Nvidia's crisis and the politics of an AI pause
Add Axios as your preferred source to
see more of our stories on Google.
Photo illustration: Aïda Amer/Axios. Photos: Kevin Carter and Jose Sarmento Matos/Bloomberg via Getty Images
A political backlash against AI and data centers — now compounded by a crisis facing the technology's most influential champion — could help force a slowdown that industry leaders have struggled to achieve voluntarily.
Why it matters: Tapping the brakes on AI development — which frontier AI developers claim to want but have so far failed to do given the risk of offering an advantage to a competitor — could give society more time to prepare for economic upheaval and the risks of superintelligent systems.
- A wave of Democratic victories and data center moratoriums are now joined by a potentially weakened Nvidia CEO Jensen Huang, who admitted to having a relationship with an alleged Chinese spy.
- Huang, who has led the industry in barreling through past resistance to AI acceleration, may be less able to do so while under a microscope.
- While both factors could stymie data center expansion in the U.S., it's far from clear that would slow AI's advance.
Between the lines: Huang has been a powerful counterweight to calls for AI restraint, dismissing fears of an AI bubble or purported catastrophic risks to humanity from the technology.
- An investigation — by the board, an ascendant Democratic congress or federal authorities — could curb his ability to make the case for continued expansion, leaving an opening for those calling for more restrictions.
- Nvidia says Huang will remain CEO and the news won't slow the company, but the potential impact on his political influence is hard to measure.
- The potential impact extends beyond that as well. Nvidia is both the dominant supplier of AI chips and an investor in companies building AI infrastructure. A crisis that raises questions about its leadership, business practices or financial relationships could ripple through the industry's expansion plans.
The intrigue: Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and Elon Musk have all warned about the potentially catastrophic risks of advanced AI, with all expressing, at times, the need to slow or pace the development of advanced AI.
- But the competitive race has made it difficult for any one company to hit the brakes without risking falling behind rivals or China in a high-stakes race for supremacy.
What they're saying: Anthony Aguirre, CEO of the Future of Life Institute, says a slowdown in data center construction would be welcome, even if it wouldn't accomplish nearly as much as more direct restrictions on frontier AI development.
- "I'll take whatever I can get," Aguirre told Axios in an interview. Aguirre and a number of other researchers have long seen the development of increasingly powerful AI as an existential threat.
- But he is skeptical that slowing data center construction alone would buy society an additional year or two, given the vast computing resources companies already possess and the growing ability of AI systems to help improve themselves.
- Aguirre said the bigger opportunity may be political: Resistance to data centers could help generate momentum for legislation and regulation that more directly limits the development of potentially dangerous AI systems.
The big picture: Aguirre argues that the goal shouldn't simply be to delay superintelligence by a year or two, but to reconsider whether humanity should build it at all.
- He favors a "pause and pivot" toward AI systems designed to augment human abilities rather than replace them.
- Other researchers and executives favor continued development with stronger safeguards, reflecting a broader debate over how much restraint is necessary and what form it should take.
Yes, but: Slowing data center construction in the U.S. doesn't necessarily mean slowing AI progress globally.
- Meta and other companies are exploring data center construction outside the U.S. Google announced a plan this week to invest $15 billion over two years in Malaysia.
- And Chinese AI developers wouldn't face the same U.S. restrictions, potentially allowing them to narrow the gap with American competitors.
The other side: The threat from China isn't as straightforward as it appears.
- Chinese AI companies benefit in part from distillation of leading American models and technical knowledge developed in the U.S., Aguirre said. Slowing the American frontier could therefore have some effect on China's pace as well.
- A U.S. decision to slow frontier development could open the door to direct conversations with China about mutually pacing the technology, rather than treating superintelligence as an unavoidable arms race.
The bottom line: The AI industry is hoping to overcome political resistance and keep building. Advocates of slowing frontier AI see an opportunity to turn that resistance into something bigger: A chance to reconsider not just how fast AI should advance, but how far it should go.
