Read one layer down. The month splits into two stories: most of the dollars sit in a slow release from a single token, while most of the fear sits in percentages attached to tokens that already fell. Neither maps cleanly onto the sell pressure a reader is trying to price.

Nearly half the month is one token, and it drips

One name dominates the August token unlocks. RAIN accounts for the single largest release of the window, worth somewhere between $569 million and $641 million depending on which tracker you trust. Even the low estimate lands near 44 percent of the entire monthly total. One token, close to half the dollars.

The size matters less than the shape. RAIN's release is linear, spread across a vesting schedule rather than dropped on one date. A linear unlock feeds the market in a thin, continuous stream. It behaves like a faucet, not a bucket tipped over at once. That is the opposite of how the $1.28 billion figure usually gets read, as if the whole sum hits a single afternoon's order book.

The percentage on the headline is not the sell pressure

The final week of August carried the token unlocks most outlets chose to feature. Three names led the lists. Humanity released 266.47 million tokens on August 25, about 7.92 percent of its float, worth $18.3 million. Huma Finance released 458.75 million tokens the next day, 16.7 percent of its float, worth $10.07 million. Plasma added 88.89 million tokens, 3.31 percent, worth $8.91 million.

Now look at the two biggest percentages. Huma unlocked more than twice the share Humanity did, yet it freed a little over half the cash. The reason is plain once you check the price history: both tokens had already collapsed. Coinliva covered Huma Finance when its token had fallen more than 80 percent, and Humanity after a 90 percent slide from its raise. A 16.7 percent unlock of a token that lost most of its value is a smaller event, in dollars, than a 7.92 percent unlock of one that held a bit more.

Percentage of supply tells you about dilution. It says nothing about how many dollars want out the door. Those are different questions, and the headline number keeps answering the one nobody asked.

The final-week releases lined up like this, with the month's one large cliff added for contrast:

A cliff into a thin book is the shape that hurts

If you want the token unlocks that genuinely move a price, look for a cliff landing in a shallow order book. August had one. On August 5, Succinct's PROVE token released 100 million coins in a single event, a supply shock near 51 percent against a float estimated around 195 million. The release roughly doubled the tokens in circulation overnight.

The danger sat in the book underneath it. At the moment of the unlock, Binance showed only about $100,000 of depth two percent below the price on PROVE, and Bybit was thinner still. A cliff that size dropping into liquidity that shallow is the setup that produces a real gap. It is nothing like a slow faucet, even though PROVE's roughly $17 million release looks small next to RAIN. The token traded near $0.17 through the event. For why the curve matters more than the calendar date, our explainer on vesting, cliffs and supply shocks lays out the mechanics.

Unlocks still matter, just not the way the table ranks them

None of this makes token unlocks harmless. A drip is still persistent supply, and a token already down 90 percent can keep falling, as Humanity has shown all year. People who receive unlocked tokens tend to sell them, on a cliff or a curve, and a market with no fresh demand feels even a slow stream. The point is narrower than dismissal. A table that ranks token unlocks by headline dollars or by raw percentage will keep pointing you at the wrong risk.

The more useful sort is by shape and by the depth of the book each release lands in. A linear $600 million stream can weigh less than a $30 million cliff that doubles a thin float. Plasma fits the same lesson from a different angle: we flagged its $883 million valuation against $573 of daily chain fees, and its 3.31 percent unlock this week barely mattered beside a chain that earns almost nothing. For the next batch of token unlocks, the figure to check first is the release curve and the order book, not the total pinned to the top of the calendar.

The mechanism sits underneath a growing share of the market. Ethereum alone had close to 39.6 million ETH locked in its staking contract by the middle of 2026, roughly a third of all the ETH in circulation, spread across about 1.24 million validators according to data compiled by news.bitcoin.com. Solana runs even hotter, with more than two-thirds of its circulating supply staked. So this is not a fringe activity. It is how two of the largest chains in the world stay secure.

What you are doing when you stake

A proof-of-stake chain does not use miners burning electricity to decide which transactions are valid. It uses validators who post collateral. That collateral is the stake. If a validator does its job, following the rules and staying online, the network pays it a reward. If it cheats or goes offline at the wrong moment, the network can take part of that collateral away. The stake is both the ticket to earn and the thing at risk, which is the whole point. It makes attacking the chain expensive for the attacker rather than free.

Most people who stake never run a validator themselves. Running one on Ethereum means putting up 32 ETH, keeping a machine online around the clock, and accepting penalties if you slip. So the market built easier doors. You can hand your tokens to a pool or an exchange that runs the hardware for you and passes back most of the reward. You can use a liquid staking service that gives you a tradeable receipt token in return. Or, more recently, you can buy a fund that stakes on your behalf and pays the yield out as cash. Each door trades a little control or a little yield for a lot less hassle. To understand why the chain needs validators at all, it helps to read how proof of work and proof of stake reach agreement in the first place.

Where the rewards actually come from

This is the part marketing pages skip. Staking rewards are not paid out of thin corporate profit. They come from two places: freshly issued tokens that the protocol creates and hands to validators, and a cut of transaction fees plus the value validators extract from ordering transactions, often called MEV.

The issuance part matters because it dilutes everyone who is not staking. If a network prints new tokens to pay stakers, and you hold that token without staking, your slice of the total supply quietly shrinks. So the real return on staking is the reward rate minus the network's inflation rate, not the flashy number on the dashboard. Ethereum's base staking yield sat near 2.7% a year in mid-2026, while the network's supply was growing at roughly 0.83% annually in a quiet market. The gap between those two figures, not the 2.7% by itself, is closer to what a staker truly gains against a holder who sits still.

Fee and MEV income is the healthier half of the reward, because it is real money paid by people using the chain rather than new supply invented to pay you. When a network is busy, that half grows. When it is quiet, staking becomes mostly a game of not losing ground to inflation.

The ways people stake, and what each one costs

The four common routes are not equal. Solo staking gives you the full reward and full control, and asks for capital, uptime, and technical care in return. Pooled or exchange staking is simple and skims a fee, and it means someone else holds the keys. Liquid staking hands you a token like stETH that keeps earning while you trade or lend it, at the cost of extra smart-contract risk and, sometimes, a price that drifts below the asset it represents. Fund-based staking, the newest door, wraps all of it in a regulated product and pays you in cash, which is why Grayscale began paying staking yield as cash across several of its crypto products.

Liquid staking is worth a closer look, because it has grown into its own layer of the market. The receipt token can be lent, used as collateral, or restaked to secure other services on top of the base chain. That is powerful and stacks risk on risk. When one restaking network paid out on a new receipt, the numbers were real but so was the added complexity, as the weETH restaking payout showed. Each layer you add is another set of contracts that has to hold.

How staking can cost you

Slashing is the penalty people fear most and see least. On Ethereum it punishes validators for double-signing or serious downtime by burning part of their stake. On Solana, slashing for double-signing exists but stays rare, with small penalties. For anyone staking through a pool or a fund, this is the operator's problem to avoid. A careless operator can still cost you, though.

The quieter risks bite more often. Your tokens are locked. Ethereum makes you wait in an exit queue to withdraw, and Solana imposes a two to three day cooldown before staked SOL turns liquid again. During that wait the price can move hard, and you cannot sell. A yield of 5% means little if the token drops 20% while you are stuck in the unbonding line.

Then there is centralization, which is a risk to the whole network rather than to your balance directly. Lido, a single liquid staking provider, controlled around a quarter of all staked ETH and more than 60% of the liquid staking market by 2026. Solana has seen the same pull toward a few large operators, to the point where one provider held 27% of staked SOL. Concentration like that makes the chain easier to pressure, and it is the reason solo staking, for all its friction, still matters.

Last, the scams. Because staking sounds like a safe yield, it is a favorite wrapper for fraud. Fake staking sites promise fixed double-digit returns and simply take deposits. One XRP staking scam pulled in 8.5 million dollars before investigators traced the funds. XRP, for the record, is not a proof-of-stake token and has no native staking, which is the tell. If a product offers staking on an asset that cannot be staked, or guarantees a rate no honest validator could pay, it is not staking.

Ethereum and Solana, side by side

The two biggest staking markets show how different the same idea can look. Ethereum pays a lower rate to a very large and dispersed validator set, with a long queue on the way out. Solana pays more, from a smaller validator set, with a short cooldown and much higher participation.

Frequently asked questions

Is staking safe?

Safer than most crypto activities, but not risk free. The token itself can fall in value while your stake is locked, and the way you stake adds its own risks, from a custodial operator to a smart-contract bug. The network mechanics are sound. The wrapper around them is where losses usually happen.

Can I lose my staked coins?

Through slashing, yes, though for ordinary stakers using a reputable pool it is uncommon. The larger practical loss comes from price movement while your tokens are locked and you cannot sell. Fraudulent staking platforms are a separate danger and can take everything.

Do I get rewards forever?

As long as you stay staked and the validator behaves, rewards keep accruing. The rate is not fixed, though. It falls as more tokens get staked and rises when fewer do, and both Ethereum and Solana have floated proposals to trim issuance over time. Treat the current yield as a snapshot, not a promise.

What is the difference between staking and a savings account?

A bank pays you from its lending profit and lets you withdraw on demand. A chain pays you from new issuance and fees, and makes you wait to exit. The yields can look similar, yet a staking reward that trails the token's inflation is not really income, and a savings deposit does not drop 20% overnight. Knowing where the money comes from is what separates a real return from a number on a screen.

Bitcoin crossed $80,000 on August 25, its first print above that mark in three months, and finished the month roughly 27% higher. The Bitcoin rally that dragged the price out of a six-week range under $67,000 to an intraday high of $81,265 looked like conviction coming back into the market. Most of the buying was not a choice.

By the close of the week before, more than $4 billion in bearish positions had been liquidated. Bloomberg logged a record $2.7 billion in short liquidations on August 19 alone, the day the range broke. CoinDesk counted roughly $3 billion cleared as the price topped $71,000, then another billion the session it passed $75,000. This was forced buying stacked on forced buying, the kind of move that packs its gains into short violent windows and then has nothing left to run on once the shorts are gone.

The buyback that moved it starts September 9

The catalyst everyone named was a Treasury decision to expand its purchases of longer-dated government debt. Read the fine print and the timing gets awkward. The liquidity-support buybacks double to $4 billion per operation from $2 billion, and they run from September 9 through November 4. Treasury Secretary Scott Bessent said the pace could go higher. None of it has happened yet.

So the market priced a bond-market plumbing change weeks before the first operation, off a cash balance most crypto traders had never watched. The Treasury General Account now holds about $950 billion, against a $550 billion to $600 billion target under the prior administration. The 30-year yield touched 5.337% before the announcement, its highest since 2007, then eased toward 5.18%. Bitcoin and gold both caught a bid on the same headlines, the old debasement reflex that treats a wall of new dollar liquidity as a reason to own scarce things.

How the squeeze built through the week

A familiar line stopped the run near $81,000

The top was not a coincidence. Bitcoin's 50-week moving average sat at $81,085 on Tuesday, and the intraday high of $81,265 tagged it almost to the dollar before the price slid back toward $79,000 the next day. That average has capped earlier attempts to break higher, and it held again. A Bitcoin rally built on covering rather than fresh demand tends to stall exactly where the last sellers are waiting.

Spot Bitcoin funds did take in real money alongside the move, more than $2.5 billion across six straight days of inflows through August 24, with $337.56 million arriving on the 24th. Yet total ETF assets jumped from $78.67 billion to $98.56 billion in a single week. Most of that gap is price, not new deposits, the same distortion that showed up when the funds grew $23 billion and only $2.6 billion was new cash.

For now the Bitcoin rally rests on two things that do not usually hold a price up for long: a technical ceiling it just failed to clear, and a Treasury operation that has not run a single time. Whether real spot demand shows up to replace the squeezed shorts before September 9 is what decides whether $80,000 holds.

Avihu Levy, StarkWare's first employee and the head of its applications work, built the method. It leans on a trick called signature grinding. Rather than accept the first valid signature a wallet produces, the software brute-forces millions of candidates until it finds one whose shape never puts the public key where a future quantum computer could read it. That grinding is slow, and it is not free.

An eight dollar payment that cost hundreds

The transaction fee alone came to 5,179 satoshis. The offchain computation behind it ran to roughly $75 to $150 by Levy's own account, and pushing the whole thing through cost several hundred dollars. For a payment worth eight. Price is the smaller problem. Bitcoin nodes relay only the transaction formats they already recognize, and this one looked odd enough that an ordinary node would drop it on sight.

So it never traveled the network at all. Levy handed it straight to a miner. MARA Pool took it through Slipstream, its private channel for nonstandard transactions, and mined it directly into the block. A quantum-safe Bitcoin transaction sits on-chain today because one large mining operation agreed to carry it past the network that would have refused it.

The seven million coins it cannot touch

Signature grinding only helps a coin while its public key stays hidden behind a hash. A lot of bitcoin no longer qualifies. Estimates put around 7 million BTC in categories where the key is already visible, and a quantum-safe Bitcoin transaction does nothing for any of them, because the thing a quantum attacker would need is public knowledge already. No machine capable of breaking that key exists yet, which is the only reason those coins are still sitting untouched.

Coinliva read the same story off the chain's own data months back: about a third of all bitcoin has already shown its public key on-chain. Migrating those coins into a hash-protected output does not close the gap either. The migration is itself a standard transaction, and it exposes the sending key on the way out.

Why the real fix runs through the miners

StarkWare framed the result as proof that guarding holdings never required a protocol change. Then its own chief executive, Eli Ben-Sasson, said the quiet part out loud. "A soft fork should happen, and I believe it will." A soft fork is not StarkWare's to grant. It needs miners to signal support over months, the same slow gate that has stranded other Bitcoin proposals this year. The grinding demo is a workaround for anyone who can afford it and knows a cooperative miner. It is not a network that defends itself.

Bitcoin mining has had a strange year on its own terms. Difficulty fell year on year, something that had happened only once before. Quantum defense now joins the list of upgrades that ask miners to move before the danger can be measured on any chart. Bitcoin still carries no protocol-level guard against a quantum break, and one clever transaction does not change that. The transaction in block 964,199 proved a narrow point. The coins that most need protecting still do not have it, and the network that would deliver that protection has not yet been asked to vote.

City Minister Lucy Rigby framed the change as a growth measure, saying tokenisation and distributed ledgers could reshape financial markets and the Bank should keep pace. Deputy Governor Sarah Breeden backed it. The wording carries more than the endorsements do: this is a secondary objective, so it sits under the Bank's primary job of guarding financial stability and can be overruled by it whenever the two collide.

The cap the Bank proposed, then walked back

The turn reads sharper next to what the Bank spent two years drafting. Its proposed rulebook would have capped individual stablecoin holdings at 20,000 pounds and business holdings at 10 million pounds per coin. Industry pushback was blunt. No firm could run settlement at scale while policing every account balance against a ceiling, they argued, and a sterling stablecoin built under those limits would struggle against the dollar tokens already moving billions a day.

In June the Bank dropped the individual caps. It swapped them for a temporary ceiling of 40 billion pounds on how much any single systemic sterling stablecoin can issue, and cut the reserves that issuers must park in non-interest-bearing central bank accounts to at least 30 percent. The table tracks what changed.

A mandate on paper, not a green light

Britain has run this play before. In 2023 Parliament handed the Financial Conduct Authority a secondary objective on competitiveness and growth, and two years on the industry still argues about whether the regulator behaved any differently. A duty to have regard to innovation is not a duty to approve anyone who applies. The one hard hook attached to the new objective is a yearly report to Parliament on what the Bank did with it.

Other countries drew their lines first

Rivals moved earlier and set firmer terms. The European Union's MiCA regime cleared hundreds of licensed firms and pushed unlicensed issuers out, and Japan's stablecoin framework wrote bank-style backing rules from the start. On the euro side, France has been backing its own euro stablecoin push.

The timeline is the part the headlines skip. Applications from firms that want to run a systemic sterling stablecoin are not expected to open until the end of 2026, and the 40 billion pound ceiling is written as temporary. So the reversal is real in law and still untested in practice. What a UK stablecoin issuer can build under it gets decided next year, not this one.

The idea is older than the hype around it. Ethereum can process a limited number of transactions per second on its base layer, and when demand spikes, the price of that space climbs fast. Rollups exist to move the heavy work somewhere cheaper while keeping Ethereum as the court of final record. This piece walks through what that actually means, how the two main designs differ, and why fees fell off a cliff in 2024.

Ethereum's block space is small on purpose

Every node in the network re-runs every transaction to check it. That redundancy is what makes Ethereum hard to cheat, and it is also what caps throughput. There is only so much a global set of home computers can verify per block. When a popular mint or a market panic fills those blocks, users bid against each other and gas fees spike into the tens of dollars.

The base layer can grow that capacity a little. A recent upgrade path lifted Ethereum's gas limit, which is roughly how much computation fits in a block, and even that modest change reshuffled long-standing assumptions about the network, as our look at the Glamsterdam gas rule showed. But raising the limit endlessly would push out the smaller machines that keep the chain decentralized. So the network took a different route. Instead of doing more work on Ethereum, do it somewhere else and post the receipts back.

How a rollup actually works

A rollup runs its own chain. Transactions happen there, fast and cheap, because a single sequencer orders them rather than a whole planet of validators. Then the rollup takes a large batch of those transactions, compresses the data, and writes it to Ethereum along with a claim about the new state of its ledger.

That last part is the whole trick. The rollup does not ask you to trust its sequencer. It posts enough information to Ethereum that anyone can reconstruct what happened and, crucially, that Ethereum itself can enforce the outcome if the operator lies. Hundreds of trades collapse into one L1 footprint, and the cost of that footprint is split across everyone in the batch. Security is inherited from Ethereum. Speed and price come from doing the busywork off chain.

The split between the two families of rollups comes down to a single question. How does Ethereum know the batch is honest?

Fraud proofs and the seven day wait

An optimistic rollup takes the batch at its word. It posts the data and says, in effect, this is valid unless someone proves otherwise. That someone is a watcher running a fraud proof. If the batch contains an invalid transaction, a challenger submits evidence, Ethereum re-checks the disputed step, and a bad batch gets thrown out.

The catch lives in the waiting. Because the network has to leave time for an honest challenger to notice fraud and respond, withdrawals back to Ethereum sit behind a challenge window. On the two largest optimistic rollups, Arbitrum and Optimism, that window runs about seven days. Move funds within the rollup and it feels instant. Pull them out to the base layer and you wait a week, unless you pay a third party to front the money. Arbitrum, Optimism, and Base all use this model.

ZK rollups prove every batch with math

A zero-knowledge rollup flips the assumption. Nothing is accepted until it is proven. For each batch, the rollup generates a validity proof, a compact piece of cryptography that mathematically demonstrates the new state follows from the old one under the rules. Ethereum verifies the proof and updates. No batch, no proof, no update.

Because the proof settles correctness on the spot, there is no week-long challenge window. Exits can finalize in hours rather than days. The cost sits elsewhere. Generating those proofs is computationally demanding, and building a ZK system that runs general smart contracts took years longer than the optimistic approach. zkSync, Starknet, Linea, and Scroll live in this camp.

A side by side comparison

Blobs made rollup data cheap in 2024

For a while, rollups had a dirty secret. Posting their batch data to Ethereum was expensive, because they were renting the same permanent storage as everything else on the base layer. That cost got passed to users. A cheap rollup transaction still carried a not-so-cheap slice of L1 data fees.

EIP-4844, which shipped with the Dencun upgrade on March 13, 2024, changed the math. It introduced a new transaction type carrying blobs, chunks of data around 128 kilobytes each that Ethereum stores temporarily and prunes after roughly 18 days. Rollups do not need their data kept forever. They need it available long enough for anyone to verify a batch. Blobs give them exactly that, in a separate fee market from regular gas.

The effect was immediate. One analysis measured an 81% drop in the calldata that optimistic rollups had to pay for. Median fees on Arbitrum, Optimism, Base, and zkSync fell from tens of cents to fractions of a cent almost overnight. A later modelling estimate tied to the Fusaka upgrade in December 2025 suggested rollup costs could fall a further 40 to 60% as blob capacity ramps, though that figure comes from a single projection and should be read as a forecast, not a settled result.

The tradeoffs rollups introduce

Cheaper and faster came with a cost that is easy to miss. Ethereum used to be one place. Now it is a hub with dozens of rollups orbiting it, each with its own liquidity, its own bridges, and its own quirks. Moving between them is not free and not always simple, and the fragmentation shows up in the data. Two respected trackers once measured total value on Ethereum's rollups and landed billions of dollars apart, a gap that says a lot about how young this layer still is.

There is also the sequencer question. Most rollups today run a single operator that orders transactions, which is efficient and also a point of control. If it goes offline or censors, users depend on escape hatches that vary in quality from chain to chain. The industry is working toward shared and decentralized sequencing, but the fragmented L2 economy is a real and current tradeoff, not a solved problem. Data availability is the other quiet dependency. A rollup is only as safe as its promise that anyone can fetch the batch data and rebuild the ledger, so where and how that data is published matters as much as the proofs sitting on top of it.

FAQ on Layer 2 rollups

Is a rollup the same as a sidechain?

No. A sidechain runs its own security, so if it fails, your funds are at risk on that chain alone. A rollup posts its data and proofs to Ethereum, which means Ethereum can enforce correct behavior. That inherited security is the line between the two.

Are my funds as safe on a rollup as on Ethereum?

Close, but read the fine print. The strongest rollups let anyone reconstruct state from Ethereum and force an exit even if the operator disappears. Younger ones still carry admin keys and upgrade powers that could, in theory, be misused. Trackers like L2Beat grade these differences openly.

Why does an optimistic rollup make me wait to withdraw?

The delay is the security. The challenge window gives honest watchers time to catch and prove a fraudulent batch before funds leave the system. Remove the wait and you remove the window in which fraud can be stopped.

Do ZK rollups make Ethereum obsolete?

They lean on it harder, not less. Every validity proof is verified on Ethereum, and the batch data lives there. Rollups scale Ethereum by borrowing its trust, which keeps the base layer central to the whole arrangement.

Rollups turned Ethereum from a network that priced out ordinary users during busy weeks into one where a swap can cost less than a cent. The design is not finished. Exit times, sequencer control, and the mess of moving value between chains are all live work. But the core bet has held up in practice, and for most people transacting on Ethereum today, a rollup is doing the work whether they know it or not.

XPL, the token behind the Plasma stablecoin chain, trades around $0.088. It touched $1.68 last September. The fall from that high runs to roughly 95%. On August 25 another 88.89 million XPL unlocked, worth about $8.9 million, adding fresh supply to a token that was hardly short of it.

The pitch was TRON's throne

Plasma launched with one target. It would be the chain built for stablecoins, moving USDT for free, aimed straight at TRON, where most of the world's tether already settles. Tether backed it. It arrived with roughly $2 billion in liquidity and cracked the top ten by total value locked within weeks. Zero-fee USDT transfers were the whole headline.

Money showed up, and money is still there. Plasma holds about $873 million in stablecoins today, with bridged deposits above $2.4 billion. In a single day the chain cleared more than 525,000 transactions across 36,584 active addresses. By raw activity, this is not a ghost network.

Free transfers, empty till

The design that wins users is the same one that starves the token. Free transfers are the whole point, and free means the base chain earns almost nothing from the volume it carries. DefiLlama put chain fees at $573 over 24 hours, with chain revenue at the same figure. Apps running on top of Plasma did far better on paper, pulling in around $121,000 in fees that day. They kept little of it, converting that gross into roughly $10,000 of app revenue. And none of it flows to XPL anyway. The token owns the settlement layer, and the settlement layer is free by choice.

The deposits tell a matching story. Stablecoins worth $873 million sit on the chain, yet on-chain trading is thin, with daily DEX volume near $3.6 million. Capital parked for yield behaves differently from capital used to pay for things. The same split showed up when a billion dollars of USDe rotated out of a delta-neutral trade and into loans. Balances chase the best return. They do not always move.

Large balances are not payment volume

None of this makes Plasma a failure. The chain works, the transfers are real, and a network that keeps its stablecoin balances through a 95% token drawdown has something to point to. The problem is narrower. A valuation near $883 million, or even a market cap closer to $240 million, is pricing in fees the chain has chosen to give away. Markets misjudge what a token holds or earns in both directions, as they did when StablecoinX held $253 million in ENA against an $87 million valuation.

What would change the read is payment flow the token can eventually monetize, or a decision to stop giving the base layer away. Neither is here yet. The stablecoins that count are the ones that turn over, and June set a record when USDC moved 16 times its own supply on chains people actually settle on. Plasma's balances are large. Whether they start to churn like that, instead of sitting, is the figure to track as the next XPL unlocks arrive.

Of that $23.3 billion in asset growth, only about $2.6 billion was money investors actually put in. The rest, close to $20.7 billion, was the coins already sitting in those funds becoming worth more. Roughly 89 cents of every dollar the crypto ETFs added last week came from price, not from new buyers.

The distinction gets blurred constantly. A fund's assets under management climb for two unrelated reasons: fresh cash walks in the door, or the tokens inside appreciate. Only the first is demand. Crypto ETFs report the two together as one AUM figure, and the figure breathes with the coin price whether or not a single new share is created.

A crypto ETF simply holds the coin for you and marks it to the market every minute. Last week the rally did the heavy lifting, and the inflow figure, real as it was, rode underneath a far larger move in those marks.

The new money was real, just smaller than it looked

None of this makes $2.6 billion trivial. It was the strongest weekly inflow into the crypto ETFs since October, and Bitcoin funds took the larger share. Spot Bitcoin ETFs pulled in $1.92 billion. Ether funds added $697 million.

BlackRock's IBIT did most of the Bitcoin work, taking $1.33 billion across the week and, on its busiest day, about 83 cents of every dollar that entered the category. Fidelity's FBTC added $293 million, with the rest scattered thinly across smaller issuers. When one fund absorbs that much of a record, the record is narrower than the word suggests. The buyers behind it are not one crowd either; big banks added to their Bitcoin ETF positions last quarter even as some institutions trimmed.

Ether's assets jumped hardest, and mostly on price

Ethereum's funds show the effect in sharper relief. Their assets rose 35.9%, from $10.5 billion to $14.3 billion, a steeper climb than Bitcoin's 25.4%. Yet Ether ETFs took in far less cash, $697 million against Bitcoin's $1.92 billion.

The gap is price. Ether ran from under $1,900 to above $2,500 on the week, close to 30%, while Bitcoin gained around 24%. The asset that rose more collected less fresh money. Its funds grew faster anyway because the coins already inside them repriced harder.

There is a fairer way to read the Ether figure, and it points the other direction. Measured against the size of the funds it entered, $697 million is a heavier bite than Bitcoin's haul: roughly 4.9% of Ethereum ETF assets versus 2.0% for Bitcoin. The Ether products are younger and smaller, with $12.17 billion in cumulative inflows since launch against Bitcoin's $62.43 billion, so each new dollar lands with more weight. Marginal appetite is tilting toward Ether, from a much lower base.

Where the extra $20 billion came from

Weekly flow trackers report net creations, the shares an issuer mints when cash arrives. Those numbers are clean and easy to headline. Total assets sit on a separate line and move with the coin price second by second.

Splice the two together and you get "$2.6 billion in, assets up $23 billion," which reads as a flood of demand when most of the sum is the market revaluing what was already held. By August 26 Bitcoin sat near $78,500 after touching $81,235, and Ether hovered around $2,470. The marks that inflated last week's assets were still largely in place, so the AUM line held without anyone buying or selling a thing.

Why the record framing keeps repeating

The packaging invites it. A single week that pairs a genuine inflow with a much bigger price gain will always throw off a dramatic AUM number, and the AUM number is the one that fits a headline. It is the same trap that made a 70x jump in Solana ETF inflows sound enormous when the flows were a rounding error next to that chain's rally.

The cause last week was specific. Bitcoin's run from roughly $62,000 followed the Treasury doubling its long-dated bond buybacks and a wave of short liquidations that forced prices up, not the arrival of a new cohort of fund buyers. Spot demand through the crypto ETFs trailed the price rather than leading it, the same dynamic that has kept the Coinbase premium negative for long stretches this year while the funds quietly absorbed supply.

The useful move now is to watch the two lines apart. If price stalls and net creations into the crypto ETFs keep landing, that is demand doing the work, and the institutional case holds. If the cash thins the moment Bitcoin stops climbing, last week was a mark-to-market event wearing an inflow record's clothes. With the Treasury catalyst already in the price, the coming weeks will settle which of the two it was.

The rule is 21,000. That is the gas a plain ETH transfer has always cost, a constant so reliable that it got hardcoded into thousands of tools. Glamsterdam keeps it for most transfers and quietly retires it for the rest.

The one number every wallet baked in

The change comes from EIP-8037, one of ten proposals bundled into Glamsterdam. It adds a second gas dimension the network has never had, called state gas, meant to charge for the cost of permanently growing Ethereum's state rather than just for computation.

Send ETH to an address that already exists on chain and nothing changes. The transfer still costs 21,000 gas. Send it to a brand new address and the network now has to create a fresh account entry, and that creation carries its own state gas charge on top of the base fee. Same action, two very different prices, decided by whether the recipient has ever touched the chain before.

For a user this is invisible. For software that assumed a single fixed figure, it is the ground shifting.

What actually gets more expensive

The repricing reaches past transfers. Under the Glamsterdam parameters developers have been testing, operations that write new data to state climb sharply, while reads and existing accounts stay flat. The exact multipliers are still moving between test rounds, so the figures below are drawn from the parameter sets floated during testing rather than a locked final schedule.

Estimates for new account creation have landed anywhere from about seven times higher in one write up to roughly 8.5 times under an earlier parameter set, which is itself a sign of how live these numbers still are. A separate proposal in the same bundle, EIP-7954, pushes the other way and raises the maximum deployed contract size from 24 KiB to 64 KiB. The upgrade is not simply making things pricier. It moves where the costs sit.

Why the Foundation started emailing developers

The warning was blunt. Software that hardcodes a maximum gas value, or assumes Ethereum has one fixed gas ceiling, will break once Glamsterdam activates. The Foundation named the categories most exposed: wallets that pre fill 21,000 for a send, indexers that decode transactions against a fixed model, and gas estimators that quote a flat number before signing. If those tools underquote a transfer to a new address, the transaction simply fails.

This is why the message arrived with a testnet and not with mainnet. The Glamsterdam fork went live on the Platåberget test network on August 20, a short lived environment the Foundation opened for exactly this kind of shakeout. Teams can watch their estimators misfire there, where the only cost is a failed test transaction, and patch before real value moves.

A mainnet slot has not been set. Past forks spent two to four months seasoning on public testnets such as Hoodi before activation, which points to a launch in the back half of 2026. That gap is the point of the early notice. Glamsterdam's 200 million gas number will get the coverage, and it earns it, yet the change most likely to reach an ordinary user first is the one deciding whether their wallet still quotes gas correctly. For a walkthrough of where these fees come from, our explainer on how gas fees are priced and who collects them covers the mechanics this upgrade is rebuilding. The extra headroom is aimed squarely at rollups, the same layer where two trackers still disagree on Ethereum L2 value by 29 billion dollars, and it fits a longer run of protocol changes that includes Vitalik Buterin's push to make Ethereum transactions private.

The names sound interchangeable. They are not. Proof of work vs proof of stake is really a choice between spending electricity and locking up money, and that single difference ripples out into cost, speed, security, and who ends up holding power over the network. This guide starts from zero and walks through both.

The problem both are trying to solve

Picture a shared notebook that anyone can write in. If two people try to spend the same ten dollars at the same instant, whose entry counts? A bank would just pick one. A blockchain has no one to pick, so it needs a rule that lets a crowd of anonymous participants settle on a single version of history and keep settling on it, block after block, forever.

That rule is called a consensus mechanism. Its job is to make cheating expensive. If rewriting the ledger costs more than you could ever steal by rewriting it, honest behaviour becomes the rational choice, and the notebook stays trustworthy. The two systems in this guide are just two ways of attaching a real, painful cost to that decision. One uses energy. The other uses capital.

Proof of work: burn electricity to earn the right to write

Bitcoin invented this model in 2009, and it is still the one most people picture when they think of mining. Computers across the world race to guess the answer to a math puzzle. The puzzle is deliberately hard and deliberately useless, a giant guessing game where the only way to win is to try trillions of combinations per second. The first machine to land on a valid answer gets to add the next block and collect the reward.

The electricity is the point, not a side effect. To attack the network you would need to out-compute everyone else combined, which means buying more hardware and paying more power bills than the entire honest crowd. That is the wall protecting the chain. It is also why the whole thing is so hungry: the Bitcoin network alone draws somewhere between 130 and 160 terawatt hours a year by current estimates from Cambridge and Digiconomist, roughly what a country the size of Poland or Argentina uses to keep the lights on.

Mining has quietly turned into heavy industry. Warehouses of specialised chips, cheap power contracts, and razor-thin margins now decide who stays profitable, and some operators are eyeing whether renting those machines to artificial intelligence pays better than mining bitcoin at all. The security is real and battle-tested. The cost is the electricity meter that never stops spinning.

How the staking model replaces the power bill

Proof of stake asks a different question. Instead of proving you spent electricity, you prove you have money on the line. Participants called validators lock up a chunk of the chain's own coins as collateral. The network then picks who writes the next block, usually at random but weighted by how much each validator has committed. Behave honestly and you earn a yield. Try to cheat, approve a fake block, go offline when you should be working, and the network destroys part of your stake. That penalty has a name that tells you everything: slashing.

Ethereum is the headline example. It launched on proof of work, ran that way for seven years, then switched over in a single event on September 15, 2022 that the community called the Merge. Running your own validator on Ethereum takes 32 ETH, though pooled and liquid services let smaller holders join with far less. If you want the fuller picture of how locking coins earns rewards, our explainer on staking in crypto covers the mechanics.

The trade the network makes is straightforward. It stops paying for security in kilowatts and starts paying for it in capital at risk. No warehouses, no puzzles, no power bills. Just money that vanishes if you misbehave.

The two side by side

The clearest way to see the gap is to line the properties up. The table numbers reflect the mainstream implementations, Bitcoin for proof of work and post-Merge Ethereum for the staking side.

Where the 99.95% energy gap comes from

That figure gets thrown around a lot, so it is worth understanding why it is real rather than marketing. When Ethereum abandoned mining, the Ethereum Foundation reported its energy use fell by roughly 99.95%, and an independent carbon study put the emissions drop near 99.99%. The reason is simple. Proof of work needs millions of machines all guessing at once, and only one wins each round, so the rest of that electricity is spent losing. The staking model needs each validator to run one ordinary computer that could sit under a desk. There is no race, so there is no wasted heat.

This is the single biggest argument in favour of proof of stake, and it is why most newer chains never touched mining at all. It is also why the environmental criticism that dogged crypto for a decade lands much harder on Bitcoin than on the rest of the market today.

What each model gives up

Neither design is free of trade-offs, and the honest version of this comparison admits both have real weaknesses.

Proof of work's strength is its brutal simplicity. The cost is external and physical. You cannot fake a power bill, and after fifteen years nobody has broken Bitcoin's core ledger. The knock against it is the energy, plus the way mining has concentrated into a handful of industrial players with access to the cheapest electricity on earth.

The staking side trades that physical cost for a financial one, and critics argue it can drift toward a rich-get-richer loop, since the more coins you commit the more rewards you earn, and the more you can commit again. Validators also have to wait in line to join or exit, and that queue can stretch for weeks when demand spikes, as it did when the Ethereum staking queue ran to 42 days. Speed is where these chains tend to pull ahead. Newer designs can confirm blocks in well under a second, and Solana's recent Alpenglow upgrade pushed its finality claims into the sub-second range. The security is younger and more complex, which is the fair counterpoint, with fewer years of adversaries trying to break it.

So which one wins? Neither, and that is the useful answer. Bitcoin will almost certainly stay on proof of work, because its whole value proposition rests on the fact that its security is anchored to something physical and unchangeable. Most of everything else has moved to proof of stake or was born there, chasing lower costs, faster blocks, and a cleaner energy story. The choice tells you what a chain values. A network that prizes maximal, boring, energy-backed security picks work. A network that wants speed and low overhead picks the other path.

Frequently asked questions

Is proof of stake less secure than proof of work?

It is younger, not proven weaker. Ethereum has secured hundreds of billions in value under this model since 2022 with no successful attack on its ledger, and proof of work simply has the longer track record, which counts for a lot when real money is at stake.

Can a blockchain switch from one to the other?

Yes. Ethereum is the proof, since the Merge swapped its entire consensus engine in place without stopping the chain or resetting anyone's balance. It was one of the more complex upgrades in the industry's history. It worked on the first try.

Do I need special hardware to take part?

For mining, yes. You need purpose-built rigs and cheap power to have any realistic chance of profit. On the staking side you need coins instead, and while running an Ethereum validator solo takes 32 ETH, pools let you contribute a fraction of that and share the rewards.

Why does anyone still use proof of work if it uses so much energy?

Because the energy is the security. That cost is exactly what makes attacking Bitcoin pointless. Supporters do not see the power draw as waste. They see it as the price of a ledger no single party on earth can quietly rewrite, and after fifteen years of the thing holding, they are happy to keep paying.

The figure is real. It is also the least useful number in the story, because it bundles holders with every reason to sell together with holders who have none.

The headline reads $1.2 billion

At prices near the record, 14.18 million tokens come to roughly $1.2 billion. Against a market cap near $19.5 billion, the release is about one part in sixteen of everything HYPE is worth, and around 1.4% of the fixed one billion token supply. Set against a token that has been setting records all year, a slug that size in a day is the kind of event traders brace for. The bracing may be aimed at the wrong number.

Under half of it goes to early backers

Split the HYPE unlock into its buckets and the picture shifts. Early investors and insiders take 46.6%, worth about $560 million. A near-equal 46.3% is community allocation, tokens spread across users and rewards rather than a single desk. The Hyper Foundation keeps the remaining 7%.

The private, early-money slice is the part with the clearest reason to take profit into a record price, and it is under half the headline. Call it 6.6 million tokens rather than 14.18 million. The community and foundation shares can sell too. But folding all three into one $1.2 billion scare figure treats a foundation treasury and a venture fund as the same seller, and they behave nothing alike. That is the first thing the HYPE unlock coverage skips.

What the buyback buys in a month

HYPE carries a structural bid that most tokens lack. Hyperliquid routes nearly all of its trading fees into an Assistance Fund that buys HYPE on the open market, automatically, every day. That mechanism is a large part of why the token trades where it does, and it is the number the HYPE unlock has to be measured against.

Here is the pace. Over the last 30 days the protocol booked about $44 million in revenue, per DefiLlama, an annualized run rate near $726 million. Near $80 a token, that buys on the order of 540,000 HYPE a month. Now hold it against Friday. The early-investor tranche alone, 6.6 million tokens, is close to a year of buybacks landing at once. The full 14.18 million is more than two years of it.

The longer arc is starker. Since launch the fund has accumulated roughly 28.5 million HYPE, spending well over a billion dollars to get there. This one unlock is about half of everything the buyback has ever bought, released on a single Friday.

None of that means HYPE has to fall. Cliff unlocks are set months ahead, and vested tokens rarely all reach exchanges on day one, as the mechanics of vesting and cliffs tend to spread the impact. The figure worth tracking sits below the $1.2 billion on every screen. Watch the 6.6 million investor tokens, whether they move at all, and how much of the daily buyback is left to meet them once the week's fees clear.

Seven private keys sat on one infected laptop

The weak point was never the cryptography. On June 8 an attacker reached seven private keys held by a member of the Humanity Foundation, keys left on a malware infected developer machine. What followed was quick. Around 141 million H were drained from Humanity Protocol's Ethereum bridge and more were minted on BNB Chain, close to $36 million. Quantstamp, which reviewed the incident, tied the methods to North Korea linked crews.

This was not a clever contract exploit. A study of last year's hacks found that code bugs cause about 60 percent of incidents but a sliver of the money lost, since the big sums walk out through stolen keys and social engineering. A MiCA licensed stablecoin lost its whole reserve to one compromised key this year, and the crews Quantstamp named have taken far more, including $578 million from DeFi in an 18 day run. The team answered with a token swap, retiring the old contracts and airdropping a new audited ERC-20 across three chains.

Under 3 percent of supply moved, and 80 percent is still waiting

The August release looked large in one frame and routine in another. Trackers called it 7.92 percent of released supply, a number built to alarm. Measured against the 10 billion token cap, the same 266 million coins come to under 3 percent, one monthly tranche in Humanity Protocol's long vesting plan.

The heavier fact is how little supply is loose. Only about 1.99 billion H circulate, near a fifth of the maximum, while more than 8 billion coins wait in the schedule ahead. That gap is why the market value sits at a fifth of the diluted figure, and why each release meets holders who would rather sell than wait.

A trust product built on a token its owner can change

There is an awkward shape to this. Humanity Protocol asks people to hand over a biometric and trust that the network stayed honest and hard to capture. The June breach came from the plainest failure in the book. A laptop, some stolen keys. And the replacement token, by CoinGecko's own contract note, runs behind a proxy whose owner can rewrite the code, minting or freezing transfers included.

None of that settles where H goes next. The identity product may still gather users, and the palm scan idea keeps its backers. For now the release calendar matters more than the technology, with the next tranche due in late September and monthly ones behind it, landing into a token already priced well under the funding rounds.

Only 15 percent of DGAI is actually trading

Of a fixed one billion supply, about 150 million tokens circulate. That is the 15 percent behind the $105 million market cap. The other 85 percent is spoken for but not yet in the market, which is how a token worth $105 million today carries a $700 million valuation on paper. If the phrase is unfamiliar, that headline number is the fully diluted value, the price of every token as if all of them traded at once.

The published allocation, on BNB Chain, breaks down like this.

Half the supply is earmarked for node operators and emissions that arrive over time. A thin float over a large locked balance is a familiar shape in this corner of the market. CHIP jumped 26 percent on a GPU-lending thesis while 80 percent of its supply stayed locked, and the overhang eventually mattered more than the pitch.

The $20 million in revenue was membership fees

The figure repeated across the launch coverage is $20 million in revenue before listing. It is real money and it is verifiable on a public BNB Chain treasury wallet. What it is not is inference revenue. According to the launch reports, the sum came from the DGrid AI Genesis premium program, which signed up more than 13,000 paying members at an average of about $1,580 each. Multiply those and you land near $20.5 million.

Members buy network access, hardware credits, monthly token credits, model services, and a membership NFT. So the $20 million measures how many people paid to join, not what the network earned once it started routing AI requests. That distinction is the one worth holding onto, because a token's price and what its network actually earns are separate facts. MANTA showed the extreme version, where the unlocks were nearly finished and the chain earned about $13 a day. Big top-line numbers can also mask a falling token: Huma Finance moved $12 billion in volume while its token fell 82 percent.

What sits under the price

DGrid AI raised a $5 million seed in July, backed by Waterdrip Capital, IoTeX, Paramita, and Zenith Capital. The token generation event landed in mid-August, and the DGrid AI sale priced DGAI near $0.705. That last detail cuts against the 93 percent headline. By ICO Analytics' reckoning the current price is roughly a 1x return on the sale, so the pop everyone cited runs off the listing floor rather than off what early buyers paid.

Decentralized AI has produced loud debuts before, and not all of them held up under a second look. Covenant AI went as far as calling Bittensor decentralization theater. The honest read on DGrid AI is that the membership money is genuine and on-chain, the network is live, and the valuation is still leaning on the 85 percent of supply that has not arrived. The number to track from here is not the launch-day candle. It is whether inference volume and node count climb toward a $700 million valuation before those locked tokens start unlocking into the float.

That gap is the story August keeps retelling.

Common break-ins are the cheap ones

Of the 207 incidents, 125 were smart contract exploits, close to 60 percent of the crypto hacks logged. Buggy code is still the everyday attack. It is also the least expensive. By value, TRM traced about 76 percent of all stolen funds to infrastructure and operational compromise, meaning leaked keys, spoofed interfaces, and staff tricked into signing a transaction they should not have. The contracts held. The people and the plumbing around them did not.

Two attacks tied to North Korea make the point at scale. The $285 million theft from Drift Protocol in April and a $292 million hit on KelpDAO moved $643 million between them, about two thirds of the entire half year. Neither turned on a clever bug in a contract. Both turned on access.

The averages say the same thing from another angle. The mean loss per incident was $4.7 million, but the median came in at just $219,000, a spread that shows the tail is doing the damage. A pile of small crypto hacks sets the count. The access failures set the total.

August's biggest DeFi loss ran on a vote

On August 23 an attacker drained about $8.5 million from Term Finance, an Ethereum lending protocol. No contract was exploited. The wallet started with 2 ETH pulled through Tornado Cash, then bought up the project's thinly held governance token on the open market. Low float and light participation meant that a small position bought a controlling one. The attacker ended with 100 percent of the votes across four of the five USDC strategy vaults and roughly 91 percent of the Ethereum Meta Vault, then proposed sending the assets to a single address and voted the proposals through.

Out went 2,843 ETH, near $6.9 million, plus 1.68 million USDC later swapped into DAI. That ETH alone was about 68 percent of the Meta Vaults' holdings. Defimon, PeckShield, and CertiK all traced the same wallets. The two ETH that captured Term Finance's vaults did what a zero-day usually does, without touching a line of the code. Term said it is working with outside security teams to recover the assets and cover what it can.

Permissions and delegates did the rest

Term was the loudest case, not the only one. Two days earlier BounceBit lost about $3 million to an authorization flaw on its own layer-1, a failure serious enough that the team is winding the chain down and reissuing its BB token on BNB Chain. Around the same window, a delegate-permission hijack let an attacker mint an absurd nominal amount of SAND through a cross-chain messaging hook, though the real, sellable damage stayed tiny.

Each of these is a control failure wearing different clothes. Who can vote. Who is authorized. Who holds a delegate slot. The attacker never had to out-engineer the contract because the contract was told, through channels it trusted, to hand the money over. This is the same category that a single stolen key drained from a licensed stablecoin back in May, scaled up and dressed in DAO governance.

The code surface did not get safer

Read this the wrong way and you conclude that audits stopped mattering. They did not. Sixty percent of the year's crypto hacks were still contract bugs, and the record incident count means more code is under fire than ever, most of it in small protocols that never get a serious review. The typical hack is a $219,000 code slip on something few people watch.

The honest version is narrower. Code exploits win on frequency, control failures win on size, and the money follows control. That was true before August through leaked keys and social engineering, and Term Finance simply added a new door onto an old floor plan, buying the vote instead of stealing the signature. Governance was supposed to be the safeguard. Priced cheaply enough, it became the entry point.

For anyone holding funds in a protocol, the questions worth asking have shifted with it. How concentrated is the governance float, and how much would a majority actually cost to buy. Who controls the signing keys, the delegate slots, the upgrade switches. As licensing rules push more assets toward a handful of permitted issuers and custodians next year, those single points of control get larger, not smaller, and that is the surface to watch through the rest of 2026.

Bitcoin's best week since 2021 did the heavy lifting

Bitcoin gained about 23.6 percent last week, its strongest stretch since February 2021, and settled near 79,000 dollars. Ether ran harder, up 31.3 percent. The money followed the price. Bitcoin funds pulled in 1.92 billion dollars, the largest weekly haul since October, and ether products added 697 million, for a combined 2.62 billion across US crypto ETFs. The dollar index slipped to 98.9, below its 200-day average of 99.1. Treasury Secretary Scott Bessent, hinting at more intervention in the bond market, gave the whole move its macro cover.

Gold spent the year doing what bitcoin was supposed to do

The twelve-month record tells the other half. Gold sits in record territory and is up around 37 percent over the year. Bitcoin, even after the rally, remains about 37 percent below the 126,000 dollar peak it set in October 2025. The link between the two assets broke down through the first half of 2026. One analysis put their correlation near minus 0.88 in the spring, among the lowest readings since the 2022 bear market. Bitcoin had stopped trading like scarce money and started tracking stocks, its correlation with the Nasdaq flipping firmly positive. That is not how a debasement hedge is supposed to behave.

Figures use the GLD and IBIT funds as of August 24, a like-for-like way to compare the metal and the coin without arguing over which spot feed to trust. The weekly column is where they finally rhyme. Every other column shows the split that ran all year, the one the return of bitcoin versus gold as a trade keeps papering over.

What the reunion actually rests on

So the debasement trade of 2026 came roaring back on a narrow window, where a Treasury-driven squeeze and a softer dollar lifted hard assets and risk assets at the same time. Bitcoin packed much of its jump into a single violent hour as shorts unwound. Gold needed no squeeze. It had been bid all year by falling real yields and steady central-bank buying, the patient version of the same thesis that has tied bitcoin, gold, and the Fed together before. Bitcoin arrived late and on borrowed positioning, a different engine even when the direction agrees.

This is not the first time the label has been declared alive in 2026. Back in June the same debasement trade unwound in one session, dragging gold, silver, and bitcoin lower together, before the metal and the coin drifted apart again. The shape of the year has been brief bursts of correlation inside a much longer stretch of divergence. Anyone who bought the phrase instead of the timing got whipsawed, twice.

For the debasement trade to mean what its name promises, bitcoin has to hold this pace after the squeeze burns off and the ETF flows cool. The past year is the base rate here, and it favors the metal. The number worth watching now is the correlation that just turned positive, and whether it stays there once the dollar stops falling.

This guide explains what a unit of gas measures, why the same action can cost wildly different gas fees, and how to read the number in your wallet before you hit confirm.

Gas measures work, not money

Every operation on Ethereum takes computational effort. Adding two numbers is cheap. Storing data forever is expensive. Gas is the unit that measures that effort, and it is deliberately separate from price. One unit of gas always represents the same amount of work no matter what ether is trading at.

The cost you actually pay has two parts multiplied together: how much gas your transaction consumes, and the price you pay per unit. A plain transfer of ether uses exactly 21,000 gas. That number is fixed in the protocol. It does not matter whether you are sending one dollar or one million. The work of moving ether from one account to another is the same, so the gas is the same.

Contract interactions are where the number climbs. Swapping a token on a decentralized exchange, approving a contract to spend your funds, minting a collectible, each of these runs code, and code costs gas. A token approval or a swap can use several times the gas of a simple transfer because the network is doing several times the work.

Treat the middle column as typical rather than exact. The transfer of ether is the one hard number. Everything that touches a smart contract depends on what that contract is written to do, so two swaps on two different exchanges will not cost the same gas.

The two fees inside every transaction

Since a 2021 upgrade called London, Ethereum gas fees split into two pieces. The base fee is set by the protocol, not by you. It rises and falls with demand, and here is the part that surprises people: it is burned. The ether spent on the base fee is destroyed, removed from supply, paid to no one. Since London went live in August 2021, more than 4.6 million ether has been burned this way.

The second piece is the priority fee, often called the tip. This one does go to someone, the validator who includes your transaction in a block. When the network is quiet, a small tip is enough. When thousands of people are competing for the next block, raising your tip is how you jump the line.

Your total fee is simple once you see it laid out:

gas used x (base fee + priority fee)

Say you send ether, so gas used is 21,000. The base fee sits at 10 gwei and you add a 2 gwei tip. That is 21,000 multiplied by 12 gwei, which comes to 252,000 gwei, or 0.000252 ether. At most ether prices that is a fraction of a dollar. Push the base fee to 100 gwei during a busy mint and the same transfer costs ten times as much, without a single thing changing on your end.

Reading gwei without a calculator

Gwei is just a convenient way to write a very small amount of ether. One gwei is one billionth of an ether. Wallets quote gas prices in gwei because writing 0.000000012 ether on every screen would be unreadable.

The practical habit is to watch the base fee in gwei before confirming anything. Single digit gwei means the network is calm and your gas fees will be small. Fifty or a hundred gwei means demand has spiked, usually because a popular launch or a market move has everyone transacting at once. There is no fixed schedule for this. It tracks whatever is happening on-chain that hour, and it can settle back down within minutes.

The base fee also cannot lurch without limit. The protocol lets it move by at most 12.5 percent from one block to the next. A block sits at a 15 million gas target with a hard ceiling of 30 million. When blocks run above the target, the base fee ticks up the next block. When they run below, it eases off. That cap is why fees rise in a climb rather than a single vertical jump, and it is what makes the next block's fee roughly predictable instead of a blind auction.

Why a Layer 2 transaction costs cents

If you have used an app on Arbitrum, Base, or Optimism, you have paid gas fees measured in cents rather than dollars. Those networks are rollups. They process transactions on their own faster lane, then post a compressed record back to Ethereum for security. You are still ultimately paying for Ethereum block space, just a tiny sliced share of it.

A 2024 upgrade named Dencun made that share far cheaper. It introduced a separate data lane, nicknamed blobs, built specifically for rollups to dump their compressed data without competing with ordinary transactions for the same expensive space. Rollup gas fees dropped sharply after it shipped. This is the current answer to the old complaint that Ethereum is too costly to use: most everyday activity has moved to these DeFi layers on top, where the same swap that once cost 30 dollars now costs less than a coffee.

Other chains handle fees differently again. Networks that prioritize raw speed, like Solana, price transactions through a compute budget rather than a burned base fee, which is part of why their costs behave nothing like Ethereum's. A recent change to Solana's slot time and block budget shows how a chain can tune throughput and cost together at the protocol level. The lesson for a newcomer is that gas is an Ethereum word for a universal idea. Every chain charges you something for the work, but the mechanism and the size of the bill vary a lot.

FAQ on gas fees

Why did my transaction fail and still charge me?

Because the validators still did the work of trying. Gas pays for computation, not for success. If a swap runs out of gas partway or a contract rejects it, the effort already happened, so the fee is taken anyway. This is why wallets estimate a gas limit for you and why setting it too low can cause a failure that still costs money.

Can I just set my fee to zero?

You can set the priority fee to zero, but you cannot avoid the base fee, since the protocol requires it. A zero tip means validators have no reason to prioritize you, so your transaction may sit unconfirmed for a long time when the network is busy. During quiet periods a very small tip is often enough.

Do I pay gas in the token I am sending?

On Ethereum you pay in ether, even when the thing you are moving is a stablecoin or another token. That trips up people who hold a wallet full of tokens but no ether and then cannot move any of it. Keep a small ether balance in any crypto wallet you use for transactions, purely to cover fees.

Are gas fees the same across every wallet and app?

The network fee is set by the chain, so it is the same regardless of which wallet you use. What differs is how each app estimates and displays it, and whether a service adds its own charge on top. Reading the raw gwei figure, rather than a rounded dollar estimate, tells you what the network itself is asking.

Once the two-part fee makes sense, the volatility stops feeling random. A quiet Sunday and a frantic token launch are the same transaction at two different prices for the same block space, and knowing how to read the base fee tells you which one you are about to pay. That habit alone will save you money over a year of transacting. For the chain underneath all of this, start with what Ethereum is. If you interact with contracts often, learn how to revoke token approvals so old permissions are not quietly costing you later.

The unlock is set for August 26. Trackers put it at 458.75 million HUMA, about 16.7% of the circulating supply and the biggest single release among this week's schedule. In dollars it is worth close to $10 million, roughly a quarter of the token's entire $37 million market value, in one day.

Small in tokens, heavy in dollars

Sixteen percent sounds routine. The dollar math does not. HUMA changes hands for somewhere between $1.6 million and $5 million a day depending on which tracker you read, so the freshly unlocked supply is worth two to six full days of trading if every recipient tried to exit at once. The point is the ratio: the release is large relative to the market it lands in.

Who receives it matters more. The tranche breaks down as 171.67 million tokens to investors, 160.83 million to the team and advisors, and 126.25 million to the protocol treasury. Add the first two and roughly seven of every ten unlocked tokens go to insiders rather than to anything user-facing. These are the holders with the lowest cost basis and the least reason to wait, the same imbalance that drew attention when a ZRO round went 94% to insiders.

The network is not the weak part

Here is where Huma Finance separates from the usual unlock story. The protocol is not a ghost chain. Huma Finance's own August milestones claim $12 billion in cumulative on-chain transaction volume, about $160 million in active liquidity, more than 100,000 depositors, and a default rate it still reports as zero since launch. Independent research earlier in the cycle logged the volume climbing from around $2.3 billion to that $12 billion figure. Whatever you think of PayFi as a category, real money has moved through it.

The token has not followed. HUMA trades near $0.021, down about 82% from the $0.1176 it printed at launch in May 2025. Usage multiplied several times over that stretch. Price went the other direction. When activity rises and the token falls, the supply side is usually the reason.

Why the price cannot catch up

Only 17.3% of the ten billion HUMA supply is live. The fully diluted value sits near $214 million against a $37 million market cap, which means almost six dollars of future tokens hang over every dollar trading today. Wednesday's release is one step down a staircase that keeps going, and each step lands in the same thin order book. A network can grow all it likes; if new supply arrives faster than buyers do, the chart reflects the float, not the fundamentals. That is the same trap that caught an unlock hitting a market too thin to absorb it, and the reason a high fully diluted valuation reads as a warning rather than a badge.

The mechanics of vesting, cliffs, and supply shocks are worth understanding before reading any of it as a verdict on the product, and we walked through them in a recent guide on how token unlocks work. For Huma Finance the question after Wednesday is narrow. The payments business keeps posting numbers. Watch whether the token can find a level where the next tranche stops mattering, or whether the float has to finish arriving first.

The 85x compares against a wall clock users rarely watch

Solana has two kinds of finality. The 12.8 seconds everyone quotes is deterministic finality under TowerBFT, the point where a block is locked in after 32 stacked votes. Almost nobody waits for it.

In practice, a block reaches optimistic confirmation at around 500 to 600 milliseconds, once at least two thirds of stake has voted on it. According to Helius, no optimistically confirmed block has ever been rolled back since the genesis block. So the experience Alpenglow actually replaces is closer to half a second, not thirteen. Measured that way, the jump is a few times faster, not 85.

Alpenglow's own targets split into two paths. One is about 100 milliseconds, when 80 percent of stake approves a block in a single round. The other is roughly 150 milliseconds, through a slower fallback that needs 60 percent in two consecutive rounds. Solana had already been shaving slot times ahead of this, cutting per-block time toward 350 milliseconds and trimming the block budget.

Half the network's stake sat out the vote

The governance vote on SIMD-0326 passed, and it passed by a lot. Yes took 98.27 percent, no drew 1.05 percent, and 0.69 percent abstained. The number that got less attention is the last one Solana published: only 52 percent of stake cast a vote at all.

A protocol changing the way it reaches agreement got a mandate from a little over half the coins that could have weighed in. That is not unusual for on-chain governance, and it is not evidence of anything sinister. It does sit awkwardly next to how the change is being sold, as a settled decision the whole network stands behind.

Turnout also runs into concentration. Coinliva has covered how one provider held 27 percent of staked SOL against a 25 percent cap, which means a small group of large validators carries an outsized share of any vote. The rules being decided are not cosmetic, either. An earlier fight over how much SOL the network burns each day showed how much sits on these votes.

Faster agreement lowers the fault ceiling

Speed does not come free. Alpenglow tightens the fault tolerance its predecessor allowed. Classic Byzantine consensus can survive up to a third of participants acting maliciously. Alpenglow accepts up to 20 percent adversarial stake, trading some of that cushion for sub-second deterministic finality. Helius frames it as a slight reduction in Byzantine tolerance, and whether 20 percent is enough headroom is a judgment the network is making collectively, not a settled fact.

The timeline reflects the caution. Agave 4.3 was at an alpha build in mid-August and carried an explicit label that it was unsuitable for production use. A bug bounty worth up to 50,000 SOL ran from August 5 to August 19, aimed squarely at finding holes in the consensus code before it ships.

None of this means Alpenglow will fall short. It means the 150 millisecond figure is a target sitting behind a validator vote, a security review, and a set of feature gates that were still dark on mainnet as of mid-August. October is the date to watch, and the honest version of the pitch is a few times faster for real users, with a thinner safety margin underneath.

The premium was the whole engine

Here is how a bitcoin treasury vehicle should work. The stock trades above the value of the crypto it holds, a premium the industry measures as mNAV above 1. While that premium holds, the company issues new shares at the inflated price, uses the cash to buy more bitcoin, and ends up with more coins per share than before. Existing holders get richer without adding a cent. The premium funds the accumulation, and the accumulation defends the premium.

Reverse it and the machine runs backward. Once mNAV drops under 1, every share sold to buy bitcoin hands away more ownership than the coins are worth. Buying becomes dilution. The engine stalled.

The premium did not vanish by accident. When a bitcoin treasury company was one of the few ways to hold the asset inside a brokerage account, investors paid up for the access. Spot bitcoin ETFs erased that scarcity. A fund now delivers the same exposure without the debt, the dilution, or an executive team taking a cut, so the reason to pay more than the coins are worth has thinned to almost nothing. Every treasury issuing shares into this market competes with a cheaper, cleaner version of itself, and the discount is the market saying so.

Strategy stopped buying and started selling

The behavior gives it away. Since May, Strategy has sold 6,948 bitcoin for roughly $432 million, and in the week ending in mid-August it raised $334 million in fresh stock while buying no bitcoin at all. The cash went to a dollar reserve that reached about $4.8 billion on August 16, which management frames as close to 2.7 years of obligations. A company built to hoard bitcoin is hoarding dollars instead.

The quarter shows why. Strategy booked a net loss of $8.2 billion for the second quarter of 2026, almost all of it an $8.3 billion unrealized loss on its digital assets as bitcoin traded near $65,000. Its stock has fallen about 20% in a month. None of this is a solvency scare. It is a treasury that has quietly switched from offense to defense, and the raise that funded dollars rather than coins was the clearest signal yet.

Twenty One's bitcoin-per-share went backward

The tell is sharper at Twenty One Capital, the Tether-backed vehicle that sold itself on a single promise: more bitcoin behind every share, quarter after quarter. That number went the wrong way. Twenty One reported 12,547 satoshi per Class A share as of June 30, down from 12,557 at the end of 2025. Ten satoshi is a rounding error. The direction is the point.

The market has drawn its own conclusion. Twenty One holds 43,514 bitcoin worth roughly $2.8 billion, yet its equity trades near $1.6 billion, about 57 cents for every dollar of bitcoin on its books. It has pledged 16,116 of those coins as collateral and carries $485 million in convertible notes against a $106 million cash position. Where Strategy plays defense with a fat cushion, the smaller bitcoin treasury vehicles have thinner ground under them.

A metric that only points up

Bulls read the discount as a bottom. A company trading below the value of its coins can buy back its own shares and lift bitcoin-per-share for everyone left, the mirror image of the premium trade. That case has teeth for the survivors. Strategy's 2.7-year dollar runway buys it time that a debt-heavy micro-treasury does not have, and a low-debt holder can wait out a bad quarter.

The harder problem sits in the numbers themselves. As one recent analysis put it, mNAV and bitcoin-per-share never had a standard definition, and both flatter the bitcoin treasury model in one direction. When premiums invert, the same metrics that sold the story stop describing it, hiding what a company can actually do with its capital. That is why the buyback cure keeps failing to convince: Coinliva has tracked the gap between the buyback pitch and its results, watched a treasury keep buying as its own value sank, and covered a vehicle the market prices at a fraction of the tokens it holds.

The next test is mechanical. Watch whether the discounted names announce real buybacks or keep issuing shares into a market that no longer pays a premium, and whether any of them must sell coins to service debt. Strategy has the cushion to choose. Most of its imitators do not. The smaller vehicles will show, over the next few earnings dates, whether a bitcoin treasury without a premium is a going concern or a slow unwind.

This guide walks through what a crypto ETF is, how the shares come into existence, what the funds cost, and what you quietly give up by using one instead of holding the asset yourself.

You own the fund, the fund owns the coin

ETF stands for exchange-traded fund. Strip out the jargon and it is a basket that someone else fills and stores, sliced into shares you can buy and sell during market hours. A gold ETF holds gold bars in a vault. A crypto ETF holds crypto in the custody of a professional custodian, and each share represents a small, fixed claim on that pile.

When the price of the underlying coin moves, the share price tracks it, minus the fee. You get the price exposure. You do not get the coin. If bitcoin is the asset you want to understand first, our explainer on what bitcoin actually is is the better place to start, because a crypto ETF only makes sense once you know what it wraps.

Spot and futures are not the same product

The first US crypto ETF was not a spot fund. It was ProShares BITO, which launched on October 19, 2021, and it held bitcoin futures contracts rather than bitcoin. Futures track the coin loosely and carry rolling costs that eat returns over long holds. For years that was the only wrapper US regulators allowed.

Spot funds changed the math. A spot crypto ETF holds the real asset, one share backed by actual coins sitting in custody. US spot bitcoin ETFs began trading on January 11, 2024. Spot ether funds followed in July 2024. The first US spot Solana ETF, Bitwise's BSOL, listed on October 28, 2025 at a 0.20% fee, and a wave of rival Solana products filed in behind it. When you read that Solana ETF inflows hit a record, that is spot money, not futures.

Where the shares come from

ETF shares are not printed at random. They are created and destroyed through a mechanism most buyers never see, and it is the part that keeps the share price honest.

Large trading firms called authorized participants sit between the fund and the market. When demand for a crypto ETF runs hot and the shares start trading above the value of the coins behind them, an authorized participant delivers cash or coins to the fund and receives a big block of new shares in return, then sells those shares into the market. That extra supply pushes the price back down toward fair value. When shares trade cheap, the process runs in reverse: the participant hands shares back and pulls out the underlying value, shrinking supply until the discount closes.

Most US crypto ETF creations use cash rather than coins changing hands directly, though in-kind coin transfers have been approved for some funds. Either way, the arbitrage loop is what ties a piece of paper on Nasdaq to a coin in a vault.

The fee is the number that decides the winner

Two crypto ETFs holding the same coin give you almost the same price exposure. What separates them over years is the expense ratio, the slice the manager takes every year whether the coin rises or falls. Small gaps compound into real money.

The spread is not trivial. Grayscale's proposed Zcash fund wanted 2.5% a year while spot bitcoin funds charge closer to 0.2%. On a long hold, a fund charging ten times more has to be worth ten times the convenience, and it rarely is. It usually is not. New issuers know this, which is why so many launch with a fee waiver for the first few months to buy market share. Bitwise's Solana fund waived its fee until it gathered its first billion dollars in assets.

Premium, discount, and the price you really pay

A crypto ETF has two prices at any moment. There is the market price you pay per share, and there is the net asset value, the actual worth of the coins each share represents. The two usually sit within a whisker of each other because of the creation and redemption loop above. On quiet, liquid funds the gap is a rounding error. On thinly traded ones, or during a market panic when authorized participants pull back, a share can drift to a premium or a discount, and you can overpay or undersell without noticing. Checking the premium before a large order is worth the thirty seconds.

There is a second, slower gap to watch, called tracking error. Fees, cash drag, and the timing of when the fund buys and sells all pull its return a little away from the coin it follows. Over a day it is invisible. Over several years a poorly run crypto ETF can lag the asset by more than its headline fee suggests, and that lag is the real cost of the wrapper, not the sticker number in the prospectus.

What you hand over for the convenience

The pitch for a crypto ETF is real. You get bitcoin exposure inside a retirement account, under the same tax paperwork as your other stocks, with a regulated custodian handling the keys and a bank like Morgan Stanley now able to issue its own spot bitcoin ETF. For someone who never wanted to run a wallet, that removes the scariest part of crypto.

The cost sits on the other side of the same coin. You do not hold the keys, so the old warning that whoever holds the keys holds the crypto applies to the fund, not to you. A crypto ETF only trades when the stock market is open, so a weekend crash or a 3 a.m. move is something you watch, not something you can act on until Monday. Most funds do not pass on staking rewards, so an ether or Solana ETF can leave yield on the table that a direct holder would earn. And the range keeps widening past the blue chips, with issuers filing for everything down to a PEPE memecoin ETF, where the wrapper is respectable but the asset inside is not.

Frequently asked questions

Is a crypto ETF safer than buying the coin?

It removes some risks and adds others. You cannot lose your keys or get phished, because a custodian holds the asset. You are still fully exposed to the coin's price, which can fall hard, and you are trusting the fund, its custodian, and their controls instead of trusting yourself.

Do I actually own bitcoin if I buy a spot bitcoin ETF?

You own shares of a fund that owns bitcoin. You cannot withdraw the coins to a wallet or spend them. If self-custody and moving the asset around matter to you, a crypto ETF is the wrong tool.

Why do two funds holding the same coin have different prices per share?

Share price is just the coin's value divided by how the fund set its shares, so one fund might price a share near 50 dollars and another near 25 for the same underlying. It says nothing about which is better. Compare fees, trading volume, and the premium to net asset value instead.

Can I hold a crypto ETF in a tax-advantaged account?

In many cases yes, which is a large part of the appeal. A crypto ETF slots into brokerage and retirement accounts that will not let you hold coins directly, though the rules depend on your country and your provider.

The wrapper keeps spreading to new assets and the fees keep grinding lower as issuers fight for flows. Before you buy any crypto ETF, the two things worth reading are the expense ratio and whether the fund gives up staking yield, because those decide what you keep. Watch the next round of Solana and altcoin launches to see how far the fees fall.