Sam Altman Sounds Alarm on China’s AI Surge as US Grid Struggles and Investment Bubble Looms

OpenAI CEO Sam Altman warns U.S. risks losing AI lead to China due to energy and infrastructure limits. China’s abundant electricity boosts its AI growth despite chip export controls.

Altman Sounds Alarm: U.S. AI Lead at Risk as China Advances Amid Power Constraints

OpenAI CEO Sam Altman has raised concerns about the United States losing its edge in artificial intelligence to China. Speaking in San Francisco, Altman emphasized that U.S. chip export controls alone won’t be enough to slow China’s rapid AI development. “I’m worried about China,” he said bluntly, pointing out that the current export restrictions likely won’t succeed.

Altman also linked China’s progress in open-source AI models to OpenAI’s recent decision to release weights for its GPT open-source models. He noted that without this move, the global AI landscape might have been dominated by Chinese open-source systems like DeepSeek and Kimi. This strategy is seen as a direct challenge to China’s growing lead in open AI technologies.

Power Availability, Not Just Hardware, Is the Key Advantage

Altman’s warnings coincide with reports highlighting China’s energy advantage. Decades of investment in electricity generation and transmission have given China a vast and reliable power supply. AI experts visiting China observed that energy availability there is taken for granted, with the country adding more annual electricity demand than Germany’s total consumption and maintaining reserve margins of 80% to 100%, backed by idle coal plants.

Meanwhile, the U.S. faces significant infrastructure bottlenecks. A study from Lawrence Berkeley National Laboratory (LBNL), supported by the Department of Energy, projects that U.S. data-center electricity demand could double or triple by 2028. Data centers currently consume about 4.4% of U.S. electricity, a figure expected to rise to between 6.7% and 12%. This surge is already straining interconnection queues and transmission systems, slowing down where and how quickly new AI data centers can be built.

LBNL’s recent “Queued Up” analysis reveals that projects in 2023 took nearly five years from interconnection request to commercial operation, a sharp increase from three years in 2015. Completion rates remain low nationwide, highlighting structural challenges rather than temporary issues.

How Companies Are Coping with Grid Limitations

To get around the grid constraints, some tech giants are locking in dedicated power sources, often from nuclear plants. Microsoft signed a 20-year deal with Constellation to help restart the Three Mile Island nuclear unit by 2027 to support data-center demand. Google and Kairos Power announced plans for a 50-MW advanced reactor in Tennessee targeting 2030, while Amazon Web Services secured up to 1,920 MW from the Susquehanna nuclear plant in Pennsylvania.

Altman did not downplay the current financial hype around AI, comparing it to the late-’90s dot-com bubble where enthusiasm outpaced reality. Still, he stressed that AI remains “the most important thing to happen in a very long time.” OpenAI plans to invest trillions in data-center construction in the near future, a scale that clashes directly with U.S. energy limitations and China’s earlier lead in available electricity.

The Growing Investment Stakes

Independent studies estimate global data-center investments will reach around $6.7 trillion by 2030, with approximately 70% dedicated to AI workloads. If U.S. permitting and transmission delays persist, these investments are likely to flow to regions with faster grid connectivity. This dynamic reinforces Altman’s warning about underestimating China’s advantage.

China’s abundant, low-cost electricity allows it to expand AI training and inference with fewer obstacles, even under chip export restrictions. This reality puts pressure on U.S. policymakers to accelerate grid modernization urgently.

Final Takeaway

While chips remain critical, the most pressing bottleneck for AI advancement in 2025 may be access to reliable electricity. Altman’s message is clear: the AI race will favor those with ready power infrastructure and streamlined regulations.