Congress Didn’t Kill AI; It Raised the Cost of Data Centers
The 2026 midterms did not shut down the AI buildout. They changed the price of participating in it.
For AI builders, founders, and investors, the immediate effect is less about a federal ban than a harder operating environment: more permitting friction, more scrutiny of who pays for grid upgrades, and less confidence that tax treatment will remain automatic. Reuters reported that the election jolted AI investors because markets began repricing the risk that local and federal policy will slow data-center expansion. 1 CRS also makes clear that Congress has tools to influence the sector even where land and water sit mostly in state hands. 2, 3
The useful distinction is this: some effects are already visible in markets and state politics, while others are still proposals or expectations. That distinction matters if you are deciding whether the new Congress changes your data-center roadmap now, or only raises the odds that your next site, grid connection, or tax assumption gets tougher later.
What changed immediately
The biggest immediate shift is political, not legislative.
Reuters said Democrats won control of both chambers, and investors reacted by selling AI-linked names on fears that data-center growth would face tighter oversight and slower approvals. 1 At the same time, the legal and procedural constraints are real: the president still has veto power, and the Senate filibuster still makes sweeping federal AI legislation difficult. 4, 5
That means the market’s first move should not be confused with enacted policy. The election changed expectations. It did not instantly rewrite federal rules.
Still, expectations can matter fast. If developers think permits will take longer, grid charges will rise, or tax advantages may be narrowed, that changes project economics before any bill is passed.
The first bottleneck is still local
The cleanest near-term constraint remains permitting.
CRS says data-center land and water use are primarily state and local matters, even though Congress can use its Commerce, Taxing, and Spending powers to influence the sector. It also notes that proposed bills in the 119th Congress have included restrictions as blunt as barring new AI-related data centers, requiring off-grid electricity, or forcing non-local water sourcing. 6
That matters because local authorities can slow projects immediately. Zoning boards, utility interconnection queues, water permits, and state moratoria do not need a federal bill to bite. 1 Minute Signal coverage of the All-In Podcast described data-center permitting as a political choke point, with some states already pausing approvals pending new legislation. 7
"The speakers identify data-center permitting as a new political choke point, noting that some states are already halting new approvals until legislation is passed, which could effectively push frontier AI development to countries like Mexico or the Middle East."
— 1 Minute Signal coverage of All-In Podcast 7
That is the part builders should treat as real now. Even if Congress moves slowly, state and local friction can hit the schedule this quarter.
The more likely federal move is to raise costs, not halt projects
The better reading of the new Democratic majority is not “ban the data center.” It is “make the developer pay more of the bill.”
The legislative agenda Reuters reported is broad but still conditional: Democrats plan hearings, scrutiny, and likely AI bills in 2027, yet any major law still faces veto risk. 4 Vox similarly reported that stronger AI legislation is expected next year, but not as a guaranteed outcome. 8
The tax side is one obvious pressure point. Senator Peter Welch has asked the IRS to stop companies from using the R&E credit to subsidize AI data-center construction, arguing the credit was meant for uncertain R&D, not physical infrastructure. 9 Foley & Lardner also says Democratic tax-writing committees have identified AI and data-center taxation, including rolling back bonus depreciation, as a priority. 5
That does not mean those changes will pass exactly as described. It does mean the political direction is clear enough to alter planning assumptions. If you are underwriting a campus, a cluster of leases, or a debt package, the risk is not just “new regulation.” It is that existing incentives become narrower, harder to claim, or more contested.
Grid costs are where Congress can change the math
The most concrete incentive lever is electricity.
CRS says Congress has already made data-center energy consumption and grid reliability a central issue in the 119th Congress, and that federal authority can reach data-center economics through taxing and spending even when states retain core land-use power. 3, 6 That gives lawmakers a practical opening: shift grid-upgrade costs away from residential ratepayers and toward the operators driving demand.
Senator Martin Heinrich’s fact sheet is the clearest example. It says the bill would require large data centers to pay for both the electricity they consume and the grid upgrades needed to support them. 10 Brookings adds that the post-midterm climate could move such rules from voluntary to mandatory. 11
"The bill directs the Federal Energy Regulatory Commission (FERC) to require utilities to establish rates that force large data centers to pay for both the cost of serving them and the grid upgrades needed to handle them."
— U.S. Senate Committee on Energy and Natural Resources 10
This is the mechanism that matters most for incentives. It does not stop AI infrastructure. It makes the cheapest version of the buildout harder to finance. That should favor operators with stronger balance sheets and dedicated power strategies, while pressuring smaller or debt-dependent builders.
Hyperscalers can absorb more pain than neoclouds can
Markets appear to agree.
Reuters reported that investors sold AI-linked stocks after the election because they expected slower buildouts and more scrutiny. 1 1 Minute Signal coverage of The Wall Street Journal said traders are distinguishing between hyperscalers with deep capital and political heft, and smaller infrastructure plays more exposed to permitting and financing risk. 12
That split is the important one. A company with diversified revenue, self-funding capacity, and broad political influence can treat higher grid costs as a margin issue. A debt-heavy operator may have to slow expansion, raise capital on worse terms, or give up market share.
1 Minute Signal coverage of Brookings framed the AI buildout as a physical infrastructure cycle that standard disclosure does not fully capture, which is useful here because the political question is really about who absorbs hidden infrastructure costs. 13 1 Minute Signal coverage of Fireship added that capital-intensive infrastructure firms are already under debt pressure while end-user developers remain protected by subsidized pricing. 14
That is why the election matters even without immediate federal legislation. It changes the financing bar.
The CHIPS layer still matters, but it is not the whole story
One gap in a lot of post-election coverage is semiconductor incentives.
That matters because AI infrastructure is not just data-center shells and power contracts. It also depends on chips, packaging, and domestic supply chains. Reuters reported in July that the Commerce Department signed letters of intent for $874 million in CHIPS-funded semiconductor research tied to AI and advanced computing. 15 The CHIPS Program Office is still active in late 2026, with funding opportunities running into 2027. 16
Congress is also still debating how to extend or expand semiconductor tax incentives. A 2026 report on Section 48D says Senate Finance Committee leadership is pushing to extend the Advanced Manufacturing Investment Credit, which is already scheduled to expire in 2026. 17 And the broader CHIPS implementation record shows the program is still making awards, though slowly and with a wider scope. 18
This is relevant for AI builders because the new Democratic Congress is not just a threat vector. It could also reinforce industrial-policy support in parts of the stack that are politically easier to defend than giant data-center campuses. But that support is not the same as a free pass for data-center expansion. In practice, Congress may become more willing to support semiconductor manufacturing while becoming less willing to subsidize the full cost of local infrastructure externalities.
The real tension is federal power versus local resistance
The election is being misread if it is treated as a single national switch.
Federal Democrats can pressure tax treatment, oversight, and grid-cost allocation. States and localities can slow permits, add water constraints, and limit new approvals. Those are related but not identical forces. A congressional committee hearing does not itself delay a build. A county planning board or utility commission can.
That is why the immediate effect is mostly a change in expected friction, not a direct shutdown. Reuters described the market reaction as anxiety over slower buildouts. 1 Fox Business reported that strategists expect more scrutiny of AI providers and data-center development, especially around local power demand, but also noted that presidential veto power limits sweeping federal action. 19 The National similarly argued that while Democrats may try to roll back tax incentives and shift grid costs, a federal moratorium is unlikely. 20
For investors, that means the biggest question is not whether AI infrastructure survives. It is where the next unit of capacity gets built, who pays the interconnection bill, and which firms can survive a higher-cost regime.
What to watch next
Three signals will show whether this becomes a durable policy shift or just a market repricing.
-
Do proposed cost-shifting ideas become actual text?
If grid-cost mandates, tax-credit restrictions, or disclosure requirements move into committee language, the market will have to treat them as more than election rhetoric. But even then, passage is not assured because of veto and filibuster constraints. 4, 5, 10 -
Do states move faster than Congress?
State-level pauses, local permit fights, and utility-rule changes may matter more in the next few quarters than federal speeches. That is where immediate project delays are most likely to show up. 6, 7, 21 -
Do hyperscalers keep building while smaller operators get squeezed?
If so, the political shift will not stop AI infrastructure. It will concentrate it in firms that can self-fund power, absorb longer timelines, and survive a less generous tax and permitting regime. 12, 21
The bottom line is modest but important: the 2026 Democratic Congress is unlikely to kill AI infrastructure, but it may end the era of easy subsidies and frictionless permitting. That makes the buildout more selective. It also makes the winners more obvious: the firms that can finance their own power, tolerate more scrutiny, and wait out a slower political process.
"There is a significant gap between data center developers and the public regarding how much data centers should pay for shared infrastructure, and particularly for infrastructure built primarily, but not exclusively, for their benefit."
— Office of Senator Elizabeth Warren 22