The Obsession with AI Investment in America

In the United States, there is a deep fascination with investing heavily in artificial intelligence (AI). High-profile figures, including President Donald Trump, are eager to assert American dominance in technology. Business leaders like Elon Musk are vying for control over influential companies such as OpenAI. CEO Sam Altman of OpenAI is intent on creating artificial general intelligence (AGI), which aims to replicate all human abilities. He calls for significantly increasing investments to achieve these ambitious goals.

Amidst this backdrop, a relatively unknown Chinese lab, DeepSeek, has showcased a more cost-effective and energy-efficient strategy for AI development. However, the U.S. tech industry seems to believe it is facing a critical moment, akin to the Soviet launch of Sputnik, leading to an increase in misguided lessons. These lessons suggest that the solution lies in spending even more on AI, distrusting Chinese technologies, and reaching back to outdated analogies from the coal industry of the 19th century to rationalize the excessive investments in the 21st century.

Despite demonstrated success in creating effective AI at a fraction of the anticipated cost, major companies continue to escalate their financial stakes. Last year, investments in AI are estimated to reach $230 billion. For this year, Amazon plans to allocate $100 billion on AI infrastructure, Alphabet aims for $75 billion, Meta could spend up to $65 billion, and Microsoft is set to spend $80 billion on AI data centers for the current fiscal year, with more funds expected for 2025. The “Magnificent Seven” tech firms will now invest more in capital than the U.S. government's total budget for research and development across all sectors.

Interestingly, while this surge of industry investment is occurring, the public sector in the U.S. is reducing its workforce and resources in pursuit of supposed efficiency. Ironically, part of the new administration's efficiency strategies involves substituting government employees with AI systems.

This raises the question: why hasn't this zeal for efficiency spilled into the private sector, where market competition would typically prompt such actions?

Three significant factors seem to trap the U.S. tech industry in this pattern of excessive investment.

One argument for increased funding stems from a modern interpretation of the Jevons paradox, which originated in the 1860s after the Industrial Revolution. Economist William Stanley Jevons posited that technological advancements that enhance the efficiency of coal usage would ultimately exacerbate England’s coal shortages by creating higher demand. This concept implies that as efficiency increases, prices decrease, resulting in heightened demand and greater necessity for coal.

This reasoning underpins the rationale of leading AI firms today. For example, Alphabet CEO Sundar Pichai mentioned in the Wall Street Journal that “we know we can drive extraordinary use cases because the cost of actually using it AI is going to keep coming down.” Microsoft CEO Satya Nadella echoed this sentiment, asserting on X that “Jevons paradox strikes again!” and emphasized his commitment to further spending.

There’s no doubt that we are still early in understanding AI’s full range of applications. However, it remains uncertain if simply making the tools cheaper will improve their adoption. A study by the Boston Consulting Group revealed that a mere 26 percent of companies have realized concrete value from AI implementation, notwithstanding the widespread excitement over its advancements.

Additionally, reliance on AI is waning. More than 56 percent of Fortune 500 companies have flagged AI as a risk in their annual reports to the U.S. Securities and Exchange Commission. Business leaders have not been able to demonstrate a satisfactory return on investment in AI to date.

Could more affordable AI lead to a greater demand for the technology itself, along with increased needs for data centers and advanced chips, as currently anticipated?

New cost-effective AI approaches are already available, such as those demonstrated by DeepSeek, which highlight ways to reduce computational needs through open-source models, utilizing a mix of AI experts, or optimizing processing by reducing the number of decimals in calculations.

Despite these innovations, leading AI companies have not explained why they aren’t revising their strategies or research and development budgets. Lower costs alone might not drive demand for extensive AI infrastructure as Jevons's theory suggests, especially when more economical methods to establish such infrastructure are available.

With vast amounts of money at stake, ignoring past technological disruptions where heavy investments from established firms resulted in significant value erosion could be a big mistake. Throughout history, incumbents have often overlooked new players equipped with more minimal investments but offering satisfactory, sometimes superior, solutions.

Consider the cases of Kodak’s downfall with digital imaging, BlackBerry’s decline in the face of the Apple iPhone, and Blockbuster being overtaken by Netflix, among others.

Another crucial factor influencing this trend is a shared reliance among major AI players. Each has gained short-term advantages from ramping up development spending. For instance, Google views generative AI as a major threat to its search engine business, compelling them to spend in defense of their top asset. Reports indicate that 2 million developers use its AI tools, resulting in billions in revenue from AI-related cloud services.

Microsoft's Azure AI has surged to around $5 billion in revenue last year, marking a 900 percent annual increase alongside a doubling of daily users for its AI-driven Copilot. Amazon has likewise generated billions from its AI cloud services and improved efficiencies within its online retail operations. Meta's CEO envisions becoming the leading assistant for a billion users while tapping into historic innovation and maintaining American technological supremacy. Practically, Meta also aims to capitalize on increased demand for data centers.

For Amazon, Google, and Microsoft, escalating AI investments boost demand for their cloud services. This interconnection fosters a cycle of mutual revenue growth, perpetuating the rationale for increasing investment. As long as each company believes that competitors will continue heavy spending, it becomes unwise for any single player to scale back, even if privately they harbor doubts.

This scenario reflects a suboptimal Nash equilibrium: a situation where all collaborators are trapped, and it’s not in any individual’s interest to stray from the status quo.

The third element steering the industry toward constant investment is the U.S. government's geopolitical strategy. The White House has reiterated its commitment to securing U.S. superiority in the AI field while curbing the influence of Chinese technologies. A noteworthy example is the launch of the ambitious $500 billion Stargate initiative, a collective venture to develop AI infrastructure led by SoftBank and OpenAI, introduced not in Silicon Valley but within the Roosevelt Room at the White House, shortly after Trump's inauguration.

Even though DeepSeek came into the limelight soon after, appearing to render such massive commitments excessive, construction for the first Stargate site is already underway in Texas. Vice President J.D. Vance recently promoted an aggressive AI opportunity agenda at the Artificial Intelligence Action Summit in Paris, cautioning against “cheap tech in the marketplace that’s been heavily subsidized and exported by authoritarian regimes.”

The Trump administration has also drawn on strategies from the previous administration to limit Chinese access to advanced chips. While the new administration seeks to allow U.S. companies to grow their AI capabilities without hindrance, it wields the potential to impose regulations, lawsuits, or tariffs on key components based on political alignment.

The emerging rules are evident: companies that comply with the administration and foster strong relations will be in advantageous positions to maneuver freely, gain government contracts, access federal AI funding, and negotiate opportunely with international regulators and other industry stakeholders.

Before the market bubble bursts, it would be prudent for at least one major company to signal a pause in escalating investments. The first step toward escaping a trap is recognizing one's entrapment. This is followed by acknowledging that competitive advantages within the industry may be shifting. Lastly, companies should have the courage to identify technology that is merely adequate rather than obsessing over the hardest computational challenges it can conquer.

Can one major corporation dare to lead the way and prioritize meaningful goals for AI that make a genuine difference in worker productivity—an aspiration that has eluded AI’s predecessor, the Internet?

Such a shift could inspire others to follow suit, creating a more beneficial Nash equilibrium. That would truly represent a breakthrough.

AI, Investment, Technology