BlackRock's 'Machine-Native Economy' Paper Mentions Bitcoin Once, in a Hedged Citation of a Prompt Survey: The 11-Page Thesis Is Stablecoins ($312B Supply, $11T 2025 Volume), x402 (>99.99% USDC, ~$0.13 Avg Payment) and Tokenized Compute ($1.1T by 2030), Not a Bitcoin Savings Account for AI Agents
What BlackRock actually published. On 22-23 September 2026 BlackRock released an 11-page whitepaper, "The Machine-Native Economy: How digital assets connect intelligence, commerce, and compute" (document code CE0926-M-5936357), authored by Will Su (Head of Digital Assets Research), Robert Mitchnick (Head of Digital Assets), Jay Jacobs (U.S. Head of Equity ETFs) and William Helm (Head of U.S. iShares Product Innovation). The phrase the headlines picked up is real: the executive summary says "AI represents machine-native intelligence, while digital assets represent machine-native money." But the paper is a stablecoin, payments-protocol and compute-tokenization thesis. Bitcoin appears in one paragraph on page 4, and only as a citation of someone else's work.
The one bitcoin sentence, and its hedge. The paper states: "Recent Bitcoin Policy Institute research offers preliminary support for this framework, reporting that model outputs across controlled simulations generally favored stablecoins for everyday payments and bitcoin for long-term value preservation. These findings reflect simulated model responses rather than observed agent behavior, but point to a potential AI-native monetary architecture in which stablecoins serve as transaction money and bitcoin as a store of value." That is the entirety of BlackRock's bitcoin argument. The parent card's claims that BlackRock argues bitcoin's "scarcity, global accessibility, and settlement independence make it a potential reserve or savings asset", and that "its volatility and transaction limitations remain obstacles", do not appear anywhere in the document. Neither "scarcity" nor "volatility" is used in the paper in relation to bitcoin. The parent card is paraphrasing a BeInCrypto headline, not the primary source; BlackRock did not make the case, it footnoted a lobby group's simulation and explicitly discounted it.
What the cited study measured. The Bitcoin Policy Institute study "Which Money Do AI Agents Prefer?" (Danielian, Brown, Egan, Zell; published 3 March 2026, moneyforai.org) ran 9,072 prompted scenarios across 36 frontier models from Anthropic, DeepSeek, Google, MiniMax, OpenAI and xAI, at temperatures 0.0, 0.3 and 0.7, using a system prompt that framed the model "as an autonomous economic agent". Results: bitcoin was chosen in 48.3% of all responses and 79.1% of store-of-value scenarios; stablecoins in 33.2% overall and 53.2% of everyday-payment scenarios; fiat or bank money in 8.9% overall; 86 responses volunteered energy or compute units (kWh, GPU-hours) as a unit of account. No model held a wallet or moved a satoshi. It is a text-preference survey of language models, which is exactly why BlackRock labelled it "simulated model responses rather than observed agent behavior". Treating it as evidence that agents are saving in bitcoin is a category error.
What agents actually pay with today. Observed agent behaviour exists and it is not bitcoin. Coinbase's x402 protocol, which BlackRock names as the emerging M2M standard, passed 100 million cumulative transactions through Q1 2026 (Chainalysis, 3 June 2026) and roughly 160 million by June 2026; over the 90 days to 24 August 2026 USDC accounted for over 99.99% of agentic transfer volume, with over 90% of transactions on Base and an average payment of about $0.13 (Crypto Briefing, 24 Aug 2026, citing Token Terminal). Payment sizes are also drifting up, not down: transactions of $1 or more went from 49% of x402 volume in early 2025 to 95% in early 2026, so the "sub-cent machine economy" is smaller than the pitch. Total 30-day x402 volume was reported at $24.24 million in August 2026, against $312.1 billion of USD stablecoins in circulation (DefiLlama, 23 Sept 2026: USDT $183.4B, USDC $75.2B). Agentic payments are a rounding error on stablecoin rails and zero on bitcoin rails.
The bitcoin-native alternative BlackRock omitted. Lightning Labs' L402 (HTTP 402 plus a Lightning invoice plus a macaroon token, implemented in the Aperture proxy) predates x402 and does the same job in sats; the whitepaper does not mention it, Lightning, or bitcoin as a payment asset at all. For value capture from settlement, the paper names "native cryptoassets (e.g. ETH)" and Circle's Arc, where USDC is the gas asset. On BlackRock's own framing, the machine-native money that captures agentic flow is the stablecoin and the chain it settles on; bitcoin's role, if any, is the balance-sheet asset an agent's principal chooses to hold, which is the existing IBIT thesis restated, not a new demand channel. Bitcoin base-layer economics do not help: the mempool.space recommended fee on 23 Sept 2026 is 1 sat/vB, i.e. the chain is empty, and it is empty because nobody is routing machine micropayments through it.
The compute leg. The paper's biggest number is not about crypto: sell-side consensus (Bloomberg, 31 Aug 2026) puts combined AWS, Microsoft Intelligent Cloud and Google Cloud revenue at roughly $1.1 trillion by 2030 (29% CAGR from 2025), cumulative AI capex above $5 trillion 2025-2030 (Goldman Sachs), and McKinsey has inference at 43% of data-centre power by 2030. BlackRock wants standardized compute claims tradeable as tokens and exchange-listed compute futures, while admitting "liquid markets for standardized compute contracts have yet to develop". For bitcoin holders the relevant read-through is the one covered yesterday: listed miners are the marginal sellers of both hashrate and power to that compute market, which is a headwind for hashrate, not a tailwind for BTC demand.
How to weigh it. Bullish: the largest asset manager, whose IBIT is the largest spot bitcoin ETF, put "stablecoins as transaction money and bitcoin as a store of value" in print, and its four authors include the two people who run BlackRock's digital-asset business. Bearish: they hedged it in the same sentence, cited a survey rather than data, and spent the other ten pages on assets BlackRock can tokenize and distribute (BUIDL-style funds, stablecoins, compute claims). Base rate: BlackRock research notes on digital assets have consistently framed bitcoin as a portfolio diversifier (the 2024 "Bitcoin: A Unique Diversifier" paper), and this one does not move that framing; it is a marketing-adjacent thought piece with a September 2027 expiry code, not a change to IBIT flows or to any BlackRock product. Net: neutral for BTC, incrementally positive for USDC/Base and for tokenization vendors.
Sources (6)
AI Research
Key Takeaway
BlackRock's 'Machine-Native Economy' whitepaper (22-23 Sept 2026) does not argue that bitcoin is AI agents' savings asset; its only bitcoin reference is a one-sentence, explicitly hedged citation of a Bitcoin Policy Institute prompt survey of 36 models, and observed agent payments (x402: >99.99% USDC, ~$0.13 average) run on stablecoins, not bitcoin.
What BlackRock actually published. On 22-23 September 2026 BlackRock released an 11-page whitepaper, "The Machine-Native Economy: How digital assets connect intelligence, commerce, and compute" (document code CE0926-M-5936357), authored by Will Su (Head of Digital Assets Research), Robert Mitchnick (Head of Digital Assets), Jay Jacobs (U.S. Head of Equity ETFs) and William Helm (Head of U.S. iShares Product Innovation). The phrase the headlines picked up is real: the executive summary says "AI represents machine-native intelligence, while digital assets represent machine-native money." But the paper is a stablecoin, payments-protocol and compute-tokenization thesis. Bitcoin appears in one paragraph on page 4, and only as a citation of someone else's work.
The one bitcoin sentence, and its hedge. The paper states: "Recent Bitcoin Policy Institute research offers preliminary support for this framework, reporting that model outputs across controlled simulations generally favored stablecoins for everyday payments and bitcoin for long-term value preservation. These findings reflect simulated model responses rather than observed agent behavior, but point to a potential AI-native monetary architecture in which stablecoins serve as transaction money and bitcoin as a store of value." That is the entirety of BlackRock's bitcoin argument. The parent card's claims that BlackRock argues bitcoin's "scarcity, global accessibility, and settlement independence make it a potential reserve or savings asset", and that "its volatility and transaction limitations remain obstacles", do not appear anywhere in the document. Neither "scarcity" nor "volatility" is used in the paper in relation to bitcoin. The parent card is paraphrasing a BeInCrypto headline, not the primary source; BlackRock did not make the case, it footnoted a lobby group's simulation and explicitly discounted it.
What the cited study measured. The Bitcoin Policy Institute study "Which Money Do AI Agents Prefer?" (Danielian, Brown, Egan, Zell; published 3 March 2026, moneyforai.org) ran 9,072 prompted scenarios across 36 frontier models from Anthropic, DeepSeek, Google, MiniMax, OpenAI and xAI, at temperatures 0.0, 0.3 and 0.7, using a system prompt that framed the model "as an autonomous economic agent". Results: bitcoin was chosen in 48.3% of all responses and 79.1% of store-of-value scenarios; stablecoins in 33.2% overall and 53.2% of everyday-payment scenarios; fiat or bank money in 8.9% overall; 86 responses volunteered energy or compute units (kWh, GPU-hours) as a unit of account. No model held a wallet or moved a satoshi. It is a text-preference survey of language models, which is exactly why BlackRock labelled it "simulated model responses rather than observed agent behavior". Treating it as evidence that agents are saving in bitcoin is a category error.
What agents actually pay with today. Observed agent behaviour exists and it is not bitcoin. Coinbase's x402 protocol, which BlackRock names as the emerging M2M standard, passed 100 million cumulative transactions through Q1 2026 (Chainalysis, 3 June 2026) and roughly 160 million by June 2026; over the 90 days to 24 August 2026 USDC accounted for over 99.99% of agentic transfer volume, with over 90% of transactions on Base and an average payment of about $0.13 (Crypto Briefing, 24 Aug 2026, citing Token Terminal). Payment sizes are also drifting up, not down: transactions of $1 or more went from 49% of x402 volume in early 2025 to 95% in early 2026, so the "sub-cent machine economy" is smaller than the pitch. Total 30-day x402 volume was reported at $24.24 million in August 2026, against $312.1 billion of USD stablecoins in circulation (DefiLlama, 23 Sept 2026: USDT $183.4B, USDC $75.2B). Agentic payments are a rounding error on stablecoin rails and zero on bitcoin rails.
The bitcoin-native alternative BlackRock omitted. Lightning Labs' L402 (HTTP 402 plus a Lightning invoice plus a macaroon token, implemented in the Aperture proxy) predates x402 and does the same job in sats; the whitepaper does not mention it, Lightning, or bitcoin as a payment asset at all. For value capture from settlement, the paper names "native cryptoassets (e.g. ETH)" and Circle's Arc, where USDC is the gas asset. On BlackRock's own framing, the machine-native money that captures agentic flow is the stablecoin and the chain it settles on; bitcoin's role, if any, is the balance-sheet asset an agent's principal chooses to hold, which is the existing IBIT thesis restated, not a new demand channel. Bitcoin base-layer economics do not help: the mempool.space recommended fee on 23 Sept 2026 is 1 sat/vB, i.e. the chain is empty, and it is empty because nobody is routing machine micropayments through it.
The compute leg. The paper's biggest number is not about crypto: sell-side consensus (Bloomberg, 31 Aug 2026) puts combined AWS, Microsoft Intelligent Cloud and Google Cloud revenue at roughly $1.1 trillion by 2030 (29% CAGR from 2025), cumulative AI capex above $5 trillion 2025-2030 (Goldman Sachs), and McKinsey has inference at 43% of data-centre power by 2030. BlackRock wants standardized compute claims tradeable as tokens and exchange-listed compute futures, while admitting "liquid markets for standardized compute contracts have yet to develop". For bitcoin holders the relevant read-through is the one covered yesterday: listed miners are the marginal sellers of both hashrate and power to that compute market, which is a headwind for hashrate, not a tailwind for BTC demand.
How to weigh it. Bullish: the largest asset manager, whose IBIT is the largest spot bitcoin ETF, put "stablecoins as transaction money and bitcoin as a store of value" in print, and its four authors include the two people who run BlackRock's digital-asset business. Bearish: they hedged it in the same sentence, cited a survey rather than data, and spent the other ten pages on assets BlackRock can tokenize and distribute (BUIDL-style funds, stablecoins, compute claims). Base rate: BlackRock research notes on digital assets have consistently framed bitcoin as a portfolio diversifier (the 2024 "Bitcoin: A Unique Diversifier" paper), and this one does not move that framing; it is a marketing-adjacent thought piece with a September 2027 expiry code, not a change to IBIT flows or to any BlackRock product. Net: neutral for BTC, incrementally positive for USDC/Base and for tokenization vendors.