Nvidia (NVDA 1.46%) CEO Jensen Huang delivered the keynote address at CES 2026, a leading annual trade show in Las Vegas for tech companies. It's an ideal venue for Huang because the CES conference is centered around connecting the tech ecosystem and highlighting future innovations.
And few companies have been as innovative as Nvidia in the last few years. The company's vaunted graphics processing units (GPUs) power, train, and run many of the artificial intelligence (AI)-powered programs offered today. AI has advanced to the point that it is now helping developers create new, more powerful AI programs; essentially, developers are building AI programs on the back of AI applications.
That's why Nvidia stock has shot up by 1,000% in the last three years, making Nvidia the biggest company in the world by market cap. And that's why its important to pay attention to Huang's bold statements about the future of AI.
"The race is on for AI," Huang said in his keynote, adding that, "Everybody's trying to get to the next level. Everybody's trying to get to the next frontier. And every time they get to the next frontier, the last-generation AI tokens, the cost starts to decline about a factor of 10x every year."
He continued:
The 10x decline every year is actually telling you something different. It's saying that the race is so intense. Everybody's trying to get to the next level, and somebody is getting to the next level. And so therefore, all of it is a computing problem. The faster you compute, the sooner you can get to the next level of the next frontier. All of these things are simultaneously happening at the same time.
What does that mean for Nvidia stock? It means AI development is far from a bubble. In fact, it's speeding up and getting cheaper at the same time, which opens the AI door to even more developers.
Cheaper AI means more AI, more computing -- and more demand for Nvidia's line of chips.
Let's break down Huang's commentary
Huang devoted some time in his keynote to discussing the rapid development of more powerful AI -- specifically, the OpenAI model o1 released in September 2024, a large language model (LLM) capable of complex reasoning. But it also has a much greater need for computing power. "o1's introduction was an inflection point for AI. Instead of a one-shot answer, inference is now a thinking process," Huang said.
That's why AI developers require more advanced AI chips. But what happens to the older generations of Nvidia chips (Ampere, Hopper, and Blackwell) when Nvidia rolls up its next-generation Rubin chips this year?
As Huang says, the price begins to decline by a factor of 10 annually. Consumers see the same scene play out with their smartphones; when Apple rolls out its newest, more powerful models, the price of Apple's older iPhones immediately drops. Consumers who want an iPhone but aren't willing to pay top prices for the newest model are content to buy the older model that's slower and less powerful but still does the job they need.
Image source: Getty Images.
We're going to see the same scenario with data center GPUs and AI infrastructure. As the cost of older chips drops, more developers and companies that may be priced out of the most powerful models will be interested in buying and building their own AI programs.
So not only is there a race for the biggest and most powerful chips, but there will be a sustainable market for less powerful chips that create a second-tier doorway to AI programs. And Nvidia will be ideally positioned to serve those markets.
The outlook for Nvidia stock
Nvidia's performance has been impressive for the last several quarters, and I'm expecting that to continue. In the third quarter of fiscal 2026 (ended Oct. 26, 2025), Nvidia's revenue was $57 billion, up 62% from a year ago. Data center revenue was $51.2 billion, up 66% from last year.
Huang said at the time of that report that the company was sold out of cloud GPUs, with Blackwell sales particularly strong. The company also said it had $500 billion in AI chip bookings through calendar 2026, with $150 billion of that already fulfilled.
NASDAQ: NVDA
Key Data Points
And now CFO Colette Kress has said the $500 billion figure has already gotten larger, as Nvidia is getting full-year orders for its Rubin chips.
In addition, Nvidia is seeing growing demand for its chips from open-source AI models, such as Meta Platforms' Llama, which are generating one in four AI tokens, meaning that they are handling about 25% of AI workload volume.
Nvidia is on target to produce $213 billion of revenue in the 2026 fiscal year and $321 billion in the 2027 fiscal year, according to analyst estimates -- roughly a 50% gain. Earnings per share are forecast to be $4.69 for this year and $7.60 in 2027.
Clearly, Nvidia's growth is going to continue, and it's in an ideal position to benefit as companies continue to make more sophisticated AI programs. But don't overlook how Huang describes the falling costs of older chips and their elasticity. As AI continues to grow, those older chips will be in plenty of demand as well.
