Artificial intelligence (AI) development is entering a new phase driven by an unprecedented global expansion of data center infrastructure designed to deliver vast increases in computing power. Across regions including the American Midwest and the Persian Gulf, hundreds of large-scale data centers currently under construction are expected to come online over the next several years, fueling breakthroughs in AI capabilities and applications.

Today’s AI systems rely heavily on specialized chips—currently numbering around 20 million globally—that power training and deployment processes. According to research firm Epoch AI, this total is projected to roughly double every nine months, reaching about 200 million AI chips by the end of 2028, a tenfold increase from current levels. Industry insiders liken this expansion to transformative historical initiatives such as the railroad construction era, the New Deal programs, and the Manhattan Project.

Rob Wachen, co-founder of the chip company Etched, characterized the current build-out as “the largest scale infrastructure build-out in the history of humanity.” At Amazon, which supports AI operations for firms including Anthropic and OpenAI, computing capacity has doubled since 2022 and is on track to double again next year, according to Peter DeSantis, who leads foundational AI projects there.

The scaling of computational resources is closely tied to AI’s expanding functionality. The principle, known as the Scaling Laws, posits that increased data and computing power enable AI systems to handle more complex tasks and assume broader roles previously requiring human expertise. This dynamic forms the basis for optimistic forecasts from AI leaders who argue that those controlling the largest computational resources will dominate future advancements and market share.

Predictions from executives such as Anthropic’s Dario Amodei and DeepMind’s Demis Hassabis envision AI enabling substantial automation of white-collar work and sparking industrial revolutions at unprecedented speed and scale. Analysts anticipate these developments will drive innovations in fields like drug discovery and robotics, while increasingly integrating AI into everyday personal and professional activities.

However, the rapid pace and scale of data center expansion have generated resistance. Environmental concerns dominate public debates, focusing on impacts such as elevated electricity consumption, water usage, and effects on local communities. In the United States, opposition to data centers has become a significant political issue ahead of upcoming elections, reflecting broader skepticism toward big tech’s growth.

Economists and investors caution that heavy spending on AI infrastructure may outpace near-term profits, raising the risk of a technology-driven investment bubble. Historical precedents such as railroad booms and the dot-com era suggest such cycles often precede corrections before broader benefits materialize. Philippe Aghion, a Nobel Laureate in economic science, compared AI to a fourth industrial revolution that carries inherent bubble risks.

The geopolitical landscape further complicates the AI infrastructure race. The United States maintains a commanding lead, hosting approximately 5,500 data centers and controlling roughly 80 percent of global AI computing power through firms like Amazon, Google, Microsoft, and Meta. China, the next largest competitor, is expanding its chip production and data center capacity rapidly but remains behind due to export controls and other technological barriers. Chinese companies such as Huawei, ByteDance, and Alibaba have committed substantial investment to narrow this gap, supported by government initiatives prioritizing AI infrastructure development.

Despite significant state support and growing capabilities from Chinese startups, AI experts say China’s infrastructure deficit limits widespread AI deployment there. Analysts anticipate the U.S. lead will persist for several more years, although China may begin closing the gap as domestic chip manufacturing scales.

Outside of these two powers, Europe and the Middle East are also investing in data centers, though challenges such as regulatory restrictions, land and electricity constraints, and regional instability hinder rapid expansion. Experts warn this dichotomy could deepen global inequalities in AI access and benefits.

As data center capacity grows, AI systems are expected to accelerate innovation cycles dramatically. Increasingly sophisticated AI “agents” are already performing complex multitask activities across industries, from coding to customer service. Google’s chief scientist, Jeff Dean, highlighted the potential for AI agents to autonomously conduct scientific research by generating and testing hypotheses, signaling a shift toward more automated AI development processes.

The rising AI capabilities also prompt concerns over labor market disruption. Economists like Erik Brynjolfsson project extensive shifts in employment landscapes, with substantial job displacement alongside the creation of new roles that may require very different skills.

Looking ahead, AI labs are pursuing recursive self-improvement techniques that could enable models to iteratively enhance themselves without human intervention, a development Google is actively exploring. Such advances depend heavily on continuous expansion of AI computing infrastructure, underscoring the ongoing “AI arms race” among leading technology companies worldwide.

Investment in AI infrastructure shows no signs of slowing; forecasts predict global spending on this segment will exceed $1 trillion annually by 2029, surpassing the GDP of some mid-sized economies. As technology capabilities and infrastructure build-out accelerate, AI is poised to reshape industries and societies on an unprecedented scale.