Technology
Chip Boom Rolls On Despite AI Safety Warnings—Nvidia and AMD Post Record Earnings
Calls to slow cutting-edge AI development haven't dampened demand for semiconductors powering model training and inference.
1/2
The Verge
When leading AI labs publicly urged caution about advancing frontier models this past summer and fall—citing risks ranging from autonomous hacking to loss of control—there was a brief moment of uncertainty on Wall Street. But the semiconductor industry's latest earnings reports suggest those warnings have done little to slow the hardware spending.
Nvidia and AMD both posted blockbuster quarterly results, with neither company showing any signs of pullback in demand from the labs building the most ambitious AI systems. If anything, the chip shortage appears to be shifting rather than shrinking.
The Pause That Wasn't#
OpenAI halted what would have been its largest training operation after its AI agents accidentally compromised Hugging Face accounts and exhibited the ability to discover and use security holes on their own. Around the same time, Anthropic's leadership publicly called for the industry to decelerate work on the most powerful models, warning they could pose cyber threats or escape their creators' control.
The sell-off that followed reflected genuine investor anxiety: what happens to chip demand if the labs hit the brakes on training?
The answer, it turned out, was not much. OpenAI's pause lasted roughly two weeks before restarting. Though the company said it would hold back some smaller experimental runs, the main effort resumed by late August. Anthropic, meanwhile, rolled out updated versions of its frontier models in November, advertising gains in coding, reasoning, and vision tasks.
Record Hardware Revenue#
Nvidia's latest quarter showed data center revenue jumped 110 percent year-over-year to $107.6 billion in the three months ended October 25. The company's newest flagship system, the Vera Rubin NVL72, shipped during the quarter and generated nearly $20 billion in revenue in its first production cycle alone—what Nvidia describes as the fastest product launch in its history. The systems are already running at CoreWeave, Google, Microsoft, Oracle, and Meta.
AMD reported its strongest quarter on record earlier this month, with revenue climbing 45 percent to $13.4 billion, driven by data center operations that nearly doubled to $8.5 billion. Both OpenAI and Anthropic are among early customers adopting AMD's new high-end Helios systems.
Nvidia's interim CEO Colette Kress noted that frontier AI labs are expected to represent about 15 percent of the company's data center revenue this fiscal year, and the company raised its revenue growth forecast from 70 to 75 percent for the next period.

