Stocks rallied after Democrats won Congress. For investors worried about the enormous cost of the AI race, the possibility of slower data-center construction offered something to celebrate: smaller spending bills.
For America’s long-term competitiveness, that same prospect raises a harder question. How does the country keep advancing AI if it becomes harder to build the infrastructure that runs it?
Blockster’s coverage of the election reactions explored the political fallout. The next question is physical: how much computing capacity can America actually bring online?
Why Markets Went Up
Wednesday’s rally had several drivers. Election uncertainty eased, oil prices fell, and investors bought large technology companies. Data-center operators moved in the opposite direction, according to Reuters’ market report. It was a divided reaction, rather than a vote of confidence in every part of the AI economy.
The logic behind Big Tech’s gains was revealing. Slower construction could reduce capital spending, protect cash flow and make existing computing capacity more valuable. Companies that already own large networks of data centers could benefit from a higher barrier to entry, as Reuters reported.
That is the uncomfortable possibility: restrictions presented as a check on Big Tech could help entrench it.
A shareholder can welcome a smaller construction budget. A startup trying to secure affordable computing capacity may see a more difficult market.
And the celebration was already proving fragile on Thursday, when AI shares fell and oil rebounded. One session’s rally cannot tell us whether the country’s infrastructure policy is working.
Trump Has a Veto. He Cannot Veto Every Obstacle.
Trump can veto federal legislation passed by a Democratic Congress. Overriding him requires a two-thirds vote in both chambers. Democratic control therefore does not automatically translate into a federal shutdown of AI development.
But his legislative veto does not apply to state and local permitting decisions.
Texas makes that distinction concrete. Governor-elect Gina Hinojosa campaigned on halting data-center approvals until lawmakers pass protections for residents. Republican Governor Greg Abbott had already imposed restrictions tied to a statewide audit, showing that resistance crosses party lines, as The Texas Tribune reports.
Post from @GinaHinojosaTXView the original post on X↗
A project can face delays over electricity connections, water use or environmental approvals without a new federal bill ever reaching Trump’s desk.
The risk is a slower, more expensive buildout—even if Washington never passes a national ban.
Businesses Are Already Paying. Can Infrastructure Keep Up?
The infrastructure debate is happening while businesses are rapidly increasing their use of AI.
On All-In, David Friedberg highlighted Anthropic’s growing customer base. Its SEC filing confirms the figures:
| Date | Anthropic customers spending $100,000+ over the preceding 12 months |
|---|---|
| December 2024 | Approximately 200 |
| December 2025 | Approximately 1,500 |
| September 2026 | More than 11,900 |
That is customer spending with one provider alone, rather than across the entire AI industry.
Anthropic says in its amended filing that demand already exceeds its available computing capacity. It also reported buying $5 billion of short-term compute in the third quarter at significantly higher costs than comparable long-term arrangements. Those pressures predate the election.
Slower construction would not suddenly switch off those customers. But if capacity falls behind demand, businesses could face tighter usage limits, slower expansion or higher costs than they otherwise would. Better efficiency can soften that pressure; it cannot guarantee that every shortage disappears.
China Is Building. America Has a Lead to Defend.
The strongest argument for accelerating U.S. construction starts with the lead America already holds.
Epoch AI’s October research estimates that the United States received 72% of global AI computing capacity arriving between 2022 and June 2026, compared with approximately 10% for China, including Chinese domestic production.
| AI hardware arrivals, 2022–June 2026 | United States | China |
|---|---|---|
| Estimated share of computing capacity | 72% | 10% |
| Capacity in H100-equivalent units | 19.8 million | 2.6 million |
These estimates measure hardware arrivals, not verified operating capacity. H100 equivalents adjust for differences in chip performance.
China is nevertheless expanding the infrastructure behind its ambitions. Chinese officials reported 52 computing facilities with more than 10,000 accelerator cards each by June, alongside more than 70 high-speed links connecting national computing hubs.
Its power buildout is also substantial. The International Energy Agency estimates China added nearly 500 gigawatts of renewable capacity in 2025, more than 60% of global additions. That is generating capacity across the economy, rather than dedicated, round-the-clock power for AI, but it demonstrates the scale of construction.
Post from @EpochAIResearchView the original post on X↗
China has problems of its own, including underused facilities and limitations in available chips. More buildings do not automatically produce better AI.
Still, an American lead is something to maintain through investment and execution. It is not a permanent entitlement.
Delaying Infrastructure Can Delay Capability
AI advances through better algorithms, more efficient hardware and experimentation. Expanding computing capacity supports that work and allows more people and businesses to use the resulting systems.
When power becomes the bottleneck, delayed infrastructure can restrict both development and deployment. As Blockster previously explored, energy is becoming a scarce input in the AI economy.
The IEA captures the competitive stakes directly:
“countries that can deliver the energy needed at speed and scale will be best placed to benefit.”
— International Energy Agency, Energy and AI
Its report notes that new transmission lines can take four to eight years to build in advanced economies. Political delays can compound an already slow process.
Residents should have enforceable protections for water supplies, electricity bills and local infrastructure. Developers should pay the costs their projects create. Those requirements should come with clear standards and a workable path to approval.
Democrats advocating moratoria need to explain how construction resumes. Republicans promising AI leadership need to deliver the power and grid connections that make it possible.
Markets can reward companies for spending less this quarter. America cannot secure its AI future by making it harder to build for the next decade.
Reporting by Sasha Steenkamp
