Marques Brownlee summed up the launch in one line:

New Surface Laptop Ultra, powered by RTX Spark starts at $2599 https://t.co/GFLe30Za71

— Marques Brownlee (@MKBHD) October 7, 2026

The starting price is not the number that tells you what local AI costs. The memory upgrade is.

Local AI is a memory problem

Running a large AI model means holding it in memory. The bigger the model, the more memory it needs, which is why Microsoft’s top configuration offers 128GB of unified memory. According to ServeTheHome’s breakdown of the lineup, the base model with 24GB of memory costs about $2,600. A 20-core version with 32GB and a 1TB drive costs about $3,700. The same 20-core chip with 128GB and the same 1TB drive costs about $5,900.

That is roughly $2,200 for the extra memory alone. ServeTheHome noted that some of Microsoft’s own models are recommended for systems with 128GB or more, and described RAM’s current cost as extreme.

That cost has a familiar cause. Memory prices have soared as AI data centers buy up supply, the same squeeze we covered in our piece on phone prices and that pushed Apple to raise the price of every iPhone by at least $100. Local AI asks the buyer to pay for one of the fastest-rising costs in electronics, and to buy a lot of it.

Who carries the cost

There is an economic shift hiding in the local AI pitch. When you use an AI model in the cloud, the company running it pays for the chips, the memory and the electricity, and recovers that cost through subscriptions or bundled services. When the model runs on your laptop, much of that cost moves to the hardware you bought. Microsoft promoted a hybrid approach in which Copilot runs some models locally and taps the cloud for extra compute. Every task that stays on the device is compute the cloud provider does not have to supply.

That is not necessarily a bad deal for users. It is a different one, and worth seeing clearly. A $5,900 laptop is a large up-front purchase of the capacity a cloud service would otherwise rent to you over time.

The case for running AI locally

There are real reasons to want AI on your own machine. Data that never leaves the device is easier to keep private. Local models work offline and can respond faster without a round trip to a data center. For developers, a machine that can run large models without paying per request can save money over time. And the Surface Laptop Ultra is not only an AI box: Engadget noted gaming support including ray tracing and DLSS, and Microsoft highlighted a replaceable SSD.

On price, ServeTheHome called the top configuration surprisingly competitive, noting that the cheapest 14-inch MacBook Pro at the same memory level costs $800 more. High-memory machines are expensive everywhere right now.

There is also a significant shift in who makes the chip. The RTX Spark is an Arm-based Nvidia processor, according to ServeTheHome. Microsoft’s most powerful Surface does not use Intel, AMD or Qualcomm, which says a lot about where the PC industry thinks AI performance will come from.

What to watch

The question for the next year is not whether AI can run on a laptop. It clearly can. It is whether enough people will pay thousands of dollars extra in memory to do it, while similar capabilities are available from cloud services for a monthly fee. If local AI catches on, expect memory to become the most important spec on a PC’s box. If it doesn’t, the 128GB laptop will remain a developer’s tool, and most people’s AI will keep running on someone else’s hardware.

Related reading: Memory Prices Are on Track to Rise 400%. Your Next Phone Is Paying for AI. · Google’s $899 Googlebook Isn’t a Chromebook Successor.

The post Microsoft’s AI Laptop Starts at $2,599. The 128GB Version Built for Local AI Costs About $5,900. appeared first on Techpinions.

On September 30, the Senate blocked the Ratepayer Protection Act, a bill meant to keep the cost of new power infrastructure for data centers off household electric bills. The vote was 57 to 43, short of the 60 needed to advance, CBS News reported, even though the bill had passed the House 417 to 3. On October 6, Alphabet signed a 20-year agreement with Constellation Energy that will add 890 megawatts of new nuclear output to the largest grid in the country, backed by more than $4.3 billion in upgrades that Constellation will fund.

Washington could not agree on who pays. One of the biggest buyers wrote its own answer.

What the Senate fight was about

Nearly everyone in the debate agreed on the principle that data centers should pay their own way. The disagreement was over whether the bill would make them. The Ratepayer Protection Act, led in the Senate by Sen. Jon Husted of Ohio, would require state utility regulators to consider standards holding large data centers responsible for the costs of the infrastructure they need. It would not require regulators to adopt them.

That word, consider, is why all but four Democrats voted against advancing it. Senate Democratic leader Chuck Schumer called it a toothless messaging bill and said Democrats want it mandatory, according to CBS. Sen. Martin Heinrich of New Mexico had objected earlier in the month when Husted tried to pass it by unanimous consent:

Sen. Heinrich objected to Sen. Husted’s attempt to pass the data center-focused Ratepayer Protection Act, argued it didn’t do enough because it was voluntary

— Burgess Everett (@burgessev) September 17, 2026

Republicans saw it differently. They argued the vote put Democrats on record opposing protection for ratepayers, and Husted said in a post on X that Democrats blocked a bill to keep families and small businesses from footing the bill for Big Tech’s energy needs. Heinrich offered his own bill, which Democrats said would require large data centers to cover specified grid costs, and Republicans objected to that one. The result is that neither version has passed.

Why the bill mattered

The costs at stake are real. As we reported in August, the price paid in PJM’s capacity auction, the market that keeps enough power available for peak demand across 13 states, rose roughly tenfold over two years, and the market’s independent monitor attributed most of that increase to data centers. Those costs are recovered from everyone on the grid, not just the companies building the data centers.

Without a federal standard, the question goes back to the states and their utility commissions, where it is being decided one rate case at a time.

What Google bought

Alphabet’s deal points to a different kind of answer. The 890 megawatts will come from upgrades to 11 existing Constellation reactors in Illinois, Pennsylvania and New Jersey, with the first increment due by 2028 and all of it online by the end of 2032, according to Big Energy News. Constellation said the added output equals that of several small modular reactors. The capacity feeds PJM, the same grid where data center demand has pushed prices up. A separate 15-year agreement covers another 2,700 megawatts from Constellation’s existing plants.

The distinction matters. When a data center simply plugs into the grid, it competes with everyone else for existing power and, as PJM’s auctions show, can raise prices for all. When a company contracts for new supply that would not otherwise exist, it adds to the pool instead of only drawing from it. Uprates, which squeeze more output from existing reactors, are one of the fastest ways to do that, because the plants and their grid connections are already in place.

Google is not alone. Amazon signed a similar deal with Constellation the week before for 690 megawatts, including upgrades at Maryland’s Calvert Cliffs plant, Bloomberg reported, and Microsoft previously agreed to buy power from the Three Mile Island plant that Constellation plans to restart.

The limits of buying your way out

It would be a mistake to treat these deals as solving the problem. The new nuclear output arrives between 2028 and 2032, while data center demand is rising now. Deals like these are signed by the largest companies with the deepest pockets, and on their own they do not address the costs of transmission lines and grid upgrades that other data center projects trigger. And a voluntary corporate contract is, by definition, voluntary, the same objection Democrats raised about the Senate bill.

Still, the contrast is instructive. The policy question everyone claims to agree on, that data centers should pay for the power they need, has stalled in Congress over whether to make it a requirement. In the meantime, some of the companies driving the demand are contracting for new supply on their own terms. Whether that becomes the norm, or stays the exception for a handful of giants, will decide how much of AI’s power bill lands on everyone else.

Related reading: AI’s Energy Use Is Both Wildly Overblown and Seriously Underrated. · 57% of Texas Voters Don’t Want a Data Center Next Door. · A Startup Wants 456 Nuclear Reactors in Utah.

The post The Senate Couldn’t Agree on Who Pays for Data Center Power. Google Just Signed Up for 890 New Megawatts. appeared first on Techpinions.

Meanwhile, the banks are not waiting. On September 28, Citi expanded its partnership with Coinbase so that its large corporate clients can accept stablecoin payments through its Spring by Citi platform, with Coinbase handling the blockchain side and converting the tokens to dollars for settlement, The Block reported. The merchant never has to hold a token.

Coinbase CEO Brian Armstrong framed it this way when the partnership began last year:

It’s not a debate anymore – crypto and stablecoins are the tools that will update the global financial system.

Excited to be collaborating with @Citi to work on improving stablecoin utility and digital asset adoption with their clients. https://t.co/IDpniQq5hh

— Brian Armstrong (@brian_armstrong) October 27, 2025

He is half right. The debate over whether stablecoins belong in the financial system is mostly over. The fight now is over something narrower and much more valuable: who gets paid for holding them.

The money is in the reserves

A stablecoin is a token redeemable for one dollar, backed by reserves the issuer holds in cash and short-term Treasuries. Those reserves earn interest. The market is roughly $300 billion, dominated by Tether and Circle’s USDC, according to a White House Council of Economic Advisers analysis from April.

The GENIUS Act bars stablecoin issuers from paying interest or yield to the people who hold their tokens. But as the council notes, the law does not explicitly prohibit affiliate or third-party arrangements that pass some of that interest along. In practice, that channel already exists: Circle shares part of the interest on its reserves with Coinbase in proportion to the USDC held on its platform, according to the Congressional Research Service.

Banks versus crypto, in their own words

That gap is where the lobbying has gone. In comments on the regulators’ proposals, bank groups including the American Bankers Association and the Bank Policy Institute urged a broad reading that would treat any economic benefit tied to holding a stablecoin as prohibited interest, American Banker reported. Their concern is deposits. Money that sits in a stablecoin earning rewards is money that is not sitting in a bank account funding loans. Estimates cited in the debate warned that if third-party rewards are allowed, community bank lending could fall by as much as $850 billion.

Coinbase and the American Fintech Council argued the opposite: the law bans only issuers from paying yield, and ordinary commercial arrangements, including rewards from platforms, should remain allowed. Sens. Thom Tillis and Angela Alsobrooks proposed a middle path that would prohibit rewards that are economically or functionally equivalent to bank deposits.

What the White House economists found

The White House economists tested the banks’ central claim. At the council’s baseline assumptions, eliminating stablecoin yield would increase bank lending by about $2.1 billion, or 0.02 percent, at a net welfare cost of $800 million. Even stacking every worst-case assumption, the council found $531 billion in additional lending, a figure that would require the stablecoin market to grow to about six times its current size relative to deposits, among other conditions it called implausible. Its conclusion was that a yield ban would do very little to protect bank lending while giving up the consumer benefits of competitive returns.

That is a striking gap: a potential $850 billion hit from the banks’ side, a $2.1 billion effect in the economists’ baseline. The two are not measuring exactly the same scenario, and models depend on their assumptions. But the distance between them is wide enough that the debate is not only about protecting lending to farms and small businesses. It is also, unavoidably, about which industry earns the interest on a growing pool of dollars.

Why the Citi deal fits

Seen that way, Citi’s move makes sense. The bank is not fighting stablecoins. It is making them a payment rail that runs through its own accounts, where the money still settles as dollars in a bank. If stablecoins become plumbing rather than savings, banks keep the deposits and the crypto firms earn fees for moving money. If they become a place to park dollars and earn rewards, the balance of power shifts.

Regulators are still writing the rule that decides which of those futures arrives. The OCC’s final GENIUS Act rule is still pending, and how it defines prohibited interest will matter more than any partnership announcement. Watch that definition. It is where the real money in stablecoins is being decided.

Related reading: Crypto’s Big Bill Failed 49–50. The Rules Coming Instead Can Be Undone Just as Fast.

The post Banks Warn Stablecoin Rewards Could Cost $850 Billion in Loans. White House Economists Put a Yield Ban’s Benefit at $2 Billion. appeared first on Techpinions.

California Attorney General Rob Bonta, who led the states, announced the settlement on X:

We’ve reached a settlement agreement with Paramount that includes court enforceable guardrails to resolve the antitrust issues at the center of our case, protect competition and consumer choice, and provide for the production of significantly more films and a minimum of $1.5… pic.twitter.com/6bnckAfj1B

— Rob Bonta (@AGRobBonta) September 21, 2026

The settlement is worth reading closely, because it shows what modern antitrust enforcement settles for. Nothing is being sold up front. Instead, Paramount made a set of promises, and most of them run for five years.

What Paramount agreed to

According to the California attorney general’s office, the combined company must release 30 films a year, including 20 wide releases, in each of the first two years, rising to 32 films with 21 wide releases in years three through five, plus at least four independent films a year. Missing those targets carries real penalties: $30 million per missed film, and potentially selling Miramax.

Paramount also committed at least $1.5 billion in additional U.S. film production spending over five years, $47.5 million for a workforce training fund, separate negotiations with cable distributors for Paramount and Warner Bros. channels, continued operation of the free Pluto TV service, and an independent monitor to check compliance. It also agreed to a News Editorial Independence Board for CNN and CBS.

Promises instead of divestitures

The states had argued that the merger would concentrate nearly a third of theatrical releases and basic cable programming in one company, Al Jazeera reported. The traditional fix for that kind of concentration is structural: make the buyer sell a studio, a network, or a library so the market keeps another competitor. That did not happen here. Earlier talks over remedies including asset sales failed, The Desk reported, and the final deal relies on commitments about behavior instead.

Behavioral remedies have a basic weakness, which is time. A divested studio stays independent. A promise to release 30 films a year lasts as long as the agreement does. After five years, the combined company will still own everything it bought, and the guardrails that justified the deal will be gone unless regulators act again.

The news provisions are the clearest example of how much rests on trust. A board to protect editorial independence at two of the country’s largest news organizations is an unusual thing to find in an antitrust settlement. But according to Al Jazeera, David Ellison, Paramount’s chief executive, retains authority over appointments to that board. Independence that the owner appoints is a meaningful safeguard only as long as the owner wants it to be.

What each side says

Bonta was careful not to endorse the merger itself. He said the settlement is not a vote of support for the deal, and added, “I don’t think these companies should merge,” ABC7 reported. Ellison said the shared aim was an outcome that serves consumers, workers and the creative community, MyNewsLA reported. Judge Araceli Martínez-Olguín called the settlement a fair, reasonable, and good-faith approach to the competitive harms the states alleged, according to Al Jazeera.

Critics were less generous. Sen. Elizabeth Warren called the outcome disastrous, arguing that it lets a single conglomerate dominate American news and entertainment, Al Jazeera reported. A coalition opposing the deal argued in a court filing, according to NBC News, that small businesses and independent contractors would not get the same protections as members of major unions. The federal Justice Department had already cleared the deal without conditions in June, according to Al Jazeera, which is why the states’ case was the last obstacle.

There is a reasonable business case on Paramount’s side, too. Traditional studios and networks are competing against streaming giants with global scale, and a larger combined library and production base is one way to stay in that fight. The settlement’s film commitments, if they hold, put a floor under how many movies the combined company releases for the next five years.

What to watch

The independent monitor’s reports will show whether the film and production commitments are being met. The more revealing signal will come from the newsrooms: who Ellison appoints to the editorial board, and whether the board ever publicly disagrees with the company. And in 2031, when the five-year terms run out, the question will be whether anything about this merger still looks different from an ordinary one.

Related reading: Streaming Prices Keep Rising and Nobody’s Leaving.

The post Paramount Can Now Buy Warner Bros. for $110 Billion. Most of Its Promises Expire in Five Years. appeared first on Techpinions.

Tim Cook introduced it on X:

Apple Watch helps millions better understand their health simply by wearing it.

Today we introduced two powerful new features: one that will notify you if it detects signs of hypertension over time, and sleep score to give you more insight into the quality of your sleep. pic.twitter.com/f1cvPYy46Y

— Tim Cook (@tim_cook) September 10, 2025

A year in, the most useful thing to understand about the feature is not what an alert means. It is what no alert means.

What the numbers actually say

The watch does not measure blood pressure. It looks at how blood vessels respond to heartbeats over 30-day periods and flags patterns consistent with hypertension. In Apple’s own validation study, about 41 percent of people with undiagnosed high blood pressure received an alert, and about 59 percent did not, MedTech Dive reported. About 7.7 percent of people without hypertension received a false alert.

That is a deliberate trade-off. The feature is tuned to avoid false alarms, so when it does alert, the signal is fairly strong. A February study in JAMA, led by Jordana Cohen of the University of Pennsylvania, applied Apple’s validation numbers to national health survey data and estimated that about 69 percent of people who get an alert actually have hypertension.

The other side of that trade-off is the part most users never think about. Because the feature misses more undiagnosed cases than it catches, silence is not reassurance. The JAMA study estimated that for adults 60 and older, someone who gets no alert still has about a 34 percent chance of having undiagnosed hypertension. Cohen’s conclusion was blunt: the absence of an alert provides limited reassurance, particularly in older and higher-risk adults.

Why that matters more than the alerts

Most people who wear an Apple Watch will not get a hypertension alert, and for many of them that will be accurate. But a feature that runs quietly in the background for a year creates an easy assumption: if something were wrong, the watch would have told me. For the people most likely to have high blood pressure, that assumption is the most dangerous part of the product.

The risk is not that the feature is bad. It is that it could replace something better. A blood pressure cuff reading at a pharmacy or doctor’s office remains the way to know your numbers. If a quiet watch becomes a reason to skip that check, a screening tool designed to catch more cases could end up delaying some diagnoses.

The case for the feature

None of that means the alerts are not worth having. About 1 in 6 adults with high blood pressure is unaware of it, according to the CDC. A tool that screens millions of wrists continuously, at no extra cost to watch owners, and correctly flags a meaningful share of them is a real public health contribution, even if it catches fewer than half. Apple’s projection of one million alerts in the first year, if it held, would mean a lot of people learning about a serious condition they did not know they had.

Apple has also been careful about how it frames the feature. It is cleared as a notification, not as a blood pressure monitor, and Apple advises people who get an alert to log readings with a cuff for a week and share them with a doctor, MacRumors reported.

There is also a sign the regulatory system is doing its job. Apple’s new Apple Watch Series 12 and Ultra 4 use redesigned heart sensors, and the hypertension feature is not yet available on them in more than 30 countries while additional regulatory clearances are in process, MacRumors reported. In the U.S., the U.K. and the EU, the feature is available at launch. New hardware means the algorithm has to be validated again, which is how a medical feature should work.

What to do with this

If your watch alerts you, take it seriously and follow up with real readings. If it never has, do not read that as a clean bill of health, especially if you are over 60 or have other risk factors. Get a cuff reading. The watch is a smoke detector that goes off for some fires, not a guarantee that the house is safe.

Related reading: Apple Raised the Price of a Year-Old iPhone by $100.

The post Apple Watch’s Blood Pressure Alert Turns One. It Misses About 59% of Undiagnosed Cases. appeared first on Techpinions.

One commissioner pointed out a problem with that number. Anna Gomez wrote that fewer than half of satellite users actually got those speeds in 2025, and called the 99.7 percent figure misleading. Then, on September 30, the FCC took a step aimed at the capacity limits behind that gap.

The vote: more airwaves for satellites

At its September meeting, the FCC freed up more than 1,000 megahertz of spectrum in the 12.7 GHz and 42 GHz bands for satellite services, and exempted satellite launches and operations from federal environmental review under the National Environmental Policy Act, Via Satellite reported. Launches remain subject to environmental review by the Federal Aviation Administration. The commission also proposed opening up more spectrum in other satellite bands.

The 12.7 GHz allocation alone represents more than a 25 percent increase in downlink spectrum for satellite operators, according to TeslaNorth’s analysis, which noted that SpaceX’s Starlink has been straining against capacity limits. Amazon, AST SpaceMobile, Globalstar and others also stand to benefit.

It is the latest in a series of moves this year to give satellite internet more room. Chairman Brendan Carr previewed a separate vote earlier in the year to lift satellite power limits:

The FCC will vote this month to supercharge your Internet.

Lifting outdated power limits on satellite Internet means faster speeds and more capacity.

pic.twitter.com/Qy7yuHv8Mx

— Brendan Carr (@BrendanCarrFCC) April 8, 2026

Why the two numbers belong together

Read on their own, the August report and the September vote look like separate stories. Together they describe a strategy. The FCC’s report found that about 10.5 million people, more than 9 million of them in rural areas, still lack fixed land-based broadband at the 100/20 benchmark. The agency is treating satellite as a large part of the answer for those people. And the main thing standing between satellite service and that benchmark is capacity: as the Benton Institute for Broadband & Society has noted, satellite service can slow during periods of congestion, when many customers share the same capacity.

More spectrum is one of the most direct fixes for that. So is launching bigger satellites, which is why Starship’s first Starlink V3 deployment matters to this story too. If the satellites and the spectrum arrive, the gap between “available” and “actually delivered” could narrow. The FCC is, in effect, betting that it will.

What each side says

Supporters see satellite as a practical way to reach remote homes that wired networks have not. Carr has framed the September changes as removing the threat of activist lawsuits and last-minute surprises from next-generation space development, and Commissioner Olivia Trusty said the goal was to avoid constraining American innovation and leadership in space. Gomez supported the spectrum item as sensible spectrum management, according to Via Satellite.

Critics focus on the report rather than the spectrum. Benton argued that the FCC overstated satellite’s reach, noting that satellite accounts for about 2.3 percent of actual subscriptions despite the 99.7 percent availability figure. Benton and Gomez also objected to the report dropping affordability from its analysis, and Benton criticized the agency for abandoning its long-term goal of gigabit download speeds. Carr, for his part, credited the administration’s policies with rising speeds, falling prices and stronger competition.

What to watch

The test of the FCC’s bet is measurable. Next year’s broadband report could show whether the share of satellite customers actually getting 100/20 has risen from under half. If it has, the 99.7 percent figure starts to mean something closer to what it says. If it hasn’t, the 10.5 million people without a land-based connection will be counted as served by a service that often falls short of the benchmark. Either way, the number to watch is not how many people satellite can reach. It is how many it actually delivers to.

Related reading: Starship Finally Reached Orbit With 26 Starlink Satellites.

The post The FCC Says 99.7% of Americans Can Get Satellite Broadband. Fewer Than Half of Satellite Users Got the Promised Speeds. appeared first on Techpinions.

Another blockbuster investment for Arizona

TSMC has announced another $100 billion for its Phoenix operations, bringing its total U.S. commitment to a historic $265 BILLION.

More leading-edge fabs, more advanced packaging facilities, more high-wage jobs. Another huge vote of… https://t.co/LObCIeF7GF

— Arizona Chamber (@AZChamber) July 16, 2026

Buried in the city’s announcement is the number that explains what Taiwan is doing. When the buildout is complete, about 30 percent of TSMC’s 2-nanometer and more advanced capacity will be in Arizona. The other 70 percent stays in Taiwan. And in late September, Taiwan’s government made clear which part it considers non-negotiable.

Factories can move. The lab stays.

In a policy report to Taiwan’s legislature at the end of September, Premier Cho Jung-tai said the island’s chip industry now spans manufacturing bases on three continents. Taiwan supports TSMC’s U.S. expansion, DigiTimes reported, but insists that core research and development stays home.

Taiwan is putting money behind that line. Earlier in September, the government laid out plans to invest NT$300 billion, about $9.5 billion, through 2033 in next-generation chip design and advanced semiconductor processes, Focus Taiwan reported. The program includes a research line for processes below 1 nanometer. Cho said Taiwan still needs to keep investing in research and talent to consolidate its advantages.

The production schedule follows the same logic. TSMC began 2-nanometer production in Taiwan in late 2025, Focus Taiwan noted. In Arizona, the first fab has been making 4-nanometer chips since late 2024, and a second fab for 3-nanometer chips is expected to reach volume production in 2027, according to the city. The newest processes start in Taiwan and arrive in Arizona later.

Why the split matters

This is a modern version of what is often called Taiwan’s “silicon shield”: the idea that the world’s dependence on Taiwanese chipmaking gives other countries a strong interest in Taiwan’s security. Moving all of TSMC’s best manufacturing to the United States would weaken that shield. Moving none of it would strain Taiwan’s most important relationship. The emerging answer is to export capacity while keeping the frontier, meaning the research that produces each new generation of chips and the first factories that make them.

The United States has pushed for more. A year ago, Commerce Secretary Howard Lutnick suggested that Taiwan help the U.S. produce half of its chips domestically. Vice Premier Cheng Li-chiun, who led Taiwan’s trade talks, rejected the idea, saying Taiwan had never committed to a 50-50 split and would not agree to such conditions, CNBC reported. For now, a 30 percent share of the most advanced capacity, running a generation or so behind Taiwan, is where things stand.

The case that 30 percent is enough

From Washington’s point of view, there is a strong argument that this is a good deal. The risk the U.S. is trying to manage is not falling a few months behind on the newest chip. It is losing access to advanced chips entirely if something goes wrong across the Taiwan Strait. Arizona State University supply-chain expert John Fowler told KJZZ that tension between Taiwan and China is the biggest threat to the industry. A dozen facilities in Arizona making and packaging chips at or near the leading edge would be an enormous cushion against that risk, even if the very newest process always debuts in Taiwan first.

The counterargument is that factories without the research behind them are a dependency, not independence. If every new process is developed in Taiwan and proven there first, Arizona can make what Taiwan has already mastered, but it cannot advance on its own. In a long disruption, the U.S. would have the capacity it built and little ability to move beyond it.

What to watch

Two things will show where this balance settles. The first is the gap: how many months separate a new process starting in Taiwan from the same process running in Arizona. If that gap keeps shrinking, the U.S. is getting closer to the frontier. The second is people. TSMC’s U.S. plans have included a research and development center since its 2025 expansion announcement. Whether it develops new technology or mainly supports production is the clearest signal of how much of the frontier Taiwan is willing to share.

Related reading: Memory Prices Are on Track to Rise 400%. Your Next Phone Is Paying for AI. · Apple Raised the Price of a Year-Old iPhone by $100.

The post TSMC Will Put 30% of Its Most Advanced Chipmaking in Arizona. Taiwan Is Keeping the Other 70%, and the Lab. appeared first on Techpinions.

The scale is alarming. The more important detail is what was being stored in the first place.

What leaked, and from whom

Security journalist Brian Krebs traced the service, called Nexus, back to IDScan’s infrastructure. According to his reporting, the cache included more than 153 million U.S. and Canadian driver’s licenses, more than 10 million other ID cards, more than 3 million travel documents, and over 579,000 medical cards. The scans carried timestamps that matched when victims had visited rental counters and dispensaries, and they included infrared and ultraviolet images of the kind IDScan’s readers capture. The FBI opened an investigation, and Nexus went offline within days of launching.

The operators claimed they had been quietly pulling new records for more than a year. That claim has not been independently confirmed, but Krebs reported that the cache grew by roughly 400,000 licenses within a single day after the service went live.

IDScan’s customers, according to Krebs, include Hertz, Target, FedEx and Caesars Entertainment, along with more than 1,000 cannabis dispensaries. In Canada, the federal privacy commissioner has opened an investigation into the company’s security and how it notified people.

Security researcher Rachel Tobac warned about what comes next:

This type of breach, affecting 153 million individual's licenses, is going to have massive impacts on identity theft, specifically identity verification platform fraud (using front and back scans from this breach to "verify" as another person for financial services and more). https://t.co/1Jc3iJKcQ5 pic.twitter.com/yEQsRkOcRb

— Rachel Tobac (@RachelTobac) September 2, 2026

The real problem is the photo

Think about what an age check actually needs to know. A bouncer or a dispensary clerk needs one answer: is this person old enough, and is this ID real? That is a yes or a no. Krebs found that IDScan instead kept complete images of each license, front and back, with the infrared and ultraviolet versions and a timestamp of the scan. That is not a record of an age check. It is a copy of the ID.

That choice is what turned a breach into a disaster. A leaked list of names and birth dates is bad. A leaked set of high-resolution, front-and-back license images is a toolkit for fraud. As Tobac notes, criminals can use front and back scans to verify themselves as someone else with financial services. The same scans that proved you were 21 at a nightclub can now be used to pass as you somewhere else.

There is a broader lesson here as more of daily life requires showing ID, from age checks on websites to purchases of age-restricted products. Every system built to check an ID faces the same choice IDScan made: verify and discard, or verify and keep. Keeping is useful to the company, for audits and disputes. But a database of ID scans is valuable to criminals in proportion to how complete it is. Collect enough of them and you have built something worth stealing.

The case for keeping records

To be fair to the businesses involved, there are real reasons to retain proof of an ID check. A bar accused of serving a minor, or a dispensary audited by state regulators, may need to show that it checked a customer’s ID. A car rental company has legitimate reasons to know who drove off in its vehicle.

But proving an ID was checked does not require a full color copy of it. A record that a specific license number passed verification at a specific time would satisfy most of those needs, at a fraction of the risk. The gap between what businesses needed and what IDScan stored is the gap the breach fell through.

What you can do, and what should change

IDScan has offered free credit monitoring and identity protection to affected people. If you think your license may be in the cache, watch for accounts opened in your name and consider a credit freeze, which is free and makes it harder for anyone to open new credit in your name. You can’t easily change your license number the way you can change a password, which is exactly why this kind of data should be collected sparingly.

The more lasting fix is for the businesses that scan IDs to ask their vendors a simple question: what do you keep, and for how long? If the answer is a picture of every license, forever, that is a breach waiting for a date.

Related reading: A Gang Hacked the FBI Over a Four-Month-Old Warning. · Passkeys Won. The Password Still Isn’t Dead.

The post 153 Million Driver’s Licenses Leaked From a Company That Only Needed to Check Your Age appeared first on Techpinions.

The Verge’s Tom Warren posted the pricing on X:

Rockstar Games is pricing GTA 6 at $79.99 for PS5 and Xbox Series S/X. The ultimate edition is priced at $99.99 pic.twitter.com/SONIycjE2i

— Tom Warren (@tomwarren) June 24, 2026

That $80 got most of the attention. It is also the least important number in Take-Two’s plan. The more revealing one is 230 million.

The game that never stopped selling

That is how many copies Grand Theft Auto V has sold, according to Take-Two figures reported by Video Games Chronicle in August. GTA V came out in 2013, and it has kept selling for more than a decade since. Red Dead Redemption 2, Rockstar’s other recent hit, has sold more than 87 million.

Seen against that history, the launch price is a single data point in a decade-long revenue curve. The question for Take-Two is not how much money GTA VI makes in its first week. It is whether GTA VI can repeat what GTA V did for thirteen years.

Why the guidance looked cautious

That framing helps explain a number that disappointed analysts in May. Take-Two guided to $8.0 billion to $8.2 billion in net bookings for the fiscal year that ends in March 2027, the year that includes GTA VI’s launch. That was up from $6.7 billion the year before, but below what analysts expected, Reuters reported. The stock rose about 7 percent in extended trading anyway, as the company reaffirmed the November date. In August, Take-Two reiterated the same range.

One reading is that management is simply being conservative before the biggest launch in its history. Another is that the fiscal year only captures about four and a half months of GTA VI sales, on two console platforms, with no PC version announced. Take-Two chief executive Strauss Zelnick said at a September investor meeting that PC is becoming more and more important to the company, GameSpot reported, while stopping short of announcing a GTA VI PC release. A PC version, whenever it comes, is a second launch that this year’s guidance does not need to include.

The case for worry

The risk in this strategy is concentration, and at least one shareholder said so directly. At the same September meeting, an investor criticized the company’s record of delays and its reliance on a small number of big franchises, and questioned whether management should stay if problems continued, GamesRadar reported. Zelnick pushed back by pointing to how far the company has come since nearly going bankrupt in 2007.

The delays are real. GTA VI was originally slated for 2025, slipped to May 2026, then slipped again to November, ESPN reported. And when one game is expected to drive a step change in a company’s revenue, anything that goes wrong with it, from a rocky launch to a slow start for its online mode, lands on the whole business.

The $80 price carries its own risk. Raising the price of the most anticipated game in years is a test of how far the industry can push, and the answer will shape what other publishers charge. If GTA VI sells at a record pace at $80, the old $70 ceiling is gone for good.

What to watch after November 19

Launch-week sales will produce the headlines, and they are likely to be enormous. The numbers that will matter more come later. How quickly does GTA VI’s online mode build a paying audience? When does a PC version arrive? And does spending from players who already own the game start to look like GTA V’s? A great first week would confirm the game is a hit. Only the next several years can show whether it is the next 230-million-copy franchise.

Related reading: Apple Raised the Price of a Year-Old iPhone by $100. · Streaming Prices Keep Rising and Nobody’s Leaving.

The post GTA 6 Costs $80. GTA V Has Sold 230 Million Copies. Take-Two Is Betting on the Second Number. appeared first on Techpinions.