Everyone's Regulating AI Infrastructure. Nobody's Using the Same Tools.

Something is happening simultaneously in Brussels, Beijing, Washington, and small-town Ohio that looks like a coordinated global response to AI. It isn't. The same problem — AI making tech infrastructure strategically critical — is hitting different governance systems, each of which reaches for the tool it already holds. The EU, a supranational union with five treaty powers, is using all five: competition law to designate AWS and Azure as "gatekeepers" under the Digital Markets Act ^[1][2]; trade policy to crack down on subsidized Chinese imports ^[3]; procurement rules to bar foreign cloud providers from public tenders in defense, healthcare, and energy ^[4]; cybersecurity regulation to revise its Cybersecurity Act ^[5]; and monetary policy to back a digital euro reducing reliance on US payment networks ^[6]. China, a centralized party-state, uses State Council decrees and capital controls: outbound investment regulations effective July 1, exit bans on senior AI professionals, and a blocked $2 billion Meta acquisition of AI startup Manus ^[7][8]. US federal agencies act individually — the FCC bars Chinese labs from testing American electronics ^[9], and FERC orders six regional grid operators to reform tariffs within 60 days ^[10] — because Congress, whose natural instrument would be antitrust law, has repeatedly failed to pass comprehensive tech legislation ^[11][12]. States and local governments fill the vacuum with the only tools they possess: Oklahoma requires 75MW+ data centers to fund their own power plants ^[13], Florida enacted the first comprehensive state hyperscale data center law ^[14], and towns from Dayton, Ohio to Hubbard Township impose moratoriums on new construction ^[15][16]. As one Ohio commissioner described the data center rush that local governments are trying to manage with zoning alone:

I think it's imperative … that we tap the brakes and we get our arms around what we're faced with and do the research, do the studies. — Jimmy Holcomb

The trigger is the same everywhere. FERC acted on the Energy Secretary's request to expedite grid connections to "remain competitive with China in the AI sector" ^[10]. Florida legislators cited AI data centers by name ^[14]. The EU framed its DMA expansion as future-proof for AI:

The DMA was designed to be future-proof and adapt to emerging challenges, for example in AI and cloud — Teresa Ribera

And the EU's Tech Sovereignty Package was explicitly motivated by the US Cloud Act and Microsoft's cancellation of the ICC prosecutor's email account ^[4]. The problem is shared. The responses are not. What connects jurisdictions is reaction, not coordination. China's Ministry of Justice ordered domestic firms to refuse cooperation with the EU's anti-subsidy investigation, calling it "unjustified extraterritorial jurisdiction" ^[17]. The US State Department instructed embassies to warn allies about Chinese AI "distillation" — training smaller models on outputs of larger American ones ^[18]. Each government is aware of the others' moves and reacts through its own instrument. No one is aligning. Coordination attempts either fail or dodge the regulatory question. The US-China Beijing summit in May 2026 produced two days of talks and no signed framework ^[19]. The G7 pledged AI coordination on child safety, but France's prime minister simultaneously demanded that France "must have its own tools" ^[20][21]. The US-led Pax Silica initiative — 35 nations including EU members — explicitly focuses on supply-chain capacity, not regulatory control:

India has the potential to be a comprehensive partner. — Jacob Helberg

India's representative said "right now it is still time for innovation" ^[22], suggesting Pax Silica actively opposes the regulatory tightening happening elsewhere. The coordination that succeeds is on building capacity. The coordination that fails is on controlling tech infrastructure. This pattern explains two things. First, why the responses look so different despite answering the same problem: each governance system can only use the authority it already holds. The EU has five treaty powers, so it uses five. China has a State Council, so it issues decrees. Municipalities have zoning, so they zone. Second, why the United States has the most fragmented response of all. The natural federal instrument for regulating tech power is antitrust law, and passing it requires congressional action — the hardest political lift in the American system. California's BASED Act, which would have prohibited self-preferencing by firms over $1 trillion market cap, was killed by lobbying from Apple, Google, and the Chamber of Commerce ^[12]. The American Innovation and Choice Online Act has failed in prior Congresses and is being revived again ^[11]. So the task falls to individual agencies and to state and local governments, each wielding whatever fragment of authority it holds. The world is converging on the same problem at the same time. It is not converging on a response.