OpenAI’s Tax Subsidy Efforts Amount to Silicon Valley Socialism

OpenAI’s push to expand the CHIPS Act tax credit into a broad artificial intelligence infrastructure subsidy shows how the government, and by extension taxpayers, are already underwriting the AI boom—just as the industry’s massive energy and land use burdens begin hitting state and local budgets.

Though some have been asking whether OpenAI wanted a government bailout, the company made a more ambitious request last month when it asked the government to expand tax credits intended to spur domestic semiconductor production to cover the entire AI infrastructure stack, from data centers to steel grid components.

Policymakers who want to support AI without handing over blank checks must rethink what a public-private partnership means in the AI context. Instead of doling out tax credits with no enforceable strings attached, Congress should tie any support to clear and measurable public benefit requirements.

In its letter to the Trump administration, OpenAI couched its request as a necessary step to ensure US leadership in AI. But stripping away the patriotic framing reveals something closer to venture capital-backed central planning.

A private company is seeking massive public subsidies for speculative infrastructure that will mostly serve its own bottom line. Combine that with data centers’ existing state-level tax breaks, their enormous energy demands, and the strains they put on local grids, and what do you get?

This is where the true public cost of this AI “boom” surfaces: higher utility bills, overloaded infrastructure, and taxpayer-funded corporate expansion gussied up as a national security strategy.

OpenAI’s request to expand the Advanced Manufacturing Investment Credit under CHIPS, a Biden era rule, would make AI infrastructure eligible for up to a 35% federal tax credit. The AMIC aimed to jumpstart domestic semiconductor fabrication—an industry that saw US manufacturing share drop from 37% in 1990 to 10% by 2022.

But AI data center projects already enjoy benefits from a patchwork of state and local subsidies. In Texas, Nevada, Ohio, and Wisconsin—where OpenAI’s Stargate data centers are being constructed—data centers routinely secure multi-decade property tax abatements, low-cost land, expedited permitting, and discounted electricity rates that shift costs to average ratepayers.

Layering an additional federal tax credit on top of those deals would do more than catalyze private investment—it would underwrite an entire industry while distorting local priorities. School districts give up tax revenue and utilities adjust where and how they generate power around siting for data centers. Voters are told this is all for economic development or jobs. But the jobs are few, and the power bill keeps climbing.

There’s nothing inherently wrong with industrial policy, but the government should only backstop the AI economy on terms that reflect the public interest.

Photographer: Kyle Grillot/Bloomberg via Getty Images

OpenAI’s letter acknowledges what has been obvious for some time: The US doesn’t generate nearly enough electricity to power big tech’s AI ambitions. They estimate we’ll need to bring online 100 gigawatts of new capacity per year—almost double the amount added nationwide last year—to support data center growth and close the “electron gap” with China.

The letter promises everything from trade training pipelines to portable expertise for US workers and implies that the AI we’re feeding electrons to may help modernize the very grid it’s overwhelming. But this is, at its core, a corporate strategy.

OpenAI’s infrastructure is proprietary, and the energy its projects consume is reserved. The supply chains it proposes to bring to the US will be vertically integrated wherever possible. This isn’t the Tennessee Valley Authority—it’s the Amazon.com-ification of our electrical grid.

There’s nothing inherently wrong with industrial policy, but the government should only backstop the AI economy on terms that reflect the public interest. Without benefit-sharing, clawbacks, transparency mandates, and job threshold requirements, all this talk about employment and economic opportunity is little more than patriotic marketing.

Recipient AI companies should have to report on energy use and emissions, share revenue with local communities, meet co-investment requirements for the grid, and agree to clawbacks when promised jobs or local benefits don’t materialize.

They should also be required to publish detailed impact assessments before construction begins. With the grid already strained, in high-demand areas they could be obligated to source a minimum share of newly generated energy from renewable sources rather than leaning on existing capacity.

At a bare minimum, any company receiving federal tax incentives should disclose overlapping state and local tax deals, in clear detail, to prevent double-dipping and the complete outsourcing of cost to taxpayers through opaque deals across levels of government. If OpenAI wants subsidies, that’s fine—but they should show their work and be made to bear their share of risk.

The public is already paying for AI through state and local tax breaks, rising electricity bills, climate impact, and now a federal infrastructure strategy that may be rewritten. The real question isn’t really whether OpenAI should get help, but whether anyone bothered to ask if we had to be the ones to help them.

Andrew Leahey is an assistant professor of law at Drexel Kline School of Law, where he teaches classes on tax, technology, and regulation. Follow him on Mastodon at @andrew@esq.social

Read More Technically Speaking