Nvidia raked in more than $19 billion in net income during the last quarter, the company reported on Wednesday, but that did little to assure investors that its rapid growth would continue. On its earnings call, analysts prodded CEO Jensen Huang about how Nvidia would fare if tech companies start using new methods to improve their AI models.

Test-time scaling" was the method that underpins OpenAI's o1 model and came up quite a lot. In a way, it's a bit of a fuzzy idea-things will get better answers if you give them more time and computing power to "think" through questions. Specifically, it adds more compute to the AI inference phase, which is everything that happens after a user hits enter on their prompt.

Nvidia's CEO was asked if he was seeing AI model developers shift over to these new methods and how Nvidia's older chips would work for AI inference.

Huang suggested that o1, and test-time scaling more generally, could become even more important for Nvidia's future business, considering it "one of the most exciting developments," and "a new scaling law." Huang did his best to reassure investors that Nvidia is well-positioned for this change.

The comments from Nvidia's CEO matched what Microsoft CEO Satya Nadella had said onstage at a Microsoft event on Tuesday: o1 represents a new approach the AI industry is taking with its models.

It's a big deal for the chip industry because it puts a much bigger emphasis on AI inference. Nvidia chips are the standard for training AI models, but there's a broad set of well-funded startups creating lightning-fast AI inference chips, such as Groq and Cerebras. This could make it a relatively more competitive space for Nvidia to operate in.

Contrary to recent reports that pace of improvement in generative models has started to slow, Huang told the analysts that AI model developers are still improving the models by adding more compute and data during the pretraining phase.

Anthropic CEO Dario Amodei also said Wednesday during an onstage interview at the Cerebral Valley summit in San Francisco that he is not seeing a slowdown in model development.

"The principle of scaling of pretraining of foundation models remains intact, and it continues to scale," said Huang on Wednesday. "As you know, this is an empirical law, not a fundamental physical law, but the evidence is that it continues to scale. What we're learning, however is that it's not enough."

That's certainly what Nvidia investors were hoping to hear, since the chipmaker's stock has increased by more than 180% in 2024 selling the AI chips that OpenAI, Google, and Meta train their models on. However, Andreessen Horowitz partners and several other AI executives had previously said that these methods were already starting to demonstrate diminishing returns.

Huang said that most of the computing workloads Nvidia has these days involve the pretraining of AI models — not inference — but he wrote that off more to where the AI world is. He said one day there simply will be enough people running AI models, meaning more AI inference will happen. Huang mentioned, "Nvidia is the world's largest inference platform today. It is quite a long way ahead of any of the startups because of the company's scale and credibility.".

Our hopes and dreams are that someday, the world does a ton of inference, and that's when AI has really succeeded," said Huang. "Everybody knows that if they innovate on top of CUDA and Nvidia's architecture, they can innovate more quickly, and they know that everything should work."