I was on a panel at a private summit in Lisbon last month, the kind where the Chatham House rule is taken seriously and the conversations get sharper than they would on a public stage. Two DeFi protocol founders, one CEX head of product, and me. The host asked the standard panel-opener: "Is the AMM dead?" The CEX product head answered first. He said, "That question is two years late. The interesting one is what comes after intents." The room agreed. The panel did not turn out to be about AMMs.
I want to write down what I think the answer to what comes after intents is, because I've now seen close to a dozen pitches in 2026 that are still answering the wrong question, and I don't want our portfolio to be on the wrong side of it.
The Part That's Done
The orderflow battle between AMMs and intent-based routing isn't over because intents are objectively better. It's over because the customer-side data made it inevitable. By Q2 2026, a majority of all retail DEX volume on Ethereum and major L2s is now routed through intent-based protocols, UniswapX, CoW Swap, 1inch Fusion, Across for cross-chain, plus the next generation of solver-driven venues coming behind them. The transition was quieter than the original AMM-versus-orderbook debate because it didn't require users to change their behaviour. They kept clicking the same swap buttons on the same aggregators. The orderflow underneath those buttons quietly routed to solvers.
The mechanical reason intents won is straightforward. From the user's perspective, intents deliver better execution, typically 5–15 basis points better on retail-sized trades, sometimes more on size, because the solver network competes to fill the order rather than the user paying the static AMM curve plus slippage. From the protocol's perspective, the intent architecture composes more cleanly with the rest of the DeFi stack, cross-chain, multi-step, with MEV protection baked into the design rather than bolted on. From the solver's perspective, the architecture creates a real new revenue line for sophisticated market-makers who were previously losing economic value to vanilla MEV extraction.
That's three rational actors agreeing on a transition. It was always going to be quick once the infrastructure caught up. It was.
What "Won" Actually Means In 2026
The part most DeFi pitches in 2026 are still missing is that "intents won" doesn't mean "intent-based venues won and AMMs lost." It means the entire orderflow stack has been re-decomposed. There are now three distinct layers — and the value capture, the competition, and the venture opportunities live at different layers than they did in the AMM era.
Layer one: the intent surface. Where users express what they want, the wallet, the aggregator, the embedded swap in a fintech app. This layer has become much more diffuse. Wallets and consumer fintech UIs are now the primary intent-creation surface, and they're not all crypto-native. The fintech-with-embedded-swap pattern that I wrote about as a thesis in early 2025 on Cointelegraph is starting to compound. The intent surface is increasingly something users encounter inside their existing product, not on a DeFi-native site.
Layer two: the solver network. Where the intent is matched against liquidity and executed. This is where the most concentrated competition is right now. The largest solvers — Wintermute, Flow Traders, GSR, Symbolic Capital's market-making arm, and the new generation of pure-DeFi solvers, are now real, profitable businesses. The barrier to entry at this layer is high: you need capital, infrastructure, and the ability to compete on basis points at machine speed. Most DeFi venture pitches I see in 2026 want to play here. Almost none of them have the operational profile to actually compete.
Layer three: the liquidity venues. Where the underlying inventory lives — AMM pools, orderbook venues, RFQ networks, increasingly tokenized inventory from CEX flow and TradFi market-makers. AMMs are still here. They didn't die. They became one liquidity source among several, addressable by solvers rather than directly by users. That's a different business than the one Uniswap, Curve, and Balancer were running in 2021–2023, but it's still a real one.
Where The Opportunity Actually Sits
If I had to point at the layer where I think the next billion-dollar DeFi business gets built in 2026, it's not the solver layer (the incumbents are too far ahead) and it's not the AMM layer (the economics are mature and the moat is already paid for). It's the intent surface and the layer just above it.
Specifically: products that aggregate intents across multiple verticals in ways that weren't possible before. Cross-chain order routing that composes a single user intent across Ethereum, Solana, Bitcoin L2s, and the major app-chains. AI-driven intent composition (this is one place where the dAI thesis I wrote about in Bridge Notes #2 actually lands cleanly — a wallet copilot that turns a high-level user goal into a multi-step intent across protocols is exactly the kind of consumer dAI surface that has real defensibility). Embedded intent infrastructure for fintechs and on-chain applications that don't want to build the stack themselves.
This is also where the cross-chain bridging and settlement work I'm closer to at Kima Network, disclosed conflict, as I noted in Bridge Notes #3, composes naturally with the intent thesis. Bridgeless cross-chain settlement makes the intent surface multi-chain by default rather than as a bolt-on. The intent-then-settle pattern is the cleaner architecture for the next wave of products, and the teams building toward it are quieter than the ones competing at the solver layer but, I think, better positioned.
What I'd Bet Against
I'd be reluctant in 2026 to underwrite new pure-AMM protocols, pure-perps DEXes that don't have an intent-native architecture, or new aggregators that compete directly on the surface UniswapX/1inch already occupy. The economics aren't going to work for newcomers in those categories.
I'd also be cautious on new solver-network plays unless the founding team has either institutional market-making credibility (the way Wintermute or Flow Traders does) or a genuinely novel architectural angle. Capital is necessary but not sufficient — the solver business is increasingly a market-making business in everything but name, and the founders who win it are going to look more like prop-trading founders than like DeFi founders.
What I would bet on, and where we at PRIM3 are spending more time in 2026, is the composability layer between intents and adjacent on-chain primitives. Intent-native lending. Intent-native structured products (this is where the SHIFT thesis on bidirectional leveraged tokenized equities composes, conflict disclosed). Intent-native cross-chain settlement. Intent-aware wallet experiences for non-crypto-native users.
The Forbes-Quotable Line
If I had to compress this to one sentence: intents already won; the founders still pitching "intents versus AMMs" are answering 2024's question, while the founders building on top of the assumption that intents won are building 2027's businesses.
The next Bridge Notes will turn to a piece I've been chewing on for two months — why verifiable inference and zkML are the trust layer decentralized AI has been missing, and which protocols I think are positioned to provide it. If you're building at the intent-surface layer and want to push back on the framing, LinkedIn or Telegram.