Lead-in: If Agents truly become the next billion users of blockchain, the more pressing question may not be "how much transaction volume will they generate," but rather: who will make money if this world actually arrives?

Historically, both the "Fat Protocol" and "Fat Application" theories assumed on-chain users were humans. Humans care about usability, brand trustworthiness, and convenient pathways—thus application layers could capture value by controlling user entry points and transaction flows. But Agents are different. They directly invoke APIs, have no brand loyalty, and can switch cost-effectively between protocols, aggregators, and trading venues.

This implies that Agents may rewrite Web3’s value distribution logic. Application layers could shift toward "headless" architectures, opening up wallet, aggregator, and on/off-ramp capabilities as APIs for Agents; protocol layers might regain opportunities by being bypassed by Agents; but in a more radical scenario, Agents could push the entire on-chain stack into price competition, compressing profit margins of applications, aggregators, and infrastructure down to near marginal cost.

What truly matters is that Agents aren’t just increasing the frequency of existing on-chain transactions—they may enable entirely new activities that were previously unfeasible: continuous portfolio rebalancing, machine-to-machine payments, and novel markets whose significance only emerges under automated, high-speed execution.

Thus, the central question of the Agent era isn’t simply whether value flows to protocols or applications, but who can ensure that Agents, despite infinite substitution options, still choose to return here. The answer may no longer be UX or branding—but liquidity, latency, settlement finality, or some new business model yet unnamed.

Below is the original text:

Many people envision Agents becoming the next billion users of blockchain. But few ask the second-order question: if this world actually arrives, who will make money?

All past theories about value capture in crypto implicitly assumed human users. The "Fat Protocol" theory posits that protocol layers best extract value from users. Meanwhile, our "Fat Application" theory—outlined in *How to Capture Value* and *The Great Repricing*—argues that application layers perform better.

But Agents redefine who the "user" is. Hence, established value-capture theories no longer hold.

In 2016, @jmonegro wrote *Fat Protocols*. For nearly a decade since, this essay has become the dominant value capture framework in crypto.

Its core idea: in the internet era, value flowed primarily to application layers—think @Google, @facebook—while foundational protocols like TCP/IP and HTTP captured little. But crypto would reverse this. On-chain data is open and shared, so applications become commoditized; meanwhile, protocol tokens essential for network usage capture speculative value as adoption grows. Every successful application drives demand for its underlying token. Ultimately, protocol layers compound faster than any application above them.

For a long time, this view seemed correct. Bitcoin and Ethereum now have higher market caps than any company built upon them. This model worked because protocol layers were scarce, expensive, and hard to replace. In 2017, Bitcoin and Ethereum indeed had scarcity—there were no dozens of general-purpose L1s competing for the same workload. Blockspace was tight enough that holding underlying assets felt like owning partial equity in every app needing that network.

Today, however, every layer of the infrastructure stack has credible alternatives: multiple high-throughput L1s, dozens of L2s, and modular settlement and data availability layers competing fiercely on price. Blockspace has shifted from scarce to abundant. As cross-chain bridges and aggregators render underlying chains nearly invisible to users, switching costs plummet rapidly. Infrastructure becomes substitutable—and substitutable things compete solely on price. Thus, pricing power of protocol layers vanishes alongside scarcity.

By 2026, entities capturing significant economic value will no longer be protocols, but applications—such as @phantom, @coinbase, @Polymarket, @Pumpfun.

In my view, the reason lies in the fact that the most valuable asset in crypto is user relationships. If you control the user interface and transaction flow, you control distribution. And as soon as users interact with any on-chain product, you can monetize almost every touchpoint: swaps, lending, staking, minting, on/off-ramps, etc. This likely explains why investors are so obsessed with neobanks.

Applications will drive infrastructure into pure price competition, compressing infrastructure margins down to near marginal cost. I documented this strategy in *How to Capture Value*. A similar dynamic is already unfolding in the stablecoin space, which I’ve discussed elsewhere.

Prices are reflecting this theory. Spencer and I call this shift the “Great Repricing”: in this cycle, value flows to the layer that owns user relationships.

The "Fat Application" theory assumes human users—who value UX, branding, and convenience. But Agents don’t care about these. They directly invoke APIs, have no brand loyalty, and can switch venues at zero cost.

When users become software, owning user relationships becomes less defensible. The entire front-end moat that the "Fat Application" theory relied upon depreciates accordingly.

So, who captures value in the Agent era?

Applications Headless

One possible future: winners in the application layer will remain dominant—but they’ll abandon UIs.

Wallets and aggregators have already built the hardest parts: deep integration across numerous protocols, routing logic, identity systems, and on/off-ramp infrastructure. The natural next step is to expose this capability as APIs for Agents, enabling Agents to route trades just as human users do today via @phantom or @JupiterExchange.

In this world, the "Fat Application" theory still holds—but without the frontend. Companies that won in the human-user era will re-platform into headless infrastructure. We’ve already seen traditional SaaS giants like Salesforce pivot in this direction.

Protocols Rise Again

Another possibility: Agents will completely bypass intermediaries.

If integration is simple enough—clear API documentation, standardized RPCs, predictable execution semantics—Agents have little incentive to pay aggregators to do what they can do themselves.

Aggregators’ advantages in the human-user era came from UX and complex routing. But Agents need neither UX nor complex routing—routing is an engineering problem solvable by code, and Agents are increasingly adept at handling such tasks.

If this future materializes, the "Fat Protocol" theory will receive a second life.

Pricing Power Collapses Across the Stack

A third possibility: Agents will exert commoditization pressure across the entire stack.

They are rational. Each time, they pick the cheapest venue—no loyalty, no friction. Applications lose the UX premium once charged to human users. Aggregators and infrastructure lose pricing power too, as there’s no human inertia left to shield them from price competition.

In this scenario, no layer can capture much value. The entire supply chain gets compressed down to near marginal cost, and economic surplus flows to whoever owns the Agents—or to the end users they represent. Crypto becomes a utility, and utilities are rarely profitable.

Agents Enable Previously Impossible Activities

The simple version of this argument: Agents do what humans already do, but at vastly higher throughput. Even with margin compression, if transaction volume explodes, the overall pie grows.

But I believe there’s a more intriguing version: Agents make certain activities previously unfeasible now viable. Examples include continuous portfolio rebalancing at sub-1-cent execution cost; machine-to-machine commercial transactions between Agents; and markets that only make sense when pricing and execution speed outpace human ability to react.

These activities won’t appear in our current frameworks for observing on-chain activity, because we assume a human participant is always present.

If this is truly Agent-driven transformation, then the question shifts from how to divide the existing pie to how much new economic activity will emerge on-chain—and which layers are best positioned to serve it.

Each cycle, we try to predict where value will flow, often assuming: the business models we know will naturally extend into the future. But this assumption routinely misses emerging models yet to appear.

When the internet was first built, no one foresaw the rise of attention economies. Today’s seemingly obvious business model—segmenting user attention into discrete units, auctioning them to advertisers, and extracting a substantial share from global ad spend—is now commonplace. Yet, at the time, it was utterly foreign. It only seems inevitable in hindsight.

AI appears to be one of the largest technological disruptions in decades. In a world dominated by Agents, part of value capture may flow to a business model no one has seriously discussed yet. The eventual value capturers may not even be the players currently dominating the market.

The most likely outcome is not one paradigm fully replacing another. Humans and Agents will coexist as users of crypto for a long time, and their respective value capture maps will differ.

As long as humans continue to interact directly with blockchains, the "Fat Application" theory remains valid: consumers willing to pay for UX, branding, and convenience will continue to pay premiums to applications that own user relationships. Meanwhile, in the layer where Agents trade, a different set of rules will govern—exactly which depends on how the scenarios above ultimately unfold.

In my view, the most critical question builders should repeatedly ask about the Agent side is: what will make an Agent return to you, rather than immediately route to the next cheaper alternative?

The answer may not be UX. It could be liquidity, latency, settlement assurance, or something else entirely.

At @bcap, we’re dedicating significant time to thinking through this question—whether in investment committee meetings or with engineering teams. We don’t yet have a definitive answer. If you’re building products around Agents and have your own views on value capture in the Agent era, we’d love to talk.

[Original Link]

Original: BlockBeats