1 SEP 2026 — Build American AI, affiliated with the Leading the Future super PAC funded by Marc Andreessen, Ben Horowitz and OpenAI president Greg Brockman, is buying multi-million-dollar advocacy advertising in Kansas, Ohio and Wisconsin. AI policy gets written in Washington, Sacramento and Albany. Kansas, Ohio and Wisconsin are where data centres get planning permission, and that fight is about zoning, water and electricity rates.
What is being funded
Leading the Future launched with more than US$50m. Build American AI is the affiliated advocacy vehicle, and its campaign is aimed at persuading voters of the virtues of data centres rather than at any named candidate.
Andreessen Horowitz has disclosed more than US$115m in federal contributions for the midterm elections, which exceeds the disclosed federal giving of donors including Elon Musk and George Soros. OpenAI publicly distanced itself from Leading the Future in June, despite Brockman's involvement, and has said that groups advocating on AI should be clear about their policy views, honest about whom they represent, and should not use tactics like astroturfing.
The state list is the tell
Federal AI legislation is drafted in Washington, and the most consequential state statutes have come from California, Colorado and New York. Kansas, Ohio and Wisconsin appear on none of those lists.
Where those three states do appear is on the shortlist for large campuses. They have land, transmission capacity and, in Ohio's case, an established cluster already generating local disputes. The decisions that determine whether a campus gets built are made by county commissions, utility regulators and zoning boards, and those bodies respond to local sentiment in a way federal legislators do not.
What an AI model may do barely enters into it. The question in front of these voters is whether a county approves a 500-acre building with a substation attached, and that decision turns on water withdrawal, ratepayer cost allocation and property tax abatements.
Why an issue campaign rather than candidate spending
Advertising that urges a policy position without expressly advocating the election or defeat of a candidate falls under lighter disclosure requirements in most jurisdictions. Organisations running it need not name their donors the way a candidate committee must, and the legal line between the two categories carries real consequences.
It is also more durable. A candidate campaign ends on election day. A campaign that shifts public opinion on whether data centres are good for a county keeps working through every planning hearing afterwards, including the ones nobody reports on.
Lighter disclosure and work that continues after the election are both reasons sophisticated money reaches for issue campaigns. Naming it accurately is not an accusation of wrongdoing. Issue advocacy is lawful, common and used by every side of every infrastructure argument.
What the local argument is actually about
The objections that stop a data centre are rarely about artificial intelligence, which is why an advertising campaign about the virtues of AI is an odd instrument for winning them.
The same three objections recur: electricity, water and tax. For electricity, the question is who pays for the transmission upgrades a large new load requires. If the cost goes into the general rate base, every household in the service territory subsidises a single customer's connection. On water, evaporative cooling consumes rather than returns what it draws, which matters in a way that industrial water use with a discharge permit does not. On tax, abatements granted to attract the investment reduce the school and county revenue the same development is meant to generate.
Each of these objections has a counter-argument. Load can be served under a special contract that isolates the cost; closed-loop and air cooling exist and are increasingly used; abatement terms are negotiable and can be tapered. Each answer is specific, checkable and negotiated locally, in a hearing room, by people who will read the numbers.
The OpenAI position is awkward
OpenAI has distanced itself from the super PAC and criticised astroturfing while its own president is one of the named backers. The company has not addressed how those sit together.
One defensible reading is that an executive's personal political giving is their own. Companies routinely decline to be associated with vehicles their officers fund privately, and the alternative of policing those donations raises its own problems.
There is also a reading in which the distance is the point, because it lets the argument be made without the company owning it. Nothing in the public record settles which applies here.
Why this matters where data centres are actually being built
Johor, Batam and the Singapore corridor are seeing the same siting questions arrive with the same substance: power availability, water for cooling, grid cost allocation and what a county or a state gets in return. The regulatory venues are entirely different; the substance of the objections is the same.
What has not yet arrived in this region is the campaign layer — organised, funded advocacy aimed at residents ahead of the approval rather than at officials during it. On the American evidence it arrives once the local objections start winning, which is a stage several ASEAN projects have not reached.
We reported that the Johor-Singapore Special Economic Zone is being built around data-centre and AI infrastructure investment. What is worth taking from Kansas is less the advertising than the timing. Operators there expect siting to become a public argument and are spending against that expectation a year before it happens.
View author profile → · Editorial policy
