulf issuing dilutive cash equity to fund the buildout directly.
The relationship is expanding, not shrinking. A larger warrant block points to a larger backstop, which points to a larger footprint at Lake Mariner. That's a vote of confidence in the underlying AI-hosting demand, even if it isn't a vote of confidence expressed through open-market stock purchases.
Warrants at $0.01 pledged as loan collateral are not the same as an active seller. Nothing in this filing shows Google distributing or selling shares. The overhang is theoretical until exercised.

The Bear Case

41 million shares at a penny is functionally free stock. If those warrants are ever exercised in full, the dilution hits per-share value directly, and the cost basis to Google is close to nothing.
A guarantee exists because the underlying credit needed one. Fluidstack's lease obligations required Google to co-sign in the first place. That is a data point about the counterparty risk in the deal, not just a marketing headline about "Google backing TeraWulf."
It landed inside a real group-wide selloff. IREN, NBIS, WULF, and KEEL all closed lower Friday. Layering a growing, penny-strike dilution overhang onto a sector already digesting a confirmed Nebius GPU-rental price hike and a Sell rating on a neocloud peer is not a coincidence this desk is ready to wave off.

The SharkWater Take

I don't read this as a bullish Google-backs-TeraWulf headline, and I don't read it as a bearish Google-is-dumping-shares headline either, because neither is what the filing says. What it says is that TeraWulf's biggest AI-hosting partner is extending credit, not cash, and getting paid in cheap dilution for the privilege. That's a normal, sensible way to finance a data center buildout. It is not, on its own, a reason to pay up for the stock on the theory that Google's name is now attached to it. Until I can trace exactly which expansion agreement accounts for the extra 8.4 million warrant shares, I'm treating this as a data point inside the broader neocloud pullback, not a standalone thesis. I am not taking a position here off this filing alone.

Tight lines. — SharkWater

Samueli's $250 Million Broadcom Sale Cleared. The Stock Didn't Blink.

noreply@blogger.com (Unknown) — Sat, 26 Sep 2026 13:46:40 +0000

SharkWater Trading • Insider Desk • Broadcom • Form 4

Samueli's $250 Million Broadcom Sale Cleared. The Stock Didn't Blink.

September 26, 2026

Bottom Line Up Front

Broadcom (NASDAQ: AVGO) director Henry Samueli sold 702,190 shares on September 23, 2026 at an average price of $356.04, for proceeds of $250,006,184, confirmed by a Form 4 filed September 25. He also gave away a further 72,474 shares with no consideration attached. He still holds 52,680,018 shares directly, so the disposal is roughly 1.5 percent of his stake.

AVGO closed September 25 at $352.81, up 0.70 percent from Thursday's $350.36. A quarter-billion-dollar insider sale moved the stock less than a rounding error. The more relevant overhang right now isn't this filing at all, it's an unresolved, reportedly ongoing Chinese state review of how dependent its data centers are on Broadcom's switching chips.

The Sale, Confirmed

This desk flagged the setup two days before it resolved. On September 23, 2026, two related filers put the market on notice: D95GT, LLC, managed by Lapland Longspur LLC, filed a Form 144 for 631,972 shares valued at $225,004,972.58, and the Samueli Foundation filed one for $25,797,598.63. A Form 144 is a notice of intent to sell, not proof a sale happened. As of Wednesday's coverage, no Form 4 had posted confirming any of it traded.

That changed Friday. Henry Samueli's own Form 4, filed September 25 at 4:15pm ET, shows the transaction dated September 23: 702,190 shares sold at $356.04 average, $250,006,184 total, plus 72,474 shares disposed by gift. The gift carries no cash consideration, the kind of transfer that shows up in estate and foundation planning rather than in a liquidity scramble.

One honest caveat: the Form 144 totals from the two entities above (roughly $250.8 million combined) don't reconcile cleanly, share for share, against the Form 4's sold-plus-gifted total. Multiple Samueli-affiliated vehicles appear to be involved. I'm not going to force a clean story out of numbers that don't quite line up. What's confirmed is the Form 4 itself: the shares moved, priced, and are on the books.

Filer	Type	Shares	Value / Price
D95GT, LLC (Lapland Longspur LLC)	Form 144, filed 9/23	631,972	$225,004,972.58
Samueli Foundation	Form 144, filed 9/23	N/A	$25,797,598.63
Henry Samueli (director)	Form 4, sale, filed 9/25	702,190	avg $356.04 ($250,006,184)
Henry Samueli (director)	Form 4, gift, filed 9/25	72,474	no consideration

Source: SEC EDGAR, Form 144 and Form 4 filings, September 23 and 25, 2026. Form 144/Form 4 totals across the listed entities do not fully reconcile against each other; treat as related but not confirmed identical transactions.

A sale this size usually throws a wake big enough to rock anything nearby. This one barely rippled the surface, which tells you either the water underneath is calmer than the headline number suggests, or the market's watching a different current entirely.

What The Market Is Actually Pricing

Broadcom's stock has a bigger, unresolved question sitting on it, and this filing isn't it. The Financial Times has reported, and multiple secondary outlets have since repeated, that Chinese authorities are reviewing how dependent state-owned data centers are on Broadcom networking-switch hardware, with one figure cited at up to 90 percent of surveyed state-owned deployments. As of this writing, that review is still at the review stage. No primary Chinese-government document has surfaced, and no confirmed policy action, a ban, a directive, anything binding, has been found.

That's the asymmetry. Samueli's sale was signaled in advance, sized small against his remaining stake, and absorbed without a ripple. The SASAC review has had no such advance notice, no size anyone can put a number on yet, and the kind of binary outcome that moves a stock on a single headline whenever it does resolve.

The Bull Case

No surprise, no new information. The Form 144 notices were public Sept 23. By the time the Form 4 posted, the market had two full sessions to price it in, and did essentially nothing.
It's a small slice of a large stake. Against 52.68 million shares still held directly, 1.5 percent moving reads as routine diversification, not an exit.
The gift component points to planning, not distress. 72,474 shares moved for no consideration is estate and foundation mechanics, not a signal about the business.

The Bear Case

A quarter-billion dollars is still a quarter-billion dollars. Co-founder selling at this scale is worth logging even when it's routine.
The actual risk sits somewhere else entirely. The SASAC review of China switching-chip dependency remains open, unresolved, and un-primary-sourced. It's the one variable here that can move the stock hard, and this filing says nothing about it.
The paper trail doesn't fully add up. When related-filer totals don't reconcile cleanly, that's worth flagging on principle, even if it turns out to be nothing more than filing mechanics across entities.

The SharkWater Take

I'm not reading a signal into this sale. A founder-director moving 1.5 percent of his stake, telegraphed two full sessions ahead of execution, is closer to bookkeeping than to conviction changing. If Samueli were trying to get out ahead of bad news, the market's total non-reaction argues he isn't, or that he isn't the one who knows something the rest of us don't. What I'm actually watching is the SASAC review. That's the variable with the power to move this stock on a single headline, it has no timeline anyone's confirmed, and this filing tells us nothing about it either way. Treat AVGO as fairly priced on the insider story and unpriced on the China story, and watch the second one, not the first.

Tight lines. — SharkWater

Iridium Holders Just Cleared an $8 Billion Rocket Lab Deal, and RKLB Rallied Into It

noreply@blogger.com (Unknown) — Fri, 25 Sep 2026 12:25:39 +0000

SharkWater Trading • Space Desk • M&A • RKLB

Iridium Holders Just Cleared an $8 Billion Rocket Lab Deal, and RKLB Rallied Into It

September 25, 2026

Bottom Line Up Front

Iridium (NASDAQ: IRDM) stockholders approved Rocket Lab's acquisition of the company on September 24, 2026, with 99.6 percent of votes cast in favor, representing 81.0 percent of shares outstanding. The deal pays Iridium holders $27.00 in cash plus Rocket Lab stock for a combined notional value of $54.00 a share, an implied Iridium enterprise value of roughly $8.0 billion. Rocket Lab (NASDAQ: RKLB) closed the day up 4.69 percent to $73.61.

The vote clears the biggest hurdle on the calendar, but closing is still targeted for mid-2027, subject to regulatory sign-off the announcement doesn't itemize. The risk for RKLB holders now sits less in the vote and more in the collar, the financing, and eighteen months of runway for something to go wrong.

What Actually Happened

At a special meeting on September 24, 2026, Iridium stockholders approved the merger agreement first announced June 29, 2026. Rocket Lab's own investor relations release put the vote at 99.6 percent of votes cast, representing 81.0 percent of Iridium's total shares outstanding. That's about as clean an approval as a shareholder vote gets. Iridium CEO Matt Desch called it "an important milestone toward bringing together two companies," and Rocket Lab founder and CEO Peter Beck used almost the identical phrase in his own statement. No 8-K formally disclosing the vote tally had posted to SEC EDGAR as of this writing, so today's confirmation rests on the companies' own release, not yet a filed document.

The Deal Mechanics, Confirmed Against the Original Filing

The consideration structure, verified against the June 29 announcement and its SEC exhibit, is straightforward on the cash side and less so on the stock side. Each Iridium share converts into $27.00 in cash plus a number of Rocket Lab shares set by an exchange ratio that is subject to a collar. That collar bands Rocket Lab's stock price between $67.50 and $112.50. RKLB closed Thursday at $73.61, comfortably inside that band. The exact formula for how many Rocket Lab shares an Iridium holder receives inside the collar was not disclosed in the announcement itself. Rocket Lab's release says the full mechanics live in the transaction agreement filed with the SEC, and this desk has not yet located a definitive merger agreement document that spells it out share by share.

A collar works like a channel marker in a harbor. Inside the marked lane, the boat has room to drift with the current and still make it to the dock at roughly the value promised. Stray outside the markers, high or low, and the deal terms lock in place rather than floating with the tide. Rocket Lab's stock is sitting well inside the channel right now, which is the calm version of this story.

The Bill Rocket Lab Is Taking On

The cash half of a $27.00-per-share payment across Iridium's outstanding share count is a real number, and Rocket Lab isn't paying it out of pocket alone. The company has committed financing for a $3.6 billion, 364-day senior secured bridge term loan from Deutsche Bank and Wells Fargo, on top of balance-sheet cash and other debt and equity sources it hasn't fully specified. Rocket Lab also completed a $1.944 billion at-the-market equity offering on September 15, 2026, which this desk reads as at least partly aimed at funding this transaction, though the company hasn't stated that directly. A 364-day bridge loan is, by design, a short-term instrument meant to be refinanced or repaid quickly. That refinancing has to happen sometime between now and mid-2027, and the terms of it aren't public yet.

The Bull Case

The vote is the cleanest kind of milestone. 99.6 percent approval on 81 percent share turnout removes stockholder risk almost entirely. What's left is regulatory, not political.
Vertical integration is the actual thesis, not just deal size. Rocket Lab builds launch vehicles and satellites; Iridium operates one of the only functioning global satellite constellations. Owning both ends of that chain is a different business than being a launch contractor for other people's satellites.
The market reaction on approval day was positive, not defensive. RKLB rallied 4.69 percent the same day the deal cleared its biggest hurdle, rather than selling off on dilution or leverage concerns, which is the more common reaction to a company taking on billions in acquisition debt.

The Bear Case

Mid-2027 is a long runway. A year and a half between a stockholder vote and a closed deal is a lot of time for regulatory review, financing markets, or Iridium's own satellite business to hand the desk a surprise.
$3.6 billion of bridge debt on a 364-day clock is a refinancing problem waiting to happen. Bridge loans get termed out or repaid, and the cost of doing either depends on credit markets Rocket Lab doesn't control eighteen months from now.
The exchange ratio mechanics still aren't fully public. Until the actual formula inside that $67.50 to $112.50 collar is confirmed, nobody outside the two companies knows exactly how many Rocket Lab shares get issued, which means nobody outside the two companies can fully model the dilution.

The SharkWater Take

The vote is good news and the market treated it that way, but the vote was always the low-risk part of this deal. Stockholders rarely torpedo a board-approved merger, and 99.6 percent approval confirms that pattern rather than telling us anything new. What I actually want before I'd treat RKLB as a clean read on this acquisition is the exchange ratio formula and a clearer picture of how the $3.6 billion bridge gets termed out. Both of those live in documents this desk hasn't located yet. Until they surface, I'm filing this as confirmed progress on a deal that was already priced as likely to happen, not as a new reason to change how I think about the stock.

Tight lines. — SharkWater

Broadcom's Chairman Just Filed to Sell $276 Million More Stock. It's the Second Time in Three Months.

noreply@blogger.com (Unknown) — Thu, 24 Sep 2026 12:29:10 +0000

SharkWater Trading • Insider Desk • Semiconductors • Broadcom

Broadcom's Chairman Just Filed to Sell $276 Million More Stock. It's the Second Time in Three Months.

September 24, 2026

Bottom Line Up Front

Three entities tied to Broadcom (NASDAQ: AVGO) co-founder and Chairman Henry Samueli filed Form 144 notices on September 23, 2026 to sell a combined 774,664 shares worth $275.8 million, at an implied price near $356 a share. It is the second such combined notice in three months. D95GT LLC and H&S Investments I LP filed to sell 694,241 shares for $250.0 million on June 26, 2026, at $377.10 to $388.50 a share.

Broadcom stock fell 2.6 percent to $354.99 the same day, on a session where the Financial Times reported China is reviewing Broadcom's data-center networking hardware in state-backed facilities. A Form 144 is a notice of intent to sell, not proof the trade executed, and none of the filings disclose what fraction of Samueli's total stake this represents.

What Actually Got Filed

Three separate Form 144 filings hit SEC EDGAR on September 23, 2026, all naming the same broker, Northern Trust Securities, and all listing the same approximate sale date: that same day. H&S Investments I LP's filing states its relationship to Broadcom in plain language: "Chairman of the Board of Broadcom." The Samueli Foundation and D95GT LLC are each listed only as "Affiliate."

Filer	Shares	Aggregate Value	Implied Price
H&S Investments I LP	70,218	$25,000,058.61	$355.99
Samueli Foundation	72,474	$25,797,598.63	$355.97
D95GT, LLC	631,972	$225,004,972.58	$356.02
Combined	774,664	$275,802,629.82	n/a

Source: SEC EDGAR, Form 144 filings for Broadcom Inc. (CIK 0001730168), filed September 23, 2026. Implied price per share is calculated from the aggregate market value and share count each filing discloses, not a quoted execution price.

This Is Not the First Time

D95GT LLC and H&S Investments I LP filed a nearly identical pair of notices on June 26, 2026, proposing to sell a combined 694,241 shares for $250,005,851, at a price range of $377.10 to $388.50 a share. That filing's disclaimer language tied D95GT directly to Samueli, stating he was "disclaiming beneficial ownership except to the extent of his pecuniary interest." Today's D95GT filing does not repeat that language. Its relationship is stated only as "Affiliate," so the link to Samueli here rests on the June precedent, the identical broker, and the same-day, same-family filing pattern, not on today's document standing alone.

A Form 144 is a flare fired before the boat leaves the dock. It tells you someone is planning to cast off and roughly how much cargo is going with them. It doesn't tell you if the trip actually happens, or how much line is still tied up back at the pier.

The Backdrop: A China Headline and a Rough Tape

Broadcom closed at $354.99 on September 23, 2026, down 2.62 percent from a prior close of $364.54, on volume of 22,438,949 shares (Yahoo Finance). The same day, the Financial Times reported, per aggregated coverage from Investing.com and SDxCentral, that Chinese authorities are reviewing Broadcom's data-center networking switches used in state-backed facilities. Separately, MarketBeat's same-day note pointed to a broader semiconductor-sector risk-off tone, alongside rising oil prices and Treasury yields. The CBOE Volatility Index closed at 15.18 that day, up 6.83 percent, consistent with a market-wide pullback rather than an AVGO-specific event.

None of these threads are confirmed to be causally connected. The Form 144 filings, the China-review report, and the broader risk-off session all landed on the same calendar day. That is a fact about timing, not a proven relationship.

The Bull Case

This is diversification, not a vote of no confidence. Founder-and-chairman stock sales through personal LLCs and family foundations are a routine estate and philanthropic-planning pattern at companies with long-tenured founders, and Samueli has sold in comparable size before without it marking a top.
The price basis argues against panic selling. Today's implied price, around $356, is roughly 3 percent below June's $377 to $388.50 range. That is a modest step down over three months, not a scramble for the exits into a falling stock.
The China review is a report, not a ruling. No confirmed Chinese regulatory action, ban, or procurement change has been announced. A review can end in nothing.

The Bear Case

A quarter-billion dollars, twice in a quarter, is not a rounding error. Whatever the motive, $275.8 million in fresh sale notices from the Chairman's own vehicles, on top of $250.0 million in June, is a meaningful and repeating supply of stock hitting the market from the most informed insider at the company.
The timing invites scrutiny even if it's coincidental. Filing a large insider sale notice on the same day a China-hardware-review story breaks and the broader tape sells off is the kind of coincidence that erodes a stock's benefit of the doubt, whether or not the filer even saw the story first.
Nobody discloses what's left. None of the three filings state what percentage of Samueli's total Broadcom holdings this represents, so there's no way to judge from the filings alone whether this is a trim or the start of a bigger unwind.

The SharkWater Take

I don't read this as a bearish signal on its own. Samueli has run this exact playbook before, in roughly the same size, three months ago, and Broadcom's business didn't fall apart in between. What I do think is worth tracking is the cadence. Two combined nine-figure sale notices in a single quarter, from the company's own Chairman, is a pattern now, not an isolated data point, and patterns are worth watching even when each individual filing has an innocent explanation. The China hardware-review report is the more interesting thread of the two stories today, because unlike the insider sale it actually has the potential to move Broadcom's revenue, and it deserves its own follow-up once there's an actual regulatory outcome to react to rather than a report of a review.

Tight lines. — SharkWater

The SEC Just Let ARK Venture Fund Build an Exchange Ramp and a Blockchain Ramp Off the Same Fund

noreply@blogger.com (Unknown) — Wed, 23 Sep 2026 12:36:55 +0000

SharkWater Trading • Venture Desk • ARKVX • Tokenized Shares

The SEC Just Let ARK Venture Fund Build an Exchange Ramp and a Blockchain Ramp Off the Same Fund

September 23, 2026

Bottom Line Up Front

On September 21, 2026, the SEC issued Release No. IC-36333, an exemptive order letting ARK Venture Fund (ticker ARKVX, currently an unlisted closed-end interval fund) add a share class listed on a national securities exchange and a second share class that trades as a tokenized instrument on alternative trading systems. Nothing launches today. No exchange is named, no ATS is named, no fee percentage is disclosed, and the fund's own site still shows blank fields where a current NAV should be.

What it means for the reader: this is a regulatory door being unlocked, not a new way to trade ARKVX yet. The interesting part is what happens to a fund that spent its whole life as an illiquid interval fund once part of it can, eventually, print on an exchange or move on a blockchain ledger.

What the Order Actually Does

ARK Venture Fund and ARK Investment Management LLC applied for, and received, an amendment to a prior November 2025 exemptive order. Reading the order directly (sec.gov/files/rules/ic/2026/ic-36333.pdf, File No. 812-16031), the operative language permits the fund to offer "a class of shares listed on a national securities exchange" and "a class of tokenized shares traded on one or more alternative trading systems." The Commission also cleared the fund to charge asset-based distribution and service fees, plus early withdrawal charges, on top of what it already collects.

The relief is effective immediately. That is a legal fact about the order, not a statement about the product. No launch date, no named exchange, no named ATS, and no fee schedule appear anywhere in the order or in the press coverage that followed it (crypto.news, CryptBull, KuCoin, CoinSpectator). ARK still has to actually build and register the specific share classes before anyone can buy either one.

What ARKVX Is Today

ARKVX, per its own fund page (ark-funds.com/funds/arkvx), is "an actively managed closed-end interval fund." That structure means no daily exchange trading. Investors get in and out through periodic repurchase offers on a schedule the fund controls, not whenever they want. Press coverage of the underlying application pegged the fund's assets near $562 million at the time it was filed; that figure is secondary-sourced, and ARK's own site did not display a current NAV or total net assets figure when checked for this post. If ARK publishes an updated number, this desk will use it. Until then, treat $562 million as directionally useful and dated to the application, not today.

The fund holds itself out as exposure to five buckets: AI and next-generation internet, space and defense, autonomous technology and robotics, digital assets and fintech, and genomic biotech. That is the pitch that made ARKVX a retail-accessible way to touch pre-IPO names without a venture fund minimum. The interval structure was always the tax retail investors paid for that access: you got the exposure, you gave up the ability to sell on your own schedule.

An interval fund is a boat with no engine. You can get on board, and eventually you can get off, but only when the dock schedule says so. This order is the SEC signing off on installing an engine. It does not say when the engine gets bolted on, and it does not promise the boat will actually leave the dock any faster once it is.

Why This Order, and Why Now

Tokenized fund shares have been circling U.S. regulators all year as tokenization platforms pushed to bring private-market and fund exposure onto public blockchain rails. An exemptive order naming a specific, well-known fund and granting both an exchange-listed class and a tokenized class in the same document is a bigger structural signal than another crypto-adjacent press release. It tells every other sponsor of an illiquid closed-end vehicle, business development companies, other interval funds, private-credit vehicles, that this path now has a precedent to point to at the SEC.

None of that changes what an investor can do with ARKVX this week. The order is dated September 21, 2026. As of this writing there is no confirmed follow-up filing naming an exchange, a transfer agent for the tokenized class, or a target launch quarter.

The Bull Case

Real liquidity upgrade, eventually. An exchange-listed share class would let holders sell in the open market instead of waiting on a quarterly repurchase window, closing the single biggest complaint about interval funds.
First-mover precedent. ARK is one of the first funds of this size and profile to get both an exchange class and a tokenized class approved together, which is a marketing and product edge if it actually ships.
Retail access to pre-IPO names gets structurally easier. If tokenized shares trade on an ATS with lower minimums or faster settlement, the fund's stated pitch, retail exposure to private companies, gets closer to what it has always advertised.

The Bear Case

There is no product yet. An exemptive order is permission, not execution. Plenty of SEC-approved structures sit unused for years, or never launch at all.
The underlying portfolio didn't get more liquid. ARKVX's actual holdings are private-company stakes. Wrapping them in an exchange-listed or tokenized share class does not make the underlying assets easier to value or sell; it just changes how the wrapper trades.
New fee authority cuts against holders, not for them. The same order that unlocks new share classes also clears the fund to add asset-based distribution and service fees plus early withdrawal charges. None of those numbers are public yet, and there is no guarantee they land in the holder's favor.

The SharkWater Take

This is a real, dated, primary-sourced regulatory event, and I'd rather flag it early than wait for the product announcement everyone else will cover at the same time. But I'm not buying ARKVX on this order, and I wouldn't tell a reader to either. There is no ticker to trade here that doesn't already exist, no new liquidity today, and no fee schedule to price into the decision. The order is the SEC saying "you're allowed to build this." It is not ARK saying "it's built." Watch for the actual S-1 or prospectus supplement naming an exchange and a tokenization partner. That filing, not this one, is the one that turns into a position.

Tight lines. — SharkWater