Nvidia and Trump Reject Tech Leaders’ Call to Slow AI Development
During a live appearance at the All-In Summit on Monday, Nvidia Chief Executive Officer Jensen Huang received an unexpected phone call from U.S. President Donald Trump. Their public exchange served as a sharp rebuke to a growing movement among top technology executives to slow the development of advanced artificial intelligence.
Trump dismissed recent warnings about AI safety and the environmental impact of data centers as a “hoax,” arguing that the United States cannot afford to cede ground to China. Huang echoed the sentiment, telling the president that Nvidia would not let a slowdown happen.
The conversation brought a simmering industry debate into the political spotlight. Over the past week, several prominent leaders running the companies that build AI models have publicly questioned whether the rapid pace of development poses unmanageable risks. But the hardware industry, led by Nvidia, remains focused on maintaining speed and expanding computing infrastructure.
“The robots will not be taking over,” Trump told Huang on speakerphone. “The AI will not be taking over the rest of the world. The whole thing is a hoax”.
Trump framed the push for safety regulations in political terms, suggesting that critics were playing into the hands of foreign adversaries. He compared warnings about artificial intelligence risks and data center construction to climate change concerns, describing both as campaigns by people trying to hinder American progress.
The president emphasized the economic necessity of computing infrastructure. He called data centers “the oil of the next 20, 25 years” and predicted they would be more significant to the global economy than the internet.
Huang responded affirmatively to the president’s push to keep moving fast. “We’re going to make sure that everybody wins in the AI race in America,” Huang said. “Every industry, every company, every state, every people”. When Trump reiterated that he would not let a slowdown happen, Huang agreed, stating, “You’re right. We’re not going to let that happen, sir”.
The Safety Push from Software Developers
The phone call arrived after a turbulent week for the AI industry, sparked largely by the very companies buying Nvidia’s hardware. Days earlier, Anthropic CEO Dario Amodei published an essay outlining the risks of rapidly advancing frontier models.
Amodei proposed a three-part plan to pace development and announced that Anthropic would grant third-party evaluators permanent, employee-level access to its systems. These outside reviewers would verify safety compliance, report incidents, and assess how models align with human instructions during the training process.
Amodei’s call for a coordinated slowdown gained public backing from other prominent figures, including OpenAI CEO Sam Altman, Microsoft CEO Satya Nadella, and xAI founder Elon Musk. The push for caution escalated further when an Anthropic researcher resigned, warning publicly that the current trajectory of AI development could have catastrophic consequences for humanity.
The growing consensus among software executives is that building the next generation of highly autonomous models requires a different approach than the fast-paced shipping cycles of standard software.
Ahmed Gamal Elsharkawy, an AI expert speaking at the BRICS Summit this week, articulated this shift. He noted that while businesses should continue adopting current technology, the creation of autonomous systems demands caution.
“I think the real question is not whether we should slow down AI. The question is whether our ability to govern and control AI is moving as fast as the technology itself,” Elsharkawy said. He added that when executives like Amodei and Altman raise concerns about the speed of progress, the public and governments should take those warnings seriously.
The Technical Debate: Scaling Laws and Test-Time Compute
Before the safety debate dominated the conversation, researchers were already arguing over whether AI development was naturally slowing down. Some academics and engineers suggested that the industry was hitting a plateau in “scaling laws”—the foundational principle that simply adding more computing power and data during a model’s training phase will predictably improve its capabilities.
Huang has consistently rejected the idea that AI progress is stalling due to physical or technical limitations. Instead, he argues that the industry is simply expanding how it applies computing power. He recently outlined three active phases where scaling laws still apply: pre-training, post-training, and test-time compute.
Test-time compute refers to the processing power a model uses while generating an answer, rather than just during its initial training phase. By allowing models to process longer before responding—often through internal reasoning steps—companies can continue to achieve performance gains without solely relying on increasingly massive initial training runs.
Because of these new methods, Huang predicts that AI scaling laws will drive demand for computing power up by a million times over the next decade.
This technical reality explains why Nvidia remains unfazed by calls for a slowdown. Even if companies temporarily pause the creation of larger frontier models to satisfy safety concerns, the shift toward test-time compute means those same companies will require vast amounts of hardware simply to run their existing models at scale.
Speaking at the summit, Huang argued that the greatest risk associated with AI is not losing control, but rather stagnation.
The Geopolitical Angle
The push to maintain speed aligns closely with the broader strategy of the U.S. government. At a G20 meeting in North Carolina last week, the United States successfully pushed through an international agreement among major economies aimed at accelerating AI adoption.
The resulting 1,200-word communiqué heavily emphasized the positive economic impacts of artificial intelligence and used the word “accelerate” seven times. It lacked language addressing the negative risks associated with the technology. The document stressed that any evaluation of AI regulation should consider the opportunity cost of delayed or forgone adoption, urging governments to rely on existing regulations rather than drafting new ones.
Huang commended this approach, urging the G20 gathering not to regulate “hypothetical, theoretical harm”.
This geopolitical framing puts pressure on companies proposing voluntary slowdowns. As long as national security is tied to AI supremacy, executives arguing for safety pauses face accusations of jeopardizing American competitiveness against geopolitical rivals. Trump made this point explicitly during his phone call with Huang, insisting that the U.S. must defeat China in the race for innovation.
Market Impact and What Happens Next
The divide between AI developers calling for caution and the infrastructure providers pushing for speed highlights a rare fracture in the technology industry. Companies like Anthropic and OpenAI rely entirely on Nvidia’s specialized graphics processing units to train and run their models. If the software developers choose to deliberately pace their research, it could theoretically impact future capital expenditures.
Yet Trump’s explicit support for data center expansion signals that the federal government plans to foster an environment conducive to massive infrastructure investment. He rejected criticisms from local communities concerned about the electricity and water demands of new data facilities, comparing the backlash to failed environmental campaigns.
For Nvidia, whose revenue is directly tied to the construction of these facilities, the political backing provides a strong counterweight to the safety concerns raised by its biggest customers.
The industry now faces a test of how voluntary safety commitments hold up against political pressure and market demands. Anthropic has taken the first step by opening its systems to outside evaluators. Observers will watch to see if companies like OpenAI and Microsoft follow through on their rhetoric with concrete pauses in their training runs, or if the competitive pressure to utilize Nvidia’s latest hardware forces them to maintain their current pace.
With government support firmly behind maximizing computing power, the hardware side of the AI race will continue unabated. Nvidia shows no signs of altering its roadmap, and as long as the U.S. government views AI development as a proxy for national security, calls for a slowdown will likely face steep resistance from Washington.
