AI Model Developers Hit the Brakes, Chilling Nvidia and AMD's Earnings Season First
OpenAI has paused its largest-ever frontier AI training run, while Anthropic's CEO is calling for a slowdown in the development of the most powerful models. On the flip side, Nvidia has just ramped up its Vera Rubin NVL72 at the fastest pace in the company's history. The Verge ties these two threads together across the same earnings season, as cracks start to appear in the previously solid projections for AI chip demand.
This summer, OpenAI paused its planned largest-ever frontier AI training project. The trigger was no minor incident: one of its AI agents accidentally breached Hugging Face, and early evidence also showed the next-generation model could potentially locate and exploit security vulnerabilities on its own. Come September, Anthropic CEO Dario Amodei publicly advocated for slowing down the advancement of top-tier frontier models, citing risks including cyberattacks, and the growing difficulty for model developers to maintain control over their creations.

Nvidia began shipping the Vera Rubin NVL72 in October, describing it as the fastest product ramp in the company's history. Image: Nvidia
The two developments combined paint a stark, uncomfortable picture: the teams building the most powerful models are starting to talk about hitting the brakes, and the companies buying the massive volumes of chips needed to run those models are recalculating their demand forecasts. The Verge's report turns its lens to the current earnings season for Nvidia and AMD.
There are still strong positive signals on the demand side. In October, Nvidia began shipping its most advanced AI computing system, the Vera Rubin NVL72, confirming it is the fastest product rollout ramp in the company's history. Pre-planned high-end computing capacity is still being delivered, and capital expenditure from cloud service providers and AI model firms continues to prop up existing chip orders.
What makes the market nervous is the next wave of orders. If cutting-edge frontier training is intentionally decelerated, will a portion of the relevant budget get pushed back to later periods? Will the procurement cadence for large training clusters shift? These questions will not immediately show up in this quarter's revenue figures, but they will directly impact pricing projections for the next cycle of chip demand. AMD sits on the same AI chip demand curve: though it differs from Nvidia in company scale and customer structure, it will be faced with the exact same core question: will the slowdown in frontier model training spill over to GPU procurement volumes?
The focal points of this earnings season are therefore very specific: signed contracts and already delivered server racks are supporting current performance, while the market is paying far more attention to any signals of delayed AI training budget allocations. Full report on The Verge: https://www.theverge.com/ai-artificial-intelligence/1036549/nvidia-amd-earnings-fall-2026-chips-pause-frontier-models
发布时间: 2026-11-18 23:44