The rapid expansion of artificial intelligence infrastructure is raising questions about whether workers and consumers could be left paying for a boom that benefits technology companies and investors. Recent developments in China and the United States have intensified concerns about the financial durability of the AI sector, its effect on employment and the cost of electricity, and the risks to communities if investment slows.

Reuters reported that China’s AI and semiconductor stock boom had suffered a sharp reversal. The STAR50 index, which tracks Beijing’s AI and chip companies in a manner comparable to the Nasdaq, fell 40 percent in less than four months, eliminating roughly $480 billion in market value. Unitree, a Chinese robotics company whose initial public offering three months earlier was oversubscribed 8,000 times by small investors, lost half its value. “And now we see the bubble bursting,” Beijing fund manager Dong Baozhen told Reuters.

In the United States, Microsoft offered voluntary buyouts to 60,000 employees—about 27 percent of its global workforce—as the company plans to replace some work with AI, according to reports. Anthropic, the company behind Claude, disclosed an $80.7 billion net loss in its filing to go public, along with approximately $657 billion in infrastructure agreements involving Microsoft, Amazon and SpaceX. About $498 billion of those commitments are noncancelable, even if Anthropic does not use the computing capacity it reserved, according to coverage of the filing.

Microsoft and Alphabet subsequently reported figures that underscored the scale of their AI spending. Alphabet used $9.1 billion in free cash flow for the second consecutive quarter. NPR’s 1A calculated that AI-related deals announced on a single Monday totaled approximately $235 billion, almost 1 percent of U.S. gross domestic product, according to its report.

Economist Dean Baker’s weekly AI Bubble Monitor has also drawn attention to the financial pressure on companies supplying computing capacity. CoreWeave, which rents computing power to major AI companies, is considered to be so deep in junk-bond territory that it must reportedly pay 16 percent interest on long-term debt to attract investors.

A central assumption behind much of the current construction of data centers, power plants and transmission lines is that demand for computing will continue to grow. But Intel co-founder Gordon Moore proposed in 1965 that the number of transistors on an integrated circuit would double roughly every two years. Moore’s Law has broadly held, while the cost of digital hardware has generally declined as its capabilities have increased.

That trend raises the possibility that facilities being built for today’s AI systems could become technologically obsolete sooner than expected. The Wall Street Journal recently described a shift from conventional microchips toward wafer-scale computing under the headline, “The Microchip Era Is About to End. The future is in wafers. Data centers will be the size of a box, not vast energy-hogging structures.”

George Gilder wrote that the technology was approaching commercial production. “Cerebras of Palo Alto, Calif., used the concept in its WSE-3 wafer-scale engine,” he wrote. “The WSE-3 boasts some four trillion transistors—14 times as many as Nvidia’s Blackwell chip—with 7,000 times the memory bandwidth. Cerebras inscribed the memory directly on to the wafer rather than relegating it to distant chips and chiplets in high-bandwidth memory mazes. The company stacked up its wafer-scale engines 16-fold, thereby reducing a data center to a small box with 64 trillion transistors.”

The comparison is significant because a conventional data center can occupy a 700,000-square-foot building, consume gigawatts of electricity and use millions of gallons of water for cooling. A much smaller system, potentially powered by rooftop solar and batteries, could require less infrastructure. Several companies have also introduced AI products designed to run on desktop computers rather than in centralized data centers. Perplexity announced a product under the headline, “Perplexity Computer now runs fully local on Mac.”

If computing becomes substantially more efficient or moves onto local devices, communities currently building data centers, power plants and transmission lines could face economic disruption. The concern recalls earlier investment booms, including Dutch tulips, 19th-century railroads, the dot-com collapse and the 2008 housing and financial crisis.

The effects are also being felt in the labor market. Peter Thiel wrote in a 2009 Cato Institute essay, “I no longer believe that freedom and democracy are compatible.” Marc Andreessen has criticized the protections afforded to government employees, including civil-service protections and public-sector unions, which he described as “two layers of insulation against ever getting fired,” according to a podcast transcript.

OpenAI CEO Sam Altman wrote in a 2021 essay that, as AI advances, “even more power will shift from labor to capital.” That warning reflects concerns that automation could increase the economic power of business and technology owners while weakening workers’ bargaining position.

Amazon has eliminated approximately 30,000 corporate jobs since last October, following CEO Andy Jassy’s warning that AI would “reduce our total corporate workforce,” according to reports. One layoff tracker counts nearly 119,000 job cuts so far this year that were explicitly tied to AI automation—about four and a half times last year’s total, according to Intellizence.

Robotics is expanding alongside generative AI. Machines equipped with AI are being developed to assemble cars, sew clothing, clean streets and drive vehicles. The industry has also created demand for workers to wear sensors that record their movements so the data can be used to train robots intended to replace them, according to reporting by The Globe and Mail.

The economic question is how consumers will afford goods and services if automation eliminates large numbers of jobs. Workers are also customers, taxpayers, homeowners and parents, and their spending supports local economies. A reduction in their purchasing power could weaken the economic base that businesses depend on.

Earlier financial crises offer examples of how the costs of economic expansion can be distributed. Deregulation of the U.S. savings-and-loan industry during the Reagan administration contributed to behavior that culminated in a major crash. The cleanup cost taxpayers more than $100 billion, equivalent to about $1 trillion in today’s money. Regulators during that period made more than 30,000 criminal referrals to the Justice Department, which secured more than 1,000 felony convictions, according to BillMoyers.com and a PBS investigation.

Bank deregulation in the late 1990s was followed by the 2008 financial crisis. Millions of Americans lost their homes, retirement accounts declined sharply and workers lost their livelihoods. The scale of that crisis was described as 70 times larger than the savings-and-loan collapse, while few senior banking executives faced criminal prosecution, according to contemporary analysis and congressional records.

The AI industry is more deeply connected than previous technology booms to the military, intelligence agencies, national-security infrastructure and the strategic competition between the United States and China. OpenAI Chief Financial Officer Sarah Friar said at a Wall Street Journal conference last November that the company wanted a federal “backstop” for financing linked to its data centers. When asked whether she meant a government guarantee, Friar replied, “Exactly,” according to reported remarks.

The One Big Beautiful Billionaire’s Bill allows companies to deduct the full cost of data-center hardware in the year it is purchased rather than using the depreciation schedules that apply to many other industries. The measure has prompted scrutiny of the public support available to AI infrastructure and of the political relationships surrounding the industry. Elon Musk is spending almost half a billion dollars this year to support Republican candidates, according to reported estimates.

Consumers are already facing higher electricity costs. Residential electricity prices rose 6.9 percent last year, more than twice the inflation rate. Goldman Sachs analysts expect data centers to account for 40 percent of electricity-demand growth through the end of the decade, according to CNBC.

Thirteen Mid-Atlantic and Midwestern states have incurred $23 billion in consumer electricity-price increases through at least 2028 as a result of demand from AI data centers, according to Fortune. A homeowner in Manassas, Virginia, received a $281 electric bill in January, compared with approximately $100 the month before. Consumer Reports has also examined the effect of data centers on electricity bills, water use and local communities in its reporting.

If the current investment boom slows or collapses, consumers and taxpayers could be left paying for partially completed transmission lines, data centers and power plants. The same concern arose after the dot-com boom, when the internet proved to be a lasting technology even as the 1999–2002 market collapse damaged retirement accounts and created immense fortunes for entrepreneurs including Mark Zuckerberg, Jeff Bezos and Bill Gates.

The concentration of corporate power has also shaped the technology sector. In 1983, the Reagan administration directed the Justice Department, Federal Trade Commission and Securities and Exchange Commission to reduce enforcement of the Sherman Antitrust Act and other antimonopoly laws. The result was a period of extensive mergers and acquisitions. Reagan also supported the decriminalization of share buybacks, creating incentives that critics associate with reduced competition, weaker worker protections and declining service standards. Similar consolidation is now occurring in AI.

The central issue is whether the AI economy will generate broad prosperity or concentrate economic and political power among the people who own the machines. The technology could make the United States considerably wealthier, but workers could still end up with less economic security, less political influence and greater exposure to the costs of financial speculation.

The experience of the internet boom and the 2008 financial crisis demonstrates that technological and financial expansions can produce lasting benefits while imposing severe losses on households. As AI investment continues, the distribution of its gains and risks remains a central economic and political question.