Author: The Beauty of Bayesian

When political slogans meet industrial interests, only through the struggle between the American two parties can the true underlying logic of the AI capital market be seen.

After the conclusion of the 2026 U.S. midterm elections, a familiar narrative inertia reappeared in the secondary market: the Democrats won, and the U.S. AI industry would suffer; data center construction might be restricted, capital expenditures of tech giants would be impacted, and the AI bull market thus faces systematic risks.

This statement seems logically self-consistent. However, if we extend the historical cycle slightly, we will uncover a very worthy fact to ponder:

From 2022 to 2024, the U.S. president was the Democratic Biden. What happened to the U.S. AI industry from 2023 to 2024?

It was a comprehensive explosion of generative AI, a surge in computing power demand, massive expansion of capital expenditures by tech giants, and a historic profit explosion from core leading companies like Nvidia.

Did these epoch-making developments occur in another parallel space-time? If the Democratic party inherently opposes AI development, then why does the most significant commercialization explosion period of the American AI revolution coincide with the presidency of a Democratic president?

Of course, a market increase does not equate to completely correct government industrial policy. However, historical facts at least remind us that directly equating Democratic governance with the decline of the AI industry is overly coarse in logic and lacks sufficient empirical support. More important than simply labeling parties is a deep understanding of real policies, economic interests, and industrial transmission mechanisms.

Respecting History First: The Truth About AI Bull and Bear Markets During the Democratic Presidency

Let’s take a look at a few sets of intuitive market data.

In 2022, the Nasdaq Composite Index fell by 33.10%.

In 2023, the Nasdaq surged by 43.42%.

In 2024, the Nasdaq continued to rise by 28.64%.

During the two years from 2023 to 2024, the cumulative increase of the Nasdaq reached approximately 84.5%.

Next, let’s look at the representative leading company in the AI industry — Nvidia.

According to Nvidia's annual report submitted to the SEC, its data center business revenue:

FY2023: $15.005 billion;

FY2024: $47.525 billion;

FY2025: $115.186 billion.

In just two fiscal years, Nvidia's data center business revenue grew by about 7.7 times. Here, FY2025 reflects the business leap for the entire year of 2024, ending January 26, 2025.

Meanwhile, Meta's capital expenditure (including financing lease principal payments) reached $39.23 billion in 2024, and Microsoft's cash expenditure on property and equipment in FY2024 reached $44.48 billion. Although not all of this spending was directed towards AI, it fully verifies the significant investments of large tech firms in infrastructure.

These figures are not speculative opinions but officially disclosed audited financial data from listed companies. It must be emphasized that from 2023 to 2024, the U.S. was under a Democratic president with a bicameral Congress of divided parties, and it is incorrect to attribute the rise entirely to a single political party. Similarly, in 2022, when the Democrats simultaneously controlled the White House and both houses of Congress, tech stocks still recorded significant declines.

What does this indicate?

The bull and bear fluctuations of tech stocks have never been a one-variable function of party affiliation. The tech bear market of 2022 was closely related to the Federal Reserve's rapid interest rate hikes, soaring real interest rates, and severe valuation compression; while the AI bull market from 2023 to 2024 was driven by breakthroughs in underlying technologies, persistent upward revisions of corporate profits, expansions in capital expenditure, and the restoration of risk appetite.

If a single party could explain the fluctuations of tech stocks, then the severe downturn in 2022 and the violent surge in 2023 under the same president could not logically reconcile.

The Biden Administration's Actual Actions: The Chip Bill and Computing Infrastructure

More worth studying than the gains in the secondary market are the actual industrial policies during Biden's tenure.

First, in August 2022, the CHIPS and Science Act was officially signed into law.

This act allocated $52.7 billion to support the domestic semiconductor industry and related programs, with approximately $39 billion for semiconductor manufacturing incentives and about $11 billion for advanced research and development. This act received bipartisan support and was not solely advanced by the Democrats, but it did complete legislation and implementation during Biden’s term. One of its core objectives is to strengthen the U.S. advanced process manufacturing capabilities and reduce external dependencies in critical computing power supply chains.

Semiconductors are the core foundation of AI computing infrastructure. If the White House was completely opposed to AI industry development, it would be impossible to explain this billion-dollar scale advanced process subsidy bill.

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

Many market participants today refer to this executive order, often remembering only model safety testing, algorithm compliance, and government regulatory requirements. However, the complete document not only covers safety risk management but also explicitly includes support for cutting-edge AI innovation, promoting market competition, attracting top global talent, advancing research, and accelerating adoption of federal government systems. This executive order reflects a policy combination of governance and development: aiming to grasp the technological high ground while attempting to integrate it into a regulatory framework.

We can discuss whether its compliance costs are excessively high, or whether some security reviews have affected the R&D pace, but we must not misinterpret “promoting AI governance” directly as “stifling technological development.”

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

On September 12, the Biden administration personally convened major cloud providers, AI laboratories, data center operators, and utility companies for closed-door discussions on ultra-large computing clusters, grid expansion, clean energy assurance, interdepartmental project approvals, and labor shortages. Subsequently, the White House led the establishment of the AI Data Center Infrastructure Working Group, directly intervening to resolve construction bottlenecks. This clearly indicates that the federal government recognizes: the competition for comprehensive national strength in AI is not only a competition of algorithm models but also a tangible engineering competition of GPU supply chains, electricity energy, and infrastructure delivery capabilities.

Fourth, in October 2024, AI was fully integrated into the core of U.S. national security strategy.

The Biden administration officially released the NSM-25 National Security Memorandum, requiring the U.S. to maintain its leading position in advanced frontier AI and enhance collaboration among computational infrastructure, defense research, and national security applications. Even on January 14, 2025, just before Biden's departure from office, the White House signed Executive Order 14141, specifically accelerating the expansion of AI infrastructure, requiring the Department of Defense and the Department of Energy to assess sites suitable for building supercomputing centers on federal land.

Of course, executive orders do not equal project implementation, and policy blueprints cannot be directly equated to final benefits. But at least from the legislative text and governing trajectory, the so-called generalization that "the Democratic party completely opposes AI development" is seriously disconnected from objective facts.

Supporting AI Industry Technology Does Not Equal Unconditional Support for Tech Giants

It is crucial to introduce a key economic distinction: supporting the technological development of a strategic emerging industry is not the same as supporting all giants in the industry to retain monopoly profits.

While the Biden administration supports semiconductor localization and computing infrastructure, it has indeed executed extremely tough regulations on tech antitrust, algorithm governance, and advanced process export controls.

In January 2024, the U.S. Federal Trade Commission (FTC) officially launched antitrust investigations into the strategic collaborations between Microsoft and OpenAI, Amazon and Anthropic, and Google and Anthropic, focusing on whether supergiants are leveraging computing power resources, exclusive funds, and distribution channels to lock in market competition barriers.

In 2022 and 2023, the U.S. Department of Commerce also continuously tightened export controls on advanced process computing chips. These measures did directly squeeze the current revenues and global layouts of some chip manufacturers and represent non-negligible policy costs. However, antitrust and export controls are aimed at market concentration and national security adversities, and do not imply that the federal government wishes to stifle domestic cutting-edge innovation.

From the perspective of institutional economics, there are three layers of stakeholder conflict:

First, at the national level, there is a desire to maintain an absolute lead in cutting-edge technology;

Second, commercial giants wish to maximize shareholder profits through technological barriers;

Third, the general public hopes for equitable distribution of social costs and benefits brought about by technological development.

The three can overlap at certain stages, but there must be structural tensions in specific interest mechanisms.

Thus, a government can very well provide hundreds of billions in subsidies for domestic chip manufacturing while imposing strict restrictions on the flow of cutting-edge chips; it can also push for construction of computing centers while requiring tech companies to bear more public utility grid renovation expenses. Simplifying the multidimensional complex games into a binary “support” or “oppose” will only mislead investors' value judgments.

The Shift in Focus: Who Will Bear the Costs of AI Data Center Expansion?

Why has computing data centers suddenly become the core issue of local elections and congressional debates in 2026?

The most direct reason is that the large-scale expansion of AI infrastructure has begun to produce hard-to-ignore negative externality costs. The U.S. Department of Energy's 2024 research report indicates that the annual electricity consumption of data centers in the U.S. was about 176 terawatt-hours in 2023, accounting for 4.4% of the national total electricity consumption. The report predicts that by 2028, electricity consumption by data centers could climb to between 325 and 580 terawatt-hours, with its share skyrocketing to between 6.7% and 12.0%.

Although this is a forecast based on scenario modeling, it visually outlines the severe scale of resource consumption by computing power. Ultra-large data centers not only require Nvidia computing cards but also the supporting natural gas or nuclear power units, ultra-high voltage transmission lines, giant substations, closed cooling towers, water resource quotas, and contiguous industrial land.

Thus the core contradiction comes to the surface: who will bear the vast costs of expanding such infrastructure?

If tech giants exclusively enjoy the high market value and commercial profits brought by AI productivity, while the costs of substation upgrades and surging electricity loads are passed on to ordinary households through utility rates, local community public opinion backlash will quickly turn into an electoral political storm. This resistance does not stem from a contradiction to AI technology itself, but from real anxiety about rising electricity prices, water resource depletion, environmental pollution, and unfair distribution of public subsidies.

In January 2026, Democratic Senator Cory Booker and others submitted the “Power for the People Act,” with the core demand being to require large data center developers to independently bear all incremental costs relating to grid expansion, firmly cutting off the channel for passing these costs to ordinary residents.

However, it is worth noting that this regulatory claim is not unique to the Democratic camp. In September 2026, the U.S. House of Representatives overwhelmingly passed the Republican-led “Ratepayer Protection Act” with 417 votes in favor and 3 against, with numerous Democratic lawmakers also voting in favor. The bill subsequently failed to surpass the 60-vote threshold in the Senate procedural vote, garnering 57 votes to 43, with opposing Democratic senators primarily arguing that the bill lacked enforceability across states while Republicans insisted on maintaining states’ rights to regulate independently.

This legislative game fully illustrates that both parties share a high consensus on the basic goal of “protecting local residents from the transfer of high electricity prices,” with the core dispute solely revolving around the execution boundaries of federal intervention versus state autonomy. The complexity of the real political landscape is far from what can be simply summarized by party labels.

The Substantial Impact of Data Center Restrictions on Corporate Costs

Rational industry observation should not swing to the other extreme, simply dismissing all statements about limiting computing infrastructure as mere campaign slogans.

After the dust settles from the midterm elections on November 3, 2026, Reuters reported that the winning Democratic candidate for the Texas governor's election, Gina Hinojosa, had publicly promised to pause the approval of new large-scale data center projects until a clear community rights protection and consumer fee regulation framework was established. Given Texas's high weight in the U.S. pending computing capacity, such approval stagnation would inevitably disrupt project delivery timelines, cross-regional asset allocations, and internal rate of return models, even if it does not alter the ultimate computing demand.

However, another undeniable fact is that prior to this election, the former Republican governor of Texas, Greg Abbott, had already introduced measures to suspend certain areas’ permit approvals for data center grid connections and launched special review procedures. By September 2026, at least five states in the U.S. had Republican candidates explicitly proposing strict controls on data center expansion. This indicates that the trend to constrain computing infrastructure at the local level is a common product of energy pressures across the states, and attributing it solely to a party's ideology is not factual.

Whether the new governor’s campaign promises can be implemented will also be subject to multiple checks and balances from state legislative powers, public utility commission (PUC) regulatory protocols, existing compliance contracts, and judicial litigation. A media announcement of victory does not equate to immediate effect of administrative restrictions. Opposition to specific computing park site subsidies and denying the development of the AI industry are two distinctly different issues.

Yet market investors should not be complacent: even if the long-term demand for AI remains strong, forced project time extensions will lead to real capital costs and depreciation losses. For third-party data center developers that operate on high leverage and rely on rapid cash flow recovery, this asset delivery risk is particularly fatal.

The Common Economic Constraints Facing Both Parties on AI Issues

The current stage of AI has entirely transcended the realm of purely software applications.

It is deeply intertwined with advanced process wafer manufacturing, national cybersecurity, defense military applications, cloud supercomputing hubs, energy investment, cutting-edge scientific computing, and the global pricing power of multinational tech giants. This means that regardless of which party is in power, when steering executive authority, they must confront highly consistent objective economic constraints.

On the one hand, cutting-edge AI is recognized as a core lever to improve total factor productivity, stimulate capital expenditure, and reshape national competitiveness; on the other hand, massive computing clusters place severe pressure on electricity supply, water resource consumption, public financial subsidies, and labor employment structures.

Both parties have different policy preferences and governance tools at various levels, but no governing official can eliminate these objective physical and economic laws just based on campaign speeches.

In the secondary market, investors often use the personal holdings of congress members as a barometer for judging market direction. For example, disclosures of investments by members of the Pelosi family and several congressional representatives show significant exposure to tech stocks, which is often seen as evidence that policies will not become severely punitive. However, personal holdings come with limitations like delayed information disclosure, wide position ranges, and independent operation of trust accounts, and one cannot infer that because a member owns Nvidia stock, they will certainly block public utility grid regulatory legislation.

Personal asset allocations can only reflect individual economic incentives and cannot substitute for systematic tracking of congressional legislative votes, federal executive orders, and industry regulatory details. More important than focusing on what assets politicians hold is understanding what policy terms ultimately reconstruct various interest patterns.

Common Misunderstandings in Capital Markets: Equating Policy Narratives with Profit Changes

From a professional investment perspective, for a macro political news event to truly impact the intrinsic value of stocks, it must pass through a tight causal transmission chain: macro political events → formation of policy texts → administrative procedures implemented → changes in project microeconomic models → reshaping of expected future cash flows for companies → repricing of discount rates and valuation premiums.

In this chain, any delay or deviation in any link will result in final market outcomes that differ significantly from the initial emotional expectations. For example, if a state raises special fees for data centers to connect to the grid, it will directly lower the internal rate of return (IRR) of projects developed by heavy-asset data center developers; but if ultimately the ultra-large cloud providers absorb this premium in full, the actual procurement demand for underlying GPUs will not experience substantial shrinkage.

Similarly, while project approval delays may negatively impact new developers in specific regions, they paradoxically increase the scarcity premium of existing computing assets that have already secured ample electricity quotas and grid connection agreements. If the capital market generically labels all political statements as “significant negatives for the AI sector,” it will lead to severe mispricing among different subtitled targets.

Within the causal attribution analysis framework, we can break down the current market risks into three competing hypotheses:

Hypothesis 1 (Demand Suppression): The commercialization demand for AI terminals is substantially curtailed at the policy level. This hypothesis needs to find direct financial evidence to validate, such as tech giants revising down their annual capital expenditure, massive cancellations of server orders, or halving budgets for cutting-edge model training computing power;

Hypothesis 2 (Cost Restructuring): The underlying demand for AI remains high, but construction costs for infrastructure and energy bills are redeployed between upstream and downstream in the industry chain. This requires close tracking of trends in grid connection fees, computing delivery periods, developers' gross profit margins, and cloud service contract quotes;

Hypothesis 3 (Valuation Clearance): Political events merely serve as a trigger for crowded trades to repricing. The real core reason for a significant stock price drop is the prior extremely high valuation multiples, the excessively crowded long positions, and liquidity pressures brought on by leveraged fund liquidations and derivatives hedging.

The third scenario deserves especially high attention in the current market environment.

The history from 2022 to 2024 has fully proved: under completely identical macro political patterns, fluctuations in corporate profit cycles, tightening or loosening of monetary liquidity, are sufficient to create drastically differing market trajectories.

The AI sector in 2026 will be the same. Even if the industry retains vast long-term development space, high capital expenditures, expensive financing rates, and high Beta concentration in holdings pose immense technical adjustment pressures in themselves.

Therefore, a long-term bullish outlook on AI does not mean blindly bottom-fishing in any adjustments, nor does it imply ignoring the macro risk exposures of the entire portfolio.

Truly independent thinking in inverse investment must be firmly based on an independent examination of the underlying fundamentals, liquidity conditions, and valuation rationality, rather than betting on simple optimism or pessimism regarding a particular party.

Beyond Party Labels: Insights into Industrial Laws and Valuation Realities

Throughout the entire AI industry cycle, the cognitive trap that capital markets most easily fall into is to habitually use extremely simplified political narratives to explain extremely complex industrial evolutions.

Democrats winning leads the market to worry about regulatory tightening, and consequently to sell off tech stocks; Republicans winning leads the market to blindly bet on regulatory dividends, thus driving up asset prices.

Such emotional narratives possess high transmissibility on social media, yet often neglect the most critical intermediary variables in rigorous asset pricing models.

The grand AI bull market from 2023 to 2024 did not halt due to Biden's government issuing AI safety executive orders; the energy constraints faced by computing infrastructure in 2026 will not dissipate simply because Republicans emphasize technological hegemony.

Policies may affect the progress of power plant grid connections and land approvals, but they cannot erase companies' rigid demands for intelligent productivity; policies may squeeze the monopoly profits of specific intermediate links, but do not hinder the penetration of advanced processes throughout society; the project cycle delays brought by policies are tangible risk variables but should never equate to an overturning of industry logic.

A mature and in-depth investment analysis framework must layer and independently model political campaign slogans, legal boundaries, energy micro-models, corporate cash flow performances, and asset valuations.

Returning to the original core question: Does the Democratic party truly oppose AI?

From the detailed policy records and legislative documents from 2022 to 2024, it is impossible to derive the rough assertion that the Democratic party comprehensively curtails AI development. However, this does not mean that there do not exist regulatory forces within the political sphere advocating for tighter controls on giants due to environmental concerns, livelihood electricity prices, or antitrust worries. A more accurate reality is that America’s AI industry expansion has transitioned from an initial technological frenzy into a complex deep water zone deeply intertwined with computing infrastructure externalities and social wealth distribution.

For secondary market investors, what determines the life and death of investment performance has never been putting a cheap label of support or opposition on parties, but rather calmly verifying whether the technical demand can be continuously realized, whether tech companies can generate ample free cash flow, and how institutional changes reshape the profit cake of various links across the industry chain.

Do not misinterpret campaign slogans as industrial facts, do not directly equate short-term electoral victories and losses with profit guidance, and do not treat extreme market sentiments as rational investment conclusions.

If the transitions between bulls and bears over the past few years have left the market with the most profound insight, it is this:

The great evolution of cutting-edge technology can transcend long political cycles, but the valuation multiples and profit quality of any tech giant must undergo the harshest real testing in every specific cycle.

True independent thinking is not naively believing that any political party will forever shield the tech bull market, but rather being able to unravel technical trends under any policy environment, verify data evidence with first principles, and ultimately see where capital will flow.