A new report making the rounds on Hacker News argues that the US federal government has quietly reorganized itself around one goal: winning the AI race. The piece, titled “AI First,” comes from the Climate and Community Institute and the Revolving Door Project, and it reframes the data center boom not as a free-market miracle but as a set of deliberate political choices. According to Hacker News, where the report climbed to a high score, the authors argue that the state is bending tax policy, energy infrastructure, and antitrust enforcement to serve the AI buildout ahead of the public.
What stands out here is the framing. Most AI coverage treats the buildout as inevitable, a natural consequence of better models and hungry demand. This report says the opposite. Subsidies, tax rates, permitting rules, and deregulation are policy levers, and someone is pulling them.
The three mechanisms
The report breaks the government’s role into three parts:
- AI First strategy. Treating the AI “race” as a national security priority, locking it into the bureaucracy through procurement, especially for military and surveillance work. The authors say weak antitrust enforcement and a revolving door of investors-turned-policymakers create a sense that AI is “too strategic to fail.”
- Antisocial infrastructure. Rolling back environmental rules and fast-tracking permits to wire the grid around data centers. The report warns that forcing coal, oil, and gas plants to delay retirement could cost ratepayers $3 billion and cause tens of thousands of excess deaths from pollution.
- Paying for the party. Low corporate and high-earner tax rates that free up cash for rapid construction. The headline number: Alphabet, Meta, and Amazon owed $65 billion at the full 21% corporate rate in 2025 but paid $15.3 billion, keeping nearly $50 billion.
Why this matters now
The timing is the story. Capital spending on data centers has become a load-bearing pillar of the stock market rally. When three companies can redirect $50 billion from public coffers into private compute, the line between industrial policy and corporate strategy gets blurry. The report calls it “a corporate takeover of state capacity.” You don’t have to accept that framing to notice the underlying facts are verifiable tax and energy numbers, not rhetoric.
There’s also a counter-current the report documents. In 2025, local opposition helped block or stall 48 data center projects worth $156 billion, often over pollution and public subsidies. That’s a real signal for anyone planning infrastructure. Community resistance is now a material project risk, not a footnote.
The other side
This is advocacy, and it reads like it. The authors want state capacity pointed at green infrastructure, public housing, and transit instead of GPUs, and they openly “reject the AI race” as worthwhile. Plenty of economists and policymakers would push back hard. They’d argue the compute buildout drives genuine productivity gains, that losing the race to China carries its own costs, and that cheap power and fast permitting benefit everyone, not just hyperscalers. The report doesn’t engage those arguments so much as dismiss them. Read it as one well-sourced perspective, not the final word.
Practical takeaways
For practitioners and businesses watching this space:
- Track the permitting and energy fights. Where your data center capacity comes from, and how contested it is locally, is becoming a supply-chain question. Factor community opposition into siting timelines.
- Watch the subsidy politics. Tax treatment and public incentives underpin the current capex pace. If the political mood shifts, the economics of aggressive buildout shift with it.
- Expect the narrative war to intensify. “AI race” framing versus “public good” framing will shape regulation over the next few years. Whichever side wins the story influences what gets built and who pays.
My read: the report overstates its case in places, but its core numbers are hard to wave away. The AI boom rests on public choices, and public choices can change. The next fight won’t be over model benchmarks. It’ll be over grids, tax bills, and who gets a say. You can find the full report and its sourcing at the original Hacker News discussion.