Who Funds the Agents?

marsbitОпубліковано о 2026-06-04Востаннє оновлено о 2026-06-04

Анотація

**Summary: Who Funds AI Agents?** OpenAI recently shut down a feature allowing AI agents to shop for users, highlighting the challenge of creating a secure and regulated environment for agent-driven transactions. While payment infrastructure exists, a crucial governance layer—defining spending limits, fraud detection, tax handling, and return policies—is largely missing. The potential is enormous: AI agents already processed $73M across 176M transactions last year, with McKinsey forecasting this could grow to $3-5T in global consumer commerce by 2030. The core competition isn't just about processing payments, which can be very cheap (especially with crypto-based settlement), but about controlling the rules that govern agent spending. Key players like Stripe and Coinbase are racing to dominate this governance layer. Stripe's acquisition of wallet provider Privy allows it to set spending policies, identity checks, and human-in-the-loop approvals directly at the wallet level. Similarly, Coinbase's stack, including its x402 protocol and AgentKit, embeds governance rules. This vertical integration across settlement, wallet, and governance layers is becoming the dominant strategy. Control over the governance layer is where significant future value lies. If agents handle trillions in transactions, even a small fee for managing compliance, fraud prevention, and policy enforcement could generate billions in annual revenue. The companies that successfully integrate across the payme...

Written by: Prathik Desai

Compiled by: Block unicorn

In March of this year, OpenAI shut down a feature that allowed AI agents to shop on behalf of users. Since its launch, in just five months, fewer than 30 Shopify merchants had used this feature. The payment infrastructure itself was not the problem; the issue lay in the lack of rules to ensure a smooth shopping experience. Which products could agents purchase? Who was responsible for collecting sales tax? How to identify fraudulent behavior? Who would handle returns—all these questions were not properly addressed.

Providing wallets for agents or building payment infrastructure is relatively easy. However, enabling individuals or businesses to use agents for consumption in a trustworthy and regulated manner is not. Only programmability and rules can ensure a trusted environment. This governance-level gap presents an opportunity for the agent economy.

Last year, AI agents processed 176 million transactions, totaling $73 million. While this number might seem insignificant now, McKinsey predicts that by 2030, AI agents will facilitate $3 trillion to $5 trillion in global consumer commercial transactions.

Companies building this economic system are racing to control the governance layer, which includes spending controls, identity verification, and policy enforcement—these determine which agents have the authority to manage budgets.

Today, we will analyze who is building the banking layer for robots and what benefits those who dominate this layer can gain.

Why Own a Multi-Layered Architecture?

The economics of handling agent payments are very low. Over the past 12 months, the average AI agent payment was just 31 cents.

Consider how much profit is left for those processing transactions through multiple backend layers for a 31-cent payment. Stripe's standard pricing is 2.9% plus a 30-cent fixed fee, leaving the merchant with less than one-tenth of a cent. Visa's interchange fees would eat away another third. In contrast, a Layer-2 stablecoin payment system processing the same transaction charges only 0.0001 USD.

These economic factors provide a rationale for the use of cryptocurrency at the settlement layer.

The payment infrastructure at the settlement layer is largely in place. Coinbase's x402 protocol handled the vast majority of last year's 176 million transactions, and currently, about 3,900 merchants accept agent payments. Stripe and Tempo jointly developed a competing protocol—the Machine Pay Protocol (MPP), launched in March, which integrates over 100 services. Google, Visa, and Mastercard also launched agent payment products around the same time. This means five competing payment architectures emerged within just 12 months.

But the problem with agent payments is that no one gets rich processing 31-cent payments. Therefore, value concentrates in the funds in circulation and the enforcement of rules governing how agents make payments.

Last week, we explained how companies capture value by owning the wallet layer that holds AI agents' stablecoin balances. But floating balances are just one of many valuable layers worth capturing. Another value layer is the rules governing how these floating balances are used.

These rules include spending controls, agent identification, policy enforcement, audit trails, and allocating liability for failed transactions. This layer is wide open.

In April this year, American Express launched the "Agent Purchase Protection" plan, an insurance product designed to cover losses from erroneous purchases made by AI agents. This effectively acknowledges the current state of the AI agent governance layer. Solving the missing governance issue holds immense value in an industry projected to reach $3-5 trillion within five years.

This is why incumbent players are now racing to seize control of the governance layer.

But at which level should this layer be built? It could be a bank, a developer API, or even a wallet.

Wallet as the Governance Layer

Every purchase an agent makes must go through a wallet. Therefore, the wallet is the optimal point to implement spending limits, identity verification, and human approval. Control the wallet, and you control the governance. Payment infrastructure company Stripe realized this early on.

In June 2025, Stripe acquired Privy, a company building embedded wallets for consumer crypto applications. Through this acquisition, Stripe gained 75 million wallets distributed across over 1,000 development teams. These wallets now sit at a critical choke point for fund flow, where all policies, spending limits, and human approvals must be executed before funds move.

Stripe has also built an entire agent payment technology stack. It acquired Bridge to handle stablecoin coordination and fiat conversion. Furthermore, it co-incubated Tempo with Paradigm, a Layer-1 blockchain focused on payments. Stripe and Tempo jointly authored the Machine Pay Protocol (MPP), an open standard for how agents request, authorize, and settle payments.

Stripe's Agent-Ready Financial Solutions now allow software to check balances, pay bills, store funds, create virtual cards, and transfer money. Agents can handle routine payments themselves, but any action outside their policy triggers a human review. Fund balances are powered by non-custodial Privy wallets available in over 150 markets.

Even Amazon, when forced to allow its developers to give AI agents spending power, chose two wallet companies—Privy and Coinbase. Not an experienced financial institution like a bank or card network, but a five-year-old wallet provider.

This is because wallets act as the ideal checkpoint, allowing an appropriate level of human intervention to ensure necessary checks and balances.

Keyrock, in its report "Who Pays for the Agent," states that the agent commerce market will "settle on an equilibrium where agents have considerable autonomy but operate within boundaries enforced by cryptography that humans can audit and revoke."

This is the position Privy holds within Stripe's stack. The wallet is responsible for setting the boundaries within which agents must operate.

Here's how governance policies work on this stack.

Privy offers two smart wallet models. In the first, the agent has full control of the wallet and executes transactions within policy constraints without human approval—ideal for fully autonomous agents like trading bots and portfolio managers. In the second, the user retains wallet ownership but grants the agent limited permissions to operate as a signer. The user can revoke access at any time.

Stripe's MPP follows similar governance logic.

The MPP introduces a feature called "Sessions" for high-frequency agent tasks. In Session mode, the agent pre-authorizes a spending budget and then makes continuous payments within that limit without needing a separate request for each on-chain transaction. The MPP has implemented sub-cent billing for LLM inference and per-query billing for data APIs.

This is a level of governance granularity that card networks cannot support.

Vertical Stack Expansion

Although Coinbase's x402 currently leads in AI agent payments, Privy's advantage is less about cryptocurrency itself and more about its distribution moat built through Stripe.

Coinbase has 3,900 merchants accepting agent payments. For every merchant accepting agent payments via Coinbase, Stripe has roughly 1,000. In February this year, Privy stated that if all Stripe merchants opted to accept machine payments, agent commerce could scale today via Privy wallets. Stripe merchants do not need to build custom crypto infrastructure.

Competition between Stripe and Coinbase is intensifying, and other traditional giants have joined the race for vertical expansion, aiming to grow across the entire technology stack.

Keyrock mapped 179 projects across the six layers of the agent payment stack (Settlement, Wallet, Routing, Protocol, Governance, and Application).

Coinbase and Stripe each cover five of these six layers. Circle covers four. Despite their size, Google covers only two layers, and Visa just one.

Over the past twelve months, incumbent payment giants spent over $80 billion filling gaps in their own stacks. Capital One acquired AI-native software platform Brex for $5.15 billion. Mastercard acquired BVNK for $1.8 billion. Among these, the wallet layer and the AI software layer saw the most active M&A activity. Stripe acquired Privy, Fireblocks acquired Dynamic, Arbitrum acquired ZeroDev. In each case, a payment infrastructure provider acquired an independent wallet provider.

Collectively, these deals indicate the market has identified a scarce resource layer. Settlement fees become cheap and interchangeable, but programming permissions, budgets, and liability are where the value lies.

Vertical integration across multiple layers also has compounding effects.

Whoever owns this checkpoint sets the spending rules, intercepts funds before they move, decides which merchants, agents, and apps get trusted access, and charges fees to do all this. We see this in the Privy-Stripe distribution moat.

Even Coinbase's stance reflects this dynamic. Every x402 payment generates USDC demand on its Layer-2, Base, producing float yield. This yield funds more agent tools via AgentKit, which features built-in session caps, single-transaction limits, and allowlists restricting transfers to audited contracts. More agents on AgentKit mean more x402 payments. Each layer influences the others.

Incumbent investment activity is significantly more active.

Coinbase Ventures also invested in Catena Labs, Skyfire, and Payman, three of the most prominent independent governance startups currently. Circle's co-founder Sean Neville founded Catena, and Circle invested in Skyfire. a16z led funding rounds for both. Visa backed Payman and partnered with Skyfire.

The same players building payment settlement infrastructure are funding the governance layer. The idea is that incumbents maximize their returns if governance functions remain part of the existing infrastructure, as Privy built in its two models. If governance becomes a separate layer, they profit via their portfolio.

What Does Controlling the Governance Layer Mean?

Payment processing was never the most valuable position because financial systems eventually commoditize. Once that happens, profit margins shift to where the decision is made about whether a transaction is allowed and under what conditions.

Historically, many industries have undergone the same commodification process.

Consider what happened after the internet commodified cable TV. All Internet Service Providers (ISPs) became nearly identical, almost interchangeable. Therefore, telecom companies had to expand vertically to stay competitive.

India's two largest telecom operators, Jio and Airtel, began bundling hundreds of TV channels, six OTT platform subscriptions, unlimited voice calls, set-top boxes, and free routers into a single broadband package. Similarly, AT&T spent $85 billion to acquire Time Warner, becoming a media-telecom giant. The aim was to combine Time Warner's premium content—like HBO, Warner Bros., and CNN—with AT&T's vast distribution network to compete with streaming platforms like Netflix and Amazon.

When broadband connectivity (the underlying infrastructure) became the cheapest part of the package, value shifted to the combination of content, relationships, and offers that best attracted customers.

We see this happening in crypto as well.

Settlement was supposed to happen at the protocol layer. Think of Ethereum as a shared ledger where everyone settles. After Coinbase launched Base as a faster, less congested Layer-2 chain, it began collecting gas fees from every transaction settled on its own chain. Today, Coinbase earns approximately $60 million annually in sequencer revenue by processing transactions on Base.

Players building agent payment systems have learned this lesson.

In "Active Buoyancy," we explained how an economy can be built by controlling the stablecoin balances agents hold between transactions. This allows companies that control the wallet layer of the stack to add a revenue stream.

The governance layer adds another revenue stream, and potentially a much larger one.

Visa processes $14.2 trillion in payment volume annually, earning a roughly 0.28% take rate. This rate includes not just fees but also the implicit management fee Visa earns for the trust built by preventing fraud, resolving disputes, and enforcing network rules.

Applying even a fraction of this rate to agent transactions gives us an idea of the immense value it brings to companies built on the governance layer. McKinsey's $3 trillion projection for agent transactions by 2030 means that even a governance fee of just 0.1% (about 35% of Visa's take) could generate $3 billion in annual revenue. For reference, Coinbase's total subscription and services revenue in 2025 was approximately $2.8 billion. Revenue from just the agent transaction governance layer alone could match what Coinbase currently earns from staking, custody, and Coinbase One combined.

Companies with a presence in the wallet, settlement, and governance layers of the agent finance stack can profit from idle agent balances (float revenue), settlement fees (sequencer revenue per transaction), and compliance fees (enforcing governance).

This is why vertical integration across the entire technology stack will be the only business model that allows companies to remain competitive in the agent era.

Пов'язані питання

QAccording to the article, what was the main reason OpenAI shut down its AI agent shopping feature in March?

AThe main reason was the lack of proper rules to ensure a smooth shopping experience. Issues like what items agents could buy, who collects sales tax, how to identify fraud, and who handles returns were not properly resolved.

QWhy does the article argue that payment processing for AI agents has low economic value, and what solution does it propose for the settlement layer?

AThe economic value is low because the average AI agent payment over the past 12 months was only 31 cents, leaving very little profit after traditional fees (e.g., from Stripe, Visa). The article proposes that cryptocurrencies, specifically Layer-2 stablecoin payment systems, are well-suited for the settlement layer due to their much lower transaction costs (around $0.0001).

QWhat role does the governance layer play in the AI agent economy, and why is it considered a major source of value?

AThe governance layer manages rules like spending controls, identity checks, policy enforcement, audit trails, and liability assignment for AI agent transactions. It's a major source of value because it decides *if* and *under what conditions* transactions can occur, shifting profit potential away from mere payment processing. Controlling this layer allows companies to charge compliance fees for enforcing these rules.

QHow does Stripe's acquisition of Privy position it in the competition for the AI agent payment stack, particularly regarding the governance layer?

AStripe's acquisition of Privy, a company building embedded wallets, gives Stripe control over a key checkpoint (the wallet layer) where governance policies (spending limits, approvals) are enforced before funds move. This, combined with Stripe's other infrastructure (Bridge, Tempo, MPP protocol), allows for vertical integration across the stack and creates a significant distribution advantage through Stripe's existing merchant network.

QWhat are the potential revenue streams for a company that is vertically integrated across the AI agent financial stack, according to the article's conclusion?

AA vertically integrated company can profit from three main streams: 1) Float income (revenue from idle agent balances held in wallets). 2) Settlement/sequencer fees (revenue from processing transactions on its own layer, like Coinbase does on Base). 3) Compliance/governance fees (revenue from enforcing rules and policies on agent spending).

Пов'язані матеріали

Silicon Valley 'Startup Guru' Steve Hoffman: Web3 + AI Could Be a Trap

Silicon Valley investor and "Godfather of Startups" Steve Hoffman warns that combining Web3 with AI is likely a trap, not a promising venture. In an interview, Hoffman argues that while AI is a foundational technology touching all industries, Web3 adds complexity, friction, and regulatory risk without solving mainstream consumer or business needs. He advises founders to focus on deep, specialized applications where startups can out-iterate giants, rather than on generic features easily replicated by large tech companies. Hoffman observes that Silicon Valley will lead foundational AI research, while China excels at rapid, large-scale application and commercialization, particularly in robotics. He stresses that AI-driven autonomous agents capable of collaborative, multi-step tasks are 2-4 years away, which will cause significant job displacement. The solution is not to slow AI but to redesign business models around human-AI collaboration and reform social systems like education and retraining. For startups, Hoffman recommends focusing on vertical, expertise-heavy domains to build defensibility. He sees major opportunities in AI fraud detection and cybersecurity. Key founder mindsets include systemic thinking over feature-focus, relentless customer centricity, building adaptive teams, and deeply understanding AI's capabilities and limits. Hoffman is also leading a non-profit initiative to establish university centers aimed at training future leaders in responsible, human-value-aligned AI innovation.

marsbit4 хв тому

Silicon Valley 'Startup Guru' Steve Hoffman: Web3 + AI Could Be a Trap

marsbit4 хв тому

Token Inefficient, Economy Tokenless

The article "Tokens Aren't Economical, Economics Aren't Tokenized" analyzes a pivotal shift in the AI industry from a technology-driven narrative to one dominated by capital efficiency. It highlights two concurrent trends: a severe capital shortage due to the exorbitant and recurring costs of compute (e.g., OpenAI's high burn rate) and a wave of corporate spin-offs where major tech companies are separating their AI units (like Kuaishou's Kling and Baidu's Kunlunxin). The core argument is that AI's "anti-internet" business model, where user growth increases costs rather than profits, has created a disconnect between high valuations and actual cash flow. Spin-offs address this by allowing AI assets to be valued independently. Within a parent company, they are seen as cost centers, but as standalone entities, they are priced based on their growth potential and scarcity in the primary market, leading to massive valuation premiums (e.g., Kling's estimated value tripling post-spin-off). The industry is at an inflection point, moving from "model worship" to "value realization." The competition is evolving from a pure compute (GPU) race to a broader focus on systemic efficiency and full-stack engineering (involving CPUs and orchestration) to achieve viable commercialization. The year 2026 is framed as a critical moment where the industry must definitively answer how to economically translate AI capability into tangible business value, reshaping the sector's future power structure.

marsbit9 хв тому

Token Inefficient, Economy Tokenless

marsbit9 хв тому

Crossing the 'Memory Wall': The Wafer-Level Revolution and Computing Power Routes in the AI Inference Era

In 2026, a historic shift occurred in AI as major cloud providers' inference spending surpassed training spending for the first time, signaling a move from "building large models" to "using large models." This shifts the core challenge from computing power to the "memory wall"—the bottleneck of data movement (model weights, activations, KV Cache) between external DRAM and processors, where energy and latency from data transfer far exceed computation itself. Companies like Nvidia face GPU idle time due to bandwidth limits. In contrast, Cerebras Systems adopts a radical "wafer-scale" approach with its Wafer-Scale Engine (WSE). Instead of cutting a silicon wafer into many chips, Cerebras uses almost the entire wafer as one massive chip (WSE-3). This design provides 44GB of on-chip SRAM, delivering memory bandwidth thousands of times higher than traditional HBM (e.g., 21 PB/s vs. Nvidia B200). For LLM inference, weights are streamed layer-by-layer from external MemoryX storage to the chip, avoiding HBM bottlenecks. This results in token generation speeds 1.5–5 times faster than Nvidia's B200 in some models and significant advantages in first-token latency and long-context tasks. Additionally, Cerebras's architecture offers much lower interconnect power consumption (0.15 pJ/bit vs. GPU's ~10 pJ/bit). However, Cerebras faces challenges: SRAM scaling has slowed with advanced nodes, limiting future capacity gains; the chip requires specialized liquid cooling and custom software stacks; and its external I/O bandwidth (150 GB/s) is low compared to NVLink, hindering multi-system scaling for very large models. Competition is intensifying. Major players are pursuing three paths: 1) Developing proprietary inference ASICs (e.g., Google TPU, Microsoft Maia), 2) Leveraging advanced packaging (e.g., TSMC's SoW) to democratize wafer-scale-like integration, potentially eroding Cerebras's process advantage within a few years, and 3) Exploring optical interconnects for ultimate bandwidth. Commercially, Cerebras is transitioning from a hardware vendor to a service provider, facing the immense challenge of building high-power, specialized data centers to meet large contracts (e.g., 250MW/year from 2026–2028). In conclusion, the AI inference era presents a fundamental architectural trade-off. Cerebras opts for extreme physical optimization for low-latency, single-task performance, while Nvidia prioritizes versatility and massive cluster throughput. The path forward remains uncertain, with technology and business models still evolving in the race toward advanced AI.

marsbit15 хв тому

Crossing the 'Memory Wall': The Wafer-Level Revolution and Computing Power Routes in the AI Inference Era

marsbit15 хв тому

Has Bitcoin's 'Rebound Ended', Officially Entering the Late Bear Market Phase?

**Title: Has Bitcoin's Rebound Ended, Entering the Late Bear Market Phase?** **Summary:** Bitcoin's price has declined by 13% this week, signaling a potential return to late-stage bear market conditions. The price fell to around $67k, positioned between the Realized Price and Realized Cap Weighted Average. For the first time since early 2022, the Short-Term Holder cost basis has dropped below this key average, confirming a hallmark of late-cycle bear markets. Profitability metrics have collapsed sharply. The 7-day average of the Realized Profit/Loss ratio plummeted from a local high of 3.16 to 0.29, mirroring the February panic sell-off. Critically, the 90-day average never breached the threshold of 2, indicating the recent rally to $82k was a bear market bounce, not a structural shift. Realized losses surged to $1.35 billion daily, with $770 million coming from Long-Term Holders selling at a loss. This accelerating redistribution of supply from weak to strong hands is a necessary but ongoing process for a market bottom. The rally stalled almost precisely at the aggregate cost basis (~$83k) of US spot Bitcoin ETF investors, turning that level into strong resistance and leaving the average ETF holder underwater again. Spot market flows have turned decisively negative, showing sellers are dominating order books despite the price drop. While a significant futures long liquidation event cleared over $400 million in leverage, providing a potential reset, sustained spot demand is yet to materialize. Options markets continue to price in higher future volatility (Implied Volatility) than recent price action (Realized Volatility) has shown, with a persistent skew towards put options, indicating ongoing demand for downside protection. In conclusion, multiple metrics point to a fragile market structure. Resistance at the ETF cost basis, accelerating realized losses, dominant spot selling, and cautious options pricing all suggest the bear market trend persists. A sustainable recovery likely requires a resurgence of spot demand, ETF holders returning to profit, and a clear reduction in selling pressure.

marsbit15 хв тому

Has Bitcoin's 'Rebound Ended', Officially Entering the Late Bear Market Phase?

marsbit15 хв тому

TechFlow Intelligence Agency: Anthropic Calls for Global Pause in AI Development While Preparing for Trillion-Dollar IPO; SpaceX IPO Roadshow Heats Up, But S&P 500 Rejects Fast-Track Inclusion

In today's TechFlow Intelligence Briefing, several major tech stories highlight a growing theme of trust and credibility gaps across AI, crypto, and finance. AI company Anthropic has publicly called for a global pause in AI development, citing risks from Claude's "recursive self-improvement." Ironically, this coincides with reports the company is preparing for a massive IPO targeting a near $1 trillion valuation. This perceived hypocrisy, coupled with widespread user complaints about Claude's declining performance, is sparking debate over whether the safety warning is genuine or a competitive tactic. Meanwhile, in a substantive security move, Anthropic open-sourced a framework for AI-powered vulnerability discovery. In the crypto market, Bitcoin's price drop below $61,000 triggered over $1.16 billion in liquidations, flipping the market into a state where more BTC is held at a loss than at a profit, a historical bearish signal. On the corporate front, SpaceX's highly anticipated IPO is generating immense Wall Street excitement, with Goldman Sachs projecting 100x revenue growth by 2030. However, the S&P 500 has refused to fast-track the company's inclusion post-IPO, potentially limiting immediate institutional demand. Separately, ByteDance's AI app Doubao lost over 6 million monthly active users after introducing a subscription model, highlighting the challenges of AI monetization. Other notable developments include Nvidia certifying HBM4 memory from Samsung, SK Hynix, and Micron; Cloudflare's acquisition of front-end tooling company VoidZero; and its CEO warning that bot traffic now exceeds human traffic online. The underlying narrative connects these events: a trust crisis. From AI firms' contradictory actions and crypto volatility to the clash between SpaceX's hyped narrative and institutional rules, a pattern is emerging where stated intentions and actual practices are increasingly misaligned.

marsbit31 хв тому

TechFlow Intelligence Agency: Anthropic Calls for Global Pause in AI Development While Preparing for Trillion-Dollar IPO; SpaceX IPO Roadshow Heats Up, But S&P 500 Rejects Fast-Track Inclusion

marsbit31 хв тому

Торгівля

Спот
Ф'ючерси
活动图片