Nvidia's impressive $19 billion net income in the last quarter failed to alleviate investor concerns about the company's future growth, particularly in the AI inference space. During the earnings call, CEO Jensen Huang was probed about the potential impact of new methods, such as "test-time scaling," on Nvidia's business.

The method, popularized by OpenAI's o1 model, involves adding more compute to the AI inference phase to improve model performance. Huang acknowledged that this approach could play a significant role in Nvidia's future, calling it "one of the most exciting developments" and "a new scaling law." However, this shift could also increase competition from startups like Groq and Cerebras, which are developing lightning-fast AI inference chips.

Despite recent reports of slowing generative model improvements, Huang and Anthropic CEO Dario Amodei expressed confidence in the continued scaling of foundation model pretraining. Nvidia's dominance in AI training chips has driven its stock up over 180% in 2024, but the company's reliance on pretraining workloads may need to adapt to the growing importance of AI inference.

Huang emphasized Nvidia's scale and reliability advantages, positioning the company as the largest inference platform in the world. As the AI industry continues to evolve, Nvidia's ability to adapt to new methods and maintain its market lead will be crucial to its future success.