On September 23, if the report is accurate, Donald Trump will host Xi Jinping for two days of talks billed as AI-focused. The sentence that has traveled furthest is that the meeting "could redefine global AI standards" and reshape technology regulation and market dynamics. That is a large claim carried in a very small vessel. The report delivering it contains roughly five discrete information points, and four of them are hedged with "may" or "could." No agenda. No delegation list. No prior bilateral reference point. No attribution.
I spent the early part of my career reading contracts for a living — in 2017 I personally audited seven utility token smart contracts — and the habit does not retire when you change job titles. You learn to weigh a document by what it refuses to name. This one refuses to name compute. And compute is the only word in this story that has a price.
Strip the report to its load-bearing facts and you are left with four: a bilateral summit is scheduled; it is framed around artificial intelligence; the framing suggests possible movement on global AI standards and technology regulation; and markets may react. Everything else is atmosphere.
The source matters here. The report comes through a crypto publication — not a wire service, not a government readout, not any desk that maintains standing sources inside a national security council or a foreign ministry. That is not disqualifying on its own. Cross-border payment researchers like me read a wide range of outlets, and specialty media often surface signals before generalist desks notice them. But specialty media also carry a structural incentive: they publish for readers who hold, trade, or build with assets that move on geopolitical headlines. The audience for an AI-and-crypto story is not a policy audience. It is a trading audience. When the reader is positioned for a price move, the framing of the story bends toward the price move.
There is a second reason for caution. I cannot confirm this summit from my own knowledge base, and the report supplies none either. So the honest posture is to treat the event as a proposition under verification rather than a fact. That is not pedantry — it changes what the analysis is for. If the summit has happened, the useful work is determining what it produced. If it is scheduled but unconfirmed, the useful work is building the framework for reading it the moment it lands, and identifying which signals separate a photo opportunity from a structural shift.
Most coverage of high-level summits collapses that distinction. It reports the meeting as though the meeting were the policy. It is not. A bilateral summit produces direction at best. The chain runs from a leader's statement to a joint communiqué to departmental interpretation to drafting to enforcement — and that chain has a well-documented tendency to take between six and eighteen months, and to snap entirely when political friction rises. Anyone who has watched a regulatory framework travel from announcement to implementation knows the gap between the two is where fortunes are made and lost. The announcement is the noise. The draft rule is the signal. Follow the money, not the noise.
So let us do the work the report declined to do. If two heads of state sit down with AI on the agenda, what are the possible deliverables, ranked by how much they would actually change?
At the bottom sits the joint statement of principle. Two governments agree that AI should be safe, that humanity should benefit, that cooperation is valuable. This costs nothing, binds no one, and is routinely mistaken for a breakthrough by people who read headlines rather than instruments. At the top sits a change to the export control architecture governing advanced semiconductors — the entity list, the license thresholds, the treatment of China-specific chip variants engineered to sit just beneath the regulatory line. Nothing else on that ladder comes close in consequence.
That ranking is not a matter of taste. It follows from where the leverage physically sits. The United States has constrained Chinese access to leading-edge accelerators through export controls, and Chinese firms have responded by pushing domestic alternatives — Huawei's Ascend line, Cambricon, and a broader effort to substitute for the Nvidia parts they can no longer freely buy. Every serious analysis of the two countries' AI competitive position eventually reduces to that exchange. It is the hardest constraint in the system — harder than model quality, harder than data, harder than talent. You can route around a missing dataset. You cannot route around a missing fab.
And yet the report offers no indication that chips are on the table at all. It reaches instead for the language of standards, which is softer, more aspirational, and — for a trading audience — more useful as a narrative, because standards sound like rules and rules sound like outcomes.
Here is what the phrase "global AI standards" quietly elides. There is no single body that issues them, and no bilateral mechanism that could produce them.
The international standards landscape is contested territory with several active claimants: the ISO/IEC joint technical committee, the ITU, the United Nations' high-level advisory machinery on AI, and the separate standard-setting trajectories of the European Union and the United States. On top of that sits a layer of binding law that is national and regional in character — the EU's risk-tiered AI Act, the American pattern of executive orders supplemented by agency rulemaking, and China's own regime of algorithm recommendation rules and large-model registration filings. These are not three drafts of the same document. They are three different theories of what AI governance is for. One treats AI as a product to be risk-classified. One treats it as a market to be guided. One treats it as an information environment to be managed. No summit merges them, because they do not disagree about details. They disagree about purpose.
What a bilateral meeting can realistically produce is a channel. A hotline. A working group. A reciprocal commitment to notify before certain classes of frontier model deployment. An agreement to keep talking about catastrophic risk while continuing to compete on capability. That is not nothing — friction-management mechanisms carry real value, especially in domains like bioweapon-adjacent model capabilities, where both governments have an interest in not being surprised. But it is a great distance from redefining global standards, and that distance is where the mispricing lives.
There is also a governance point that gets lost whenever standards are discussed as a purely technical exercise. Standards are not neutral coordination devices; they are market-access instruments. A testing regime that a large firm can absorb and a small one cannot is a moat wearing a lab coat. An interoperability requirement authored by the incumbents of one jurisdiction is a tariff in technical clothing. When the report says the summit may affect technology regulation, it is describing something real — but it is describing a contest over who writes the rules, not a convergence toward shared ones. The ethical dimension is not decorative here. Follow the money, not the noise: the money in an AI standards regime is compliance, and compliance is a business model.
I care about this less as an abstract governance question than as a payments question, because that is the terrain I have spent this decade on. In 2020 I produced a long report on how unstable stablecoin pegs were propagating into cross-border remittance corridors in Latin America — the connection between an algorithmic yield incentive in one place and a family's grocery budget in another. That work taught me that the most revealing question about any technology policy is not what it announces, but which intermediaries it leaves standing.
Apply that question to an AI-focused bilateral. If the two governments move toward reciprocal rules on data flows, model deployment, or AI-generated content, the first places those rules touch ground are the financial and communications rails that already carry value across borders. Stablecoin issuance sits exactly there. So does on-chain attestation. So does decentralized compute. So does every project that has ever described itself as compliant by design.
And here is where my own experience turns skeptical. When I audited token contracts in 2017, the pattern that killed projects was rarely a bug in the code. It was a governance structure that assumed founders would behave well because the whitepaper said they would. When I studied DeFi liquidity mechanics three years later, the failure mode had evolved but not changed: incentives that looked self-correcting in a spreadsheet and turned predatory in production.
The AI-crypto intersection is repeating that pattern at a larger scale. In 2026 I worked on a framework for verifying AI-generated content on-chain — trustless attestation of provenance for model outputs, so a downstream system could check what produced an artifact without trusting the producer. I built it with a small group of cryptographers and AI researchers, and the hardest problem was never cryptographic. It was jurisdictional. A ledger is global; a regulator is not. The moment attestation standards become a matter of bilateral or multilateral agreement rather than open protocol design, the verification layer inherits the politics of the governments that wrote it.
That is the blind spot in the bullish reading of this summit. If AI standards tighten — even for legitimate safety reasons — the pressure lands on open-weight models, on permissionless verification, and on any system whose compliance story is "the cryptography proves it" rather than "the licensed auditor approved it." The crypto projects with the strongest claim to decentralization hold the weakest position in a standards negotiation, because you cannot send a DAO to a summit.
This is where I have to say something uncomfortable about the industry I work in. On-chain governance voter turnout has spent years stuck below five percent of token holders on most major proposals. The phrase community decision-making describes something that, in practice, is a handful of whales and venture funds pushing proposals across a quorum line. That is not a moral failing unique to crypto; it is what happens when you attach governance rights to capital and assume participation will follow. But it matters enormously here, because if AI standards and compliance frameworks become conditions of doing business, the entities that sit at the negotiating table will not be communities. They will be the ones with legal departments.
I have written before about the traceability problem: projects that preach decentralization while foundation wallets and team allocations move on a schedule any competent analyst can reconstruct. Those allocations are not hidden. They are merely unexamined. When a regulatory regime arrives, that traceability cuts both ways. It makes the decentralization claim falsifiable — and it makes the foundation a natural point of enforcement.
So expect this shape if standards work accelerates. Large, well-advised AI-crypto projects will embrace compliance and be rewarded with listings, banking access, and enterprise partnerships. Small, genuinely decentralized ones will face rising costs and drift toward jurisdictional arbitrage. And the middle — projects with real technology and no legal team — will be squeezed hardest. The cost of a rule is always paid by whoever can least afford to read it.
Here is where I part company with the consensus read.
The conventional interpretation of a Trump–Xi AI summit is directional and simple: a thaw is bullish for AI hardware with cross-border exposure, and bearish for the domestic Chinese compute-substitution narrative. Buy the supply chain, sell the substitute. That reading treats the summit as a policy event. I think it is more useful to treat it as a temperature reading.
Consider what a thaw would actually do. If the two governments establish a working channel and signal that some category of chips might be reviewed, the immediate effect is not a change in the physical supply of accelerators. Factories do not move on communiqués. The immediate effect is a change in urgency — and urgency, not capability, is what funds domestic substitution programs. A summit that looks conciliatory can slow the Chinese substitution pipeline precisely because it makes substitution feel less existential. Sentiment moves first, capex moves last, and capex is what determines who can train frontier models in 2030.
The deeper contrarian point concerns convergence itself. The prevailing assumption in crypto media is that AI governance and crypto governance are merging — that as AI becomes regulated, blockchain becomes the compliance layer, and everyone who arrived early benefits. I think the opposite is likelier. AI governance is being built by nation-states, in bilateral and multilateral rooms, using the instruments states actually possess: procurement, export control, licensing, standards bodies. Crypto governance is being built by nobody in particular, in public, with contested legitimacy and sub-five-percent participation. Those are not two systems converging. They are two systems on diverging tracks, and the AI track is the one with enforcement power.
That divergence has a specific consequence for anyone holding assets here. The value created by AI regulation will accrue to firms that can satisfy it. Much of the value created by crypto is the option not to need to. If the summit moves standards in a way that makes that option more expensive, the market will reprice the crypto assets whose thesis rested on regulatory ambiguity — and it will do so faster than it reprices the AI names the headlines will be about.
Meanwhile, the assets that survive a fragmented compute regime are the ones without a counterparty to sanction. Bitcoin's security model has always rested on fee revenue, and that revenue has never depended on any government's permission — which is why the inscription wave of 2023 and 2024 mattered more than the discourse admitted. It demonstrated that demand for blockspace can materialize from a direction nobody's regulatory map covered. In a world where compute is a licensed commodity, the ability to buy space in a blockspace market no one can gate is not a curiosity. It is the whole thesis.
Watch for four things, in this order. First, official confirmation from both governments — without it, the entire event remains a proposition, and the correct response is patience rather than positioning. Second, any wording that touches semiconductors, the entity list, or China-specific chip variants, because that language, not the word standards, tells you whether the meeting changed anything. Third, the output format: a communiqué is convertible into policy; a photograph is not. Fourth, the departmental follow-through in the six to eighteen months after the cameras leave, because that is where rules are actually written.
Everything before that is narrative. Volatility is the tax on impatience, and geopolitical headlines are the most expensive form of it.
Which leaves one question worth sitting with. If AI standards are written in rooms no decentralized protocol can enter, and compute is a commodity only states can license — what exactly is the crypto industry's seat at that table, and is it building one, or waiting to be handed one?