The artificial-intelligence boom is entering a more complicated phase. The central question is no longer whether companies can build larger models, but whether governments and communities will accept the infrastructure, strategic risks and economic disruption required to operate them.

That tension is visible on both sides of the Atlantic and across the U.S.-China relationship. Financial analysts continue to view hyperscalers and AI infrastructure firms as attractive investments, while policymakers are increasingly focused on energy use, supply-chain dependence and national security. The result is a contest between acceleration and control—one that will shape the technology’s next phase.

From software story to industrial policy

AI now depends on physical assets: advanced semiconductors, electricity, cooling systems and large data centers. The technology’s economic promise has therefore become inseparable from industrial policy. The Financial Times reported that former Federal Reserve official Kevin Warsh argued the AI boom could sustain U.S. growth at a level substantially above its long-run average, without necessarily producing equivalent inflation.

That optimism reflects the productivity case. If AI enables workers to complete routine cognitive tasks more quickly, companies could produce more with the same labor force. It could also accelerate research, logistics and software development. Investors are betting that these gains will justify enormous capital spending by the largest cloud providers.

But the gains are not guaranteed to arrive evenly. Productivity improvements may accrue first to firms that control computing capacity and proprietary data, while smaller companies face higher costs to access models. Workers whose tasks are easily automated may experience weaker bargaining power even if national output rises. A higher-growth economy does not automatically mean a more equal one.

The security cost of scale

The U.S.-China technology rivalry adds another constraint. Advanced chips are not merely commercial products; they are inputs into military systems, intelligence analysis and autonomous technologies. The Financial Times has described the AI chip relationship with China as a delicate balance, reflecting the difficulty of restricting sensitive equipment without severing a market that remains important to global companies.

Supporters of export controls argue that limiting access to the most capable chips can slow the development of military AI and protect a strategic advantage. Critics counter that broad restrictions encourage China to develop substitutes, divide global supply chains and impose costs on allied businesses. Controls can also be difficult to enforce when equipment, software and expertise move through multiple jurisdictions.

The controversy surrounding Nvidia’s leadership has further illustrated how commercial and security questions are colliding. NPR reported that Jensen Huang resigned as the company’s chief executive after revelations involving alleged workplace affairs and a potential Chinese spy. The episode does not establish wrongdoing by the company, but it underscores the scrutiny facing firms at the center of the semiconductor rivalry.

Public resistance is becoming a business risk

Data centers are also provoking a more local form of opposition. Their electricity demand can pressure grids, compete with households and intensify debates over who pays for new transmission infrastructure. Communities may welcome construction jobs and tax revenue while resisting higher utility costs, water consumption or land use.

That backlash is already pushing companies to look beyond traditional technology hubs. The World Politics Review reported that U.S. resistance to the AI boom is encouraging technology giants to build data centers overseas, while warning that similar scrutiny is likely to follow them abroad. Moving facilities may diversify risk, but it does not remove the underlying demands for power, water and public subsidies.

Governments face a difficult trade-off. Fast approvals could attract investment and strengthen national AI capacity. Strict requirements could protect consumers and local infrastructure, but also make projects more expensive or send them to countries with weaker standards. The debate is likely to intensify as electricity demand grows and voters connect AI expansion with household bills.

What comes next

The next stage will probably bring more conditional support rather than a simple embrace or rejection of AI. Governments are likely to tie permits and incentives to grid investment, environmental safeguards, cybersecurity and disclosure of major incidents. Companies, meanwhile, will need to demonstrate that their systems create measurable value rather than merely justify rising valuations.

Three tests will determine whether the boom becomes durable. First, productivity gains must spread beyond a small group of technology firms. Second, chip controls must protect security without fragmenting innovation into incompatible blocs. Third, data-center growth must be reconciled with affordable, reliable electricity.

The AI race is therefore becoming less about who can build the biggest model and more about who can build a political coalition around the infrastructure beneath it. The winners will be the countries and companies that make expansion credible to investors, useful to workers and acceptable to the public.

Sources