AMM Design in 2026: When the Chain Dictates the Curve

Two decentralized exchanges can run the exact same formula and still behave like different species. A constant-product pool on Ethereum and a constant-product pool on Solana both obey x*y=k, the rule that fixed the price of an on-chain trade long before order books came back into fashion. Yet the way each one quotes a price, defends against front-running, rebalances inventory, and even which shapes of curve are economically possible, diverge sharply once you look past the math. The formula is portable. The machine around it is not.

That gap matters more in 2026 than it ever has, because the center of gravity in on-chain trading has moved. Solana has led all chains in spot decentralized-exchange volume for five straight quarters, posting roughly $197 billion in the third quarter, about 36% of cross-chain spot activity. It overtook Ethereum on monthly DEX volume back in May, around $117 billion to Ethereum’s $52 billion. The venues doing that volume were not Ethereum designs ported over; they were built for a different runtime, and it shows.

This piece reads AMM design through the chain it runs on. Same invariant, two machines: a deliberate, expensive, globally ordered settlement layer on one side, and a fast, cheap, parallel one on the other. The result is two distinct schools of market-maker design, and a running argument about whether they will converge or keep pulling apart.

One Formula, Two Runtimes

The core idea is almost embarrassingly simple. Vitalik Buterin formalized the constant-product market maker in a March 2018 ethresear.ch post about front-running resistance, crediting the formulation to Gnosis; Hayden Adams turned it into Uniswap later that year. A pool holds two reserves, x and y, and keeps their product constant. Every trade slides along that hyperbola, and the slope at your point on the curve is the price you pay.

Picture a pool with 100 ETH and 250,000 USDC, so k equals 25,000,000 and the spot price is $2,500. Buy 10 ETH and the pool is left with 90 ETH, which forces the USDC side up to 25,000,000 divided by 90, or about 277,778. You hand over roughly 27,778 USDC for 10 ETH, an average price near $2,778, about 11% worse than spot, and the marginal price after your trade sits around $3,086. Nothing in that calculation knows or cares which blockchain it is on.

What does care is everything wrapped around the swap: how the transaction reaches a validator, how it is ordered against other transactions, how much it costs to update the pool, and how often a market maker can afford to re-quote. Those are properties of the chain, not the formula. Change the runtime and the same invariant produces a different market.

Four Properties That Shape Every On-Chain Market

Four features of a base layer do most of the work in explaining why Ethereum and Solana AMMs look different. The first is the execution model. Ethereum processes a block’s transactions sequentially, one touching the next, so a swap and an arbitrage trade against the same pool line up in a single queue. Solana’s Sealevel runtime executes transactions in parallel whenever they touch non-overlapping accounts, which rewards designs that keep state local and contention low.

The second is the fee market. Ethereum runs a global gas auction with an EIP-1559 base fee, so a memecoin frenzy in one corner raises the price of every unrelated swap. Solana uses local fee markets, pricing congestion per account, so a hot pool does not tax the rest of the chain. The third is block time: Ethereum settles roughly every 12 seconds, Solana in slots of about 400 milliseconds. The fourth is the mempool. Ethereum has a public one, where pending trades sit in the open; Solana has no protocol mempool at all, forwarding transactions straight to the upcoming leader through its Gulf Stream mechanism.

None of these are value judgments. They are constraints, and they push designers toward different answers.

The Ethereum School: Composability and the Singleton

On Ethereum, gas is the tax that shapes design. Every storage write costs money and every extra contract call adds to the bill, so the winning patterns minimize on-chain steps and lean hard on composability, the ability to stack protocols like Lego inside a single transaction. Uniswap’s history is a story of squeezing more market out of fewer operations. V1 arrived in November 2018 with flat 0.30% ETH-paired pools; V2 added arbitrary ERC-20 pairs and an on-chain price oracle in 2020; V3 in May 2021 introduced concentrated liquidity, letting providers bunch capital into a price band for up to 4,000 times the capital efficiency, with the position tracked as an NFT and fee tiers of 0.05%, 0.30% and 1.00%.

Then came the gas-driven leap. Uniswap V4 shipped on 31 January 2025 with three ideas aimed squarely at Ethereum’s cost model: a singleton contract that holds every pool in one place, so multi-hop swaps stop paying for token transfers between separate pool contracts; flash accounting that nets balances once at the end of a transaction instead of after each step; and hooks, external code that runs at set points in a pool’s life. V4 has since overtaken V3, carrying roughly half of Uniswap’s volume. Curve’s StableSwap and Balancer’s weighted pools round out the school, each a different curve tuned to waste as little gas and as little capital as possible.

The through-line is that Ethereum rewards doing more inside one expensive transaction. The singleton, flash accounting, and composable hooks are all ways of amortizing a high fixed cost, and they only look clever because the cost is high in the first place.

The Solana School: Speed, Bins, and Oracles

Solana removes the gas constraint and replaces it with a different one: get in and out fast, because blocks come every 400 milliseconds and compute is cheap. That makes designs viable that would bankrupt a user on Ethereum. The three largest venues each take a different tack. Orca’s Whirlpools is a concentrated-liquidity AMM in the Uniswap V3 mold, letting providers set a price range and earn full fees while the market sits inside it. Raydium pairs a classic constant-product pool with its own concentrated-liquidity product and, historically, posted liquidity into the Serum order book, now the community-run OpenBook, a hybrid that only makes sense where placing and canceling orders costs almost nothing.

Meteora’s DLMM takes the boldest structural swing. Instead of a smooth curve, it slices the price line into discrete bins; liquidity inside a bin trades at a single price with zero slippage, and a volatility-sensitive dynamic fee rises when the market moves fast. The bin model traces back to Trader Joe’s Liquidity Book on Avalanche, and it leans on cheap transactions to let providers spread capital across many bins and shift it quickly. The common thread across the Solana school is tempo: these are machines built to re-quote constantly, something a 12-second, dollar-a-swap chain cannot support.

Cheap compute also changes the capital-efficiency calculus. On Ethereum, a provider thinks hard before rebalancing a concentrated position, because every adjustment costs gas; on Solana, moving liquidity between bins or re-centering a range is close to free, so active management is the default rather than a luxury. That cuts both ways. It lets sophisticated market makers hug the price tightly and earn more fees per dollar deployed, but it also means a passive provider who does nothing is competing against bots that re-center all day. The chain that makes active liquidity cheap also makes lazy liquidity relatively worse off.

Lifinity and the Oracle-Priced Curve

The sharpest example of a chain dictating a curve is Lifinity, a Solana AMM that mostly ignores its own reserves when setting a price. A standard constant-product pool only discovers price when an arbitrageur shows up to correct it, which means liquidity providers are structurally late and pay for that lag. Lifinity instead reads the market price from Pyth, a first-party oracle that publishes a fresh value in every Solana slot, roughly every 400 milliseconds, and quotes around it. Its creators call it a proactive market maker: it does not wait to be arbitraged, it moves with the oracle.

This design is only practical because of the chain underneath it. Re-pricing a pool against an external feed several times a second would be absurd on Ethereum, where each update burns gas and the next block is 12 seconds away. On Solana the update is nearly free and the oracle refreshes at slot speed, so a market maker can track an off-chain price in near real time, cut the arbitrage leakage that drains passive pools, and claim lower impermanent loss as a result. It is the cleanest illustration of the thesis: the same job, market making, produces a structurally different machine once the runtime changes.

Oracle pricing is not a free lunch, of course. It trades arbitrage risk for oracle risk, so the market maker is only as safe as the feed it trusts, a trade-off Ethereum pools historically avoided by letting arbitrage do the pricing for them. But that is precisely the point: Solana made a different trade-off economically sensible.

Why MEV Looks Different on Each Chain

Maximal extractable value is where the mempool difference bites hardest. On Ethereum, pending swaps sit in a public mempool, so searchers can see a trade coming and wrap it in a sandwich: buy in front, let your purchase push the price up, then sell into you after. The practice is industrialized through block-building auctions, and it reaches everyone; a well-known bot front-ran one of Vitalik Buterin’s own swaps in 2026. Builders and relays now sell blockspace to the highest bidder, which at least drags the value into the open.

Solana has no public mempool to snipe from, but that did not make it immune. Searchers found their opening inside Jito’s Block Engine, which had added an off-chain mempool to surface MEV. After a surge of sandwich attacks on ordinary users, Jito Labs switched that mempool off in March 2024, ending what it described as a six-week cat-and-mouse fight with front-runners. The deeper lever is who gets to order transactions. Solana co-founder Anatoly Yakovenko has argued that the chain’s Alpenglow consensus upgrade changes the economics of MEV, making delay-based reordering more expensive for validators to pull off. On both chains the underlying leak is the same, arbitrage against a slightly stale AMM price, but the plumbing that exposes it is chain-specific, and so are the defenses.

The defenses rhyme across chains even when the attacks do not. On Ethereum, traders increasingly route through private order flow and intent auctions that never touch the public mempool, so there is nothing to sandwich; on Solana, scrapping Jito’s mempool pushed the same logic down to the base layer. Neither fix is total. Private flow concentrates power in a handful of builders and solvers, and a validator that controls block ordering can still extract value whether or not a mempool exists. MEV is less a bug of one chain than a property of putting a market on any public ledger.

The Order Book Comes Back, On Solana

AMMs exist because order books were impractical on early Ethereum: posting and canceling limit orders on a chain that charges dollars per write and settles every 12 seconds is a non-starter. Flip those constraints and the old design becomes attractive again. Solana’s cheap, fast blocks have revived the central limit order book on-chain, from OpenBook, the community fork of Serum, to Phoenix, a fully on-chain order book that many professional market makers prefer for its tight spreads. Raydium’s hybrid routing between its own pools and an order book is a product of the same economics.

A word of caution on a recurring mix-up: Hyperliquid, often cited in the same breath as these venues, runs a central limit order book for perpetual futures on its own layer-one chain. It is a formidable exchange, but it is not an AMM, and comparing it to Uniswap or Orca confuses two different mechanisms. The honest read of 2026 is that cheap blockspace did not kill the AMM so much as end its monopoly on on-chain liquidity, letting curves and order books compete on their merits again.

What is emerging is not a winner but a division of labor. Order books tend to win for large, liquid pairs, where professional market makers can quote tight two-sided prices; AMMs still win for the long tail, where no human will sit on a book for an obscure token but a passive curve will always quote something. Aggregators paper over the seam, pulling from both so a trader never has to know which one filled the order. On Solana that coexistence is native; on Ethereum it mostly lives on cheaper rollups, for the same reason it could not live on mainnet a few years ago.

LVR and Impermanent Loss Do Not Care Which Chain You Use

For all the architectural divergence, the economics of being a liquidity provider are stubbornly chain-agnostic. Impermanent loss, the gap between holding two assets and pooling them while their relative price moves, follows the same curve on Solana as on Ethereum. A 2x move costs about 5.7% versus holding; a 5x move costs about a quarter of the position. Fees have to make up that difference, or the provider would have been better off doing nothing at all.

The sharper framing is loss-versus-rebalancing, or LVR, the value a passive pool bleeds to arbitrageurs who always trade against a slightly stale price. Research from a16z crypto, led by Tim Roughgarden and co-authors, modeled it as proportional to volatility squared over eight; a plain ETH-USDC pool at 5% daily volatility loses on the order of 11% a year to this effect before fees. Their conclusion is the one every design in this article is really chasing: “An AMM can be successful only if it has happy LPs, which means that fee revenues need to scale with LVR.” That is as true for a Meteora bin on Solana as for a Uniswap tick on Ethereum, which is exactly why oracle pricing, dynamic fees and order books keep resurfacing on both sides. The impermanent-loss figures above are drawn from the Binance Academy explainer.

pump.fun and the Memecoin Factory

No design embodies cheap-and-fast more than the Solana memecoin pipeline. On pump.fun, a new token launches onto a bonding curve, a one-sided pricing function where the cost rises as more is bought, with no liquidity provider required. When a token’s market value crosses roughly $69,000 it graduates: the accumulated SOL seeds a standard constant-product pool, these days on pump.fun’s own PumpSwap AMM rather than Raydium, and ordinary two-sided trading begins. The whole flow fires in seconds for a few cents.

That machine has swallowed the category. pump.fun has at times accounted for roughly 80% of Solana memecoin launches, and only a small fraction of tokens ever graduate; the rest fade or were never meant to last. It is also a magnet for scams and drainers, which is why the same chain that makes launching a token trivial makes it trivial to lose a wallet to a malicious approval. Anyone moving fast through these venues should treat every unfamiliar token as hostile until proven otherwise, the discipline the crypto phishing kill chain rewards. A bonding-curve-to-AMM conveyor belt is simply not a design Ethereum’s fee market would tolerate at this volume.

Ethereum’s Answer: Rollups, Hooks, and the Aero Merger

Ethereum’s response to Solana’s speed has been to move the cheap-and-fast activity onto rollups and to make its AMMs programmable. Layer-two networks like Base, Arbitrum and Optimism push swap costs down toward Solana territory, at the price of fragmenting liquidity across dozens of chains. Uniswap V4 hooks turn that fragmentation into a feature of sorts: a pool can attach custom logic for dynamic fees, time-weighted average market making, limit orders or oracle protection, so one codebase can host hundreds of specialized markets. The flip side is new attack surface, because arbitrary code running inside a pool is arbitrary risk, which is why serious teams lean on auditors like Trail of Bits before shipping a hook that touches real money.

The consolidation instinct is strongest in the incentive-driven corner of the EVM world. On 21 October 2026, Aerodrome and Velodrome, the dominant liquidity layers on Base and Optimism, merged into a single protocol called Aero spanning seven chains, including Base, Ethereum, Optimism, Arbitrum and Robinhood Chain, with AERO holders migrating one-for-one and VELO at a fixed ratio. The pitch was scale: one vote-escrow system directing emissions across a far larger addressable market instead of two sister protocols competing for the same flow. Where Solana concentrates liquidity by default on one fast chain, Ethereum’s ecosystem has to keep re-stitching a liquidity fabric that its own scaling strategy pulled apart.

Routing: Jupiter Versus the Aggregators

When liquidity scatters across many pools and designs, the aggregator becomes the real front door. On Solana, Jupiter has become the default router, splitting a single order across Orca, Raydium, Meteora, Lifinity and Phoenix to find the best net price, and it handles the large majority of Solana aggregation volume. On Ethereum and its rollups the same job falls to 1inch, Matcha, CoW Swap and UniswapX, each walking a liquidity graph to minimize price impact and, increasingly, routing through intent-based auctions where professional solvers compete to fill an order.

For an ordinary user the distinction between a pool and the router in front of it has mostly dissolved, part of a broader blurring of the line between wallet and exchange. You tap swap, a router you never see picks the venue, and the underlying AMM design matters only through the price you get and the slippage you eat. That abstraction is convenient, and it is also why so few traders could tell you whether their last swap touched a constant-product curve, a bin, an oracle-priced quote or an order book.

Aggregation has also quietly become a safety layer. A good router does not just hunt for the best headline price; it can split an order to limit price impact, set tight slippage, and, on Ethereum, send the trade through a private channel so no bot can front-run it. The practical upshot is that the most important choice a retail trader makes is often not which AMM to use but which router to trust, since the router picks the venue, the path and the protection. That is a strange place for the industry to land after a decade of arguing about curve math, but it is where 2026 actually sits.

The Flippening in DEX Volume

The scoreboard has tilted. Decentralized exchanges as a whole took a record share of spot crypto trading in 2026, with the DEX-to-CEX spot ratio hitting about 24% in July, the highest in The Block’s series, before easing back toward the high teens through the autumn. Within that total, Solana has done the heavy lifting on raw turnover, while Uniswap remains the single largest venue by volume, clearing more than the next several DEXs combined across Ethereum and its rollups. Two different winners, measured two different ways.

The backdrop is a jittery market. As this piece went out, a broad risk-off move had dragged the majors lower, with Bitcoin in the low $80,000s, Ether around $2,380, Solana near $102 and Uniswap’s UNI token close to $6.96, all down on the day. Volume leadership and token price have clearly decoupled: Solana can out-trade every chain while its token sells off with the rest of the market, and UNI can lag even after Uniswap switched on a protocol fee. For where the cycle might head from here, our year-end outlook lays out the bull and bear cases. Spot figures here are drawn from CoinGecko.

Regulation: The SEC, DEXs, and Why the Chain Does Not Matter

Regulators do not care about Sealevel versus the EVM. In the United States, the Securities and Exchange Commission issued Uniswap Labs a Wells notice in 2024, then closed the investigation in February 2025 without an enforcement action, a marked softening under the current leadership. The unresolved questions, whether an AMM pool is an exchange, whether a governance token is a security, and where the Commodity Futures Trading Commission’s remit over derivatives begins, attach to the activity, not to the chain beneath it.

The practical pressure tends to land on the points a protocol cannot fully decentralize: the web front end, the token treasury, the governance process, the fiat on-ramps. That is the thesis of following the compliance chokepoints, and it applies identically to an Orca pool on Solana and a Uniswap pool on Ethereum. A clever curve does not change who holds the keys to the interface, and it is the interface that a regulator can actually reach.

Convergence or Divergence: The 2026 Outlook

The two schools are already borrowing from each other. Ethereum’s V4 hooks can now host oracle-priced quoting and dynamic fees that look a lot like Lifinity or Meteora; Solana keeps importing concentrated liquidity and order books that started on Ethereum and Avalanche. Faster Ethereum rollups and cheaper blockspace narrow the cost gap that justified much of the divergence in the first place. If settlement everywhere trends toward cheap and fast, the design space may converge on a common toolkit, with the chain becoming a deployment detail rather than a design constraint.

The counter-case is that chains have personalities, and the AMMs that fit them will keep diverging at the edges. Yakovenko keeps betting that raw speed reshapes market microstructure in ways slower chains cannot copy; Uniswap founder Hayden Adams, pushing back on claims that AMMs are a dead end, insists that AMMs are only just getting started, pointing to cheaper capital and composability as durable edges. Both can be right at once. The formula is a commodity now; the interesting work is in the machine you wrap around it, and in 2026 the chain you pick still writes a large part of that machine for you.

Frequently Asked Questions

Why is Solana’s DEX volume higher than Ethereum’s in 2026?

Solana’s sub-second blocks and fractions-of-a-cent fees make high-frequency trading, memecoin launches and constant re-quoting cheap, which carried it past Ethereum on monthly DEX volume around May 2026 and to roughly $197 billion in the third quarter, leading all chains for a fifth straight quarter. Ethereum still hosts the single largest venue, Uniswap, by volume.

Does the same AMM formula work differently on Solana and Ethereum?

The math is identical: x*y=k produces the same price curve on any chain. What differs is the runtime around it, namely parallel execution, local fee markets, 400-millisecond blocks and no public mempool on Solana versus sequential execution, a global gas auction, 12-second blocks and a public mempool on Ethereum. Those constraints push designers toward different fee models, oracle use and order-book hybrids.

What is Lifinity and why can it not run on Ethereum?

Lifinity is a Solana AMM that prices trades from the Pyth oracle, which refreshes every slot, about 400 milliseconds, rather than waiting for arbitrageurs to correct its reserves. Re-pricing a pool several times a second against an external feed would be prohibitively expensive on Ethereum, where each update costs gas and blocks are 12 seconds apart, so the design is effectively chain-specific.

Is Hyperliquid an AMM?

No. Hyperliquid runs a central limit order book for perpetual futures on its own layer-one chain, not a constant-product or concentrated-liquidity AMM. It is often grouped with DEXs in volume tables, but its mechanism is an order book, so comparing it directly to Uniswap or Orca mixes two different market structures.

Which chain is safer for providing liquidity?

Neither chain removes the core risks. Impermanent loss and loss-versus-rebalancing hit liquidity providers the same way on both, and smart-contract or hook bugs can exist anywhere. Solana adds dependence on oracle feeds and validator behavior, while Ethereum adds gas costs and public-mempool MEV. The safer choice depends on the specific pool, its fees and its audits, not the chain alone.

By Yuki Tanaka, senior DeFi correspondent at HOGE Wire.