Democrats oppose AI?

Original source: The Beauty of Bayesian

When political rhetoric meets industry interests: cutting through the U.S. two-party battle to see the real logic of AI capital markets. After the 2026 U.S. midterm elections, a familiar narrative has resurfaced in the market: the Democrats won, and the U.S. AI industry is doomed; data center construction may be restricted, the capital expenditure of tech giants will be hit, and the AI bull market therefore faces systemic risk. This argument sounds logically complete. But if we stretch the historical lens a little, we will find a question very much worth asking:

From 2022 to 2024, the U.S. president was Democrat Joe Biden. From 2023 to 2024, what did the U.S. AI industry experience?

It was the explosion of generative AI, a surge in computing power demand, the expansion of capital expenditure by tech giants, and historic profit growth for AI leaders such as Nvidia.

Did these things happen in another America? If the Democratic Party were naturally opposed to AI development, then why did the most important commercialization explosion of the U.S. AI revolution occur precisely during a Democratic president's term?

Of course, a market rally does not mean government policy is necessarily correct. But historical facts at least remind us: directly equating Democratic governance with AI industry decline is too crude in logic and lacks sufficient empirical support. What matters more than party labels is an understanding of real policies, economic interests, and industry transmission mechanisms.

1. First, Respect History: How Much Did AI Rally During Democratic Governance?

Let us first look at a few numbers.

In 2022, the Nasdaq Composite fell 33.10%.

In 2023, the Nasdaq rose 43.42%.

In 2024, the Nasdaq continued to rise 28.64%.

In just two years, 2023–2024, the Nasdaq's cumulative gain reached approximately 84.5%.

Now look at Nvidia, the most representative company in the AI industry.

According to the annual report Nvidia filed with the SEC, its data center business revenue was:

FY2023: $15.005 billion;

FY2024: $47.525 billion;

FY2025: $115.186 billion.

In just two fiscal years, data center business revenue grew about 7.7 times. FY2025 here ends on January 26, 2025, and mainly reflects business development in 2024. At the same time, Meta's capital expenditure in 2024 (including principal payments on finance leases) reached $39.23 billion, and Microsoft's cash spending on property and equipment in fiscal 2024 reached $44.48 billion. Not all of these expenditures were for AI, but they fully demonstrate that large tech companies were making enormous infrastructure investments at the time.

These figures are not media speculation but financial data formally disclosed by companies. It should be emphasized that from 2023 to 2024, the United States had a Democratic president and a divided Congress, so the rally cannot be attributed entirely to the Democrats. Likewise, in 2022, when Democrats controlled both the White House and both chambers of Congress, tech stocks still fell sharply.

What does this show?

Tech stock movements have never been a single-variable function of party politics. The 2022 tech bear market was closely related to rapid Fed rate hikes, rising real interest rates, and valuation compression; the 2023–2024 AI bull market was driven jointly by technological breakthroughs, upward earnings revisions, capital expenditure expansion, and a recovery in risk appetite.

If one party could explain tech stock movements, then the 2022 bear market and the 2023 bull market under the same president could not possibly be explained at the same time.

2. Did the Biden Administration Really Oppose AI? Look at What It Actually Did

More worthy of study than market gains is the actual policy of those years.

First, in August 2022, the CHIPS and Science Act was formally signed.

The act appropriated $52.7 billion for the U.S. semiconductor industry and related programs, of which about $39 billion was for semiconductor manufacturing incentives and about $11 billion for R&D. The act received bipartisan support and was not completed by Democrats alone. But it was indeed signed and implemented during Biden's tenure. One of its core goals was to strengthen U.S. semiconductor manufacturing capacity and reduce vulnerabilities in key technology supply chains.

Semiconductors are an important component of AI computing infrastructure. If the federal government at the time had been comprehensively opposed to AI industry development, it would be hard to explain the existence of such large-scale industrial policy.

Second, in October 2023, Biden signed AI Executive Order EO 14110.

Many people today mention this executive order and remember only safety testing, algorithmic risks, and government regulation. But the full document not only addressed risk management but also included support for AI innovation, promotion of competition, attraction of talent, and advancement of scientific research and government applications. This executive order reflected a policy mix: it wanted AI to develop and also wanted AI to be governed to some degree. One can debate whether its compliance costs were too high or whether some safety requirements suppressed innovation. But one cannot directly rewrite "regulating AI" as "opposing AI technology development."

Third, in September 2024, the White House actively coordinated AI data center construction.

On September 12, the Biden administration convened hyperscale cloud providers, AI companies, data center operators, and power companies to discuss large AI data centers, grid construction, clean energy, project approvals, and labor demand. Subsequently, the White House pushed to establish an AI data center infrastructure task force to coordinate relevant departments in resolving construction bottlenecks. This precisely proves that the Democratic administration at the time had already realized that AI competition is not only competition over algorithms but also competition over GPUs, energy, power, data centers, and infrastructure delivery capacity.

Fourth, in October 2024, AI formally entered the core of U.S. national security strategy.

The Biden administration issued National Security Memorandum NSM-25, requiring the United States to maintain leadership in advanced AI and coordinating computing infrastructure, R&D, and national security applications. Even on January 14, 2025, on the eve of Biden's departure, he signed EO 14141, specifically promoting U.S. AI infrastructure construction, including requiring the Department of Defense and the Department of Energy to identify federal land suitable for building advanced AI data centers.

Of course, executive orders do not mean projects have been built, and policy goals are not the same as final outcomes. But at least from policy texts and administrative actions, the generalization that "Democrats oppose U.S. AI development" does not match these facts.

3. The Real Difference: Supporting AI Technology Does Not Mean Unconditionally Supporting Tech Giants

Here an important distinction needs to be introduced: supporting the development of an industry is not the same thing as supporting all companies in that industry in retaining all profits.

The Biden administration, on the one hand, supported semiconductors and AI infrastructure, and on the other hand, also strengthened tech antitrust enforcement, AI safety governance, and export controls.

In January 2024, the U.S. FTC formally investigated partnerships such as Microsoft with OpenAI, Amazon with Anthropic, and Google with Anthropic, focusing on whether large tech companies used advantages in funding, computing power, and contracts to restrict competition.

In 2022 and 2023, the Biden administration also gradually tightened export controls on advanced computing chips to China. These policies may indeed affect tech companies' revenue, costs, and competitive advantages, and cannot be glossed over. But they targeted market structure, national security, and technology diffusion, and do not mean the U.S. government wanted the domestic AI industry to stop innovating.

From an economic perspective, there are three different interests here:

The nation wants to possess advanced AI technology.

Companies want to use AI technology to create profits.

The public wants the costs and benefits brought by AI development to be reasonably distributed.

The three can broadly align, but they cannot always be perfectly aligned.

Therefore, a government can absolutely subsidize domestic semiconductor manufacturing while restricting some high-end GPU exports; promote AI data center construction while requiring large tech companies to bear more grid costs. Compressing these policies at different levels into a single sentence of "supporting AI" or "opposing AI" would instead obscure the questions truly worth investors' study.

4. The 2026 Controversy Has Essentially Shifted: Who Pays for AI Expansion?

Why did data centers suddenly become an important issue in U.S. local elections in 2026?

The answer first lies in the fact that AI infrastructure expansion has begun to generate increasingly significant external costs. Research published by the U.S. Department of Energy in 2024 showed that in 2023, U.S. data centers consumed about 176 TWh of electricity, accounting for 4.4% of national electricity consumption. The study projected that by 2028, data center electricity use could reach 325–580 TWh, accounting for 6.7%–12.0% of national electricity consumption.

Note: this was a scenario projection at the time, not electricity consumption that had already materialized, but it clearly revealed the scale of the problem. Data centers need not only GPUs but also new generation capacity, transmission lines, substations, cooling systems, water resources, and land.

Who bears these infrastructure costs? If large tech companies obtain computing power and commercial profits while part of the grid construction cost is passed on to ordinary households through utility rates, local political resistance will naturally rise. This resistance does not necessarily come from antipathy toward AI technology; it may also come from concerns about electricity prices, land use, environmental protection, and public subsidies.

In January 2026, Democratic Senator Cory Booker and others proposed the Power for the People Act, requiring large data centers to bear the relevant grid expansion costs and reduce the burden on ordinary residents.

But it is worth noting that this position is not unique to Democrats. In September 2026, the House of Representatives passed the Republican-led Ratepayer Protection Act by a vote of 417 to 3, with many Democratic lawmakers also voting in favor. The bill subsequently failed to reach the 60-vote threshold needed to advance in a 57–43 procedural vote in the Senate. Democratic senators opposed argued that the bill only required states to consider relevant standards and lacked mandatory force; Republicans emphasized that state regulatory authority should be respected.

This episode shows that the two sides have considerable overlap in the goal of protecting residential electricity bill payers, but they still differ on the strength of constraints and the manner of implementation. Real political divisions are often much more complex than the two-party confrontation summarized by the media.

5. It Cannot Be Denied: Democratic Candidates' Data Center Restriction Proposals May Indeed Have Real Impact

Rational analysis cannot go to the other extreme either—believing that all statements restricting AI infrastructure are campaign slogans.

After the November 3, 2026 midterm elections, Reuters reported that Gina Hinojosa, the Democratic candidate who won the Texas gubernatorial election, promised to suspend approvals for new data center projects until clearer community and consumer protection rules are established. Texas accounts for a significant share of planned U.S. data center capacity. Therefore, even if an approval pause does not affect ultimate AI demand, it may change project delivery timelines, regional allocation, and financing returns.

But another fact is equally noteworthy: before this election, Republican Texas Governor Greg Abbott had also taken measures to suspend some new grid connections and strengthen review of data center projects. In September 2026, Republican candidates in at least five states proposed or strengthened data center restriction proposals. This makes it lack sufficient basis to interpret local data center controversies entirely as "Democrats opposing AI."

Whether new gubernatorial campaign promises are implemented also depends on administrative authority, state legislatures, public utility commissions, existing contracts, and judicial review. Media declarations of victory do not mean the relevant restrictions have formally taken effect. Opposing the siting or subsidies of certain data center projects is not the same proposition as opposing AI itself. But it equally cannot be ignored: even if long-term AI demand remains strong, project delays may cause real cash flow losses. For data center developers with high leverage and reliance on project commissioning and continuous financing, this risk is especially important.

6. On AI, the Two U.S. Parties Share an Economic Constraint That Cannot Be Ignored

AI is no longer a purely software technology issue.

It relates to semiconductor manufacturing, national security, military applications, cloud computing, energy investment, scientific research competition, and the competitive position of large U.S. tech companies in global markets. This means that no matter which party holds political power, it must face multiple simultaneous economic incentives.

On the one hand, advanced AI may raise productivity, expand economic output, and create new industrial opportunities. On the other hand, AI infrastructure requires large amounts of electricity, capital, and public resources, and may also bring employment shocks, market concentration, and cost pass-through problems.

The two parties do not attach exactly the same degree of importance to these issues or use exactly the same policy tools, but neither can eliminate these constraints with a single campaign slogan.

Here one can also discuss the personal investment holdings of political figures. The Pelosi family, some Democratic lawmakers, and related congressional trading-tracking ETFs have observable exposure to tech stocks, which is an interest thread worth studying. But public holdings involve issues such as disclosure lags, wide amount ranges, and the need to distinguish family members' holdings from lawmakers' own holdings. Even less can one assert that a lawmaker will certainly not support regulation affecting data center profit margins simply because their family holds NVDA.

Investment holdings can provide information about economic incentives, but they cannot replace direct analysis of legislative votes, administrative actions, and policy implementation. More important than observing what is in politicians' pockets is examining what policies they ultimately allow to take effect.

7. The Most Common Mistake in Capital Markets: Equating Policy Narratives Directly with Earnings Changes

From an investment perspective, for a piece of negative political news to truly affect stock value, it must at least pass through a transmission chain: political event → policy text → administrative implementation → project economics → corporate earnings expectations → valuation and risk premium.

Any link may break, and may also differ from the market's initial speculation. For example, a state raising data center grid connection costs may cause developer project IRRs to decline; but if ultimately cloud providers simply bear more energy costs, GPU demand may not necessarily decline in tandem.

For another example, project delays may be bearish for data center developers in a specific region, yet make existing facilities that already have sufficient power and interconnection capacity more scarce. If capital markets simplify all these differences into "bearish for the AI sector," mispricing among different companies may occur.

Using a causal attribution framework, the risk can be broken down into three competing hypotheses:

H1: Final demand for the AI industry is substantially suppressed by policy. Evidence to watch includes companies cutting capital expenditure, order cancellations, and downward revisions to computing power demand.

H2: AI demand still exists, but infrastructure construction costs are redistributed. Watch grid connection fees, project delays, developer profit margins, and investment returns.

H3: The political event is merely a trigger for market repricing. What truly amplifies the decline may be high valuations, crowded positioning, deleveraging, options hedging, and insufficient liquidity.

The third case deserves particular vigilance.

The experience of 2022–2024 has already shown that corporate earnings, monetary policy, and valuation changes can produce radically different market outcomes under the same political environment.

The same is true of the 2026 AI market