Jensen Huang, founder, chairman and CEO of Nvidia, speaks about the future of artificial intelligence and its impact on energy consumption and production at the Bipitarian Policy Center in Washington, D.C., September 27, 2024.
Somodevilla chip | Getty Images
Nvidia revealed on Monday new chips for desktop and laptop computers that use the same Blackwell architecture that powers the company's fastest AI processors for servers and data centers.
The chips, called GeForce RTX 50 series, will be pre-installed in computers priced from about $550 to $2,000, the company said. Shipments of laptops with chips will begin in March.
Nvidia unveiled the processors at CES in Las Vegas, where CEO Jensen Huang delivered a keynote speech on Monday.
“Can you imagine having this amazing graphics card, Blackwell? I'll shrink it and put it there,” Huang said, holding the laptop.
Nvidia, which has he rose with a market capitalization of over $3.5 trillion thanks to sales of AI chips to giant cloud service providers and other technology companies, until the last few years it was known for selling graphics processing units (GPUs) to power video games. Nvidia's first chip, released in late 1999, was designed to quickly draw triangles and polygons for 3D games.
“Obviously back then we were a gaming company and these GPUs were built to speed up games,” Justin Walker, senior director of product at Nvidia, said during the press call.
Wall Street is less enthused about Nvidia's gaming business these days, given the explosion of artificial intelligence and the ever-increasing demand for more processing power. In quarter that ended in October, Nvidia's game sales accounted for less than 10% of total revenue compared to 88% of data center chip sales.
Nvidia holds the vast majority of the data center AI GPU market, outpacing rivals Advanced micro devices AND Intel.
But CES, formerly known as the Consumer Electronics Show, focuses solely on consumer products, and the new chips unveiled on Monday are aimed primarily at gaming.
Nvidia says the RTX 50 series chips will support a feature called DLSS 4, which uses artificial intelligence to increase frame rates in games. They can also display characters' faces in greater detail and generally provide users with better graphics and higher resolution.
Nvidia's gaming business is growing, with revenue up 15% year-over-year in the most recent quarter. However, data center sales have at least doubled for six quarters in a row, most recently reaching $30 billion.
Nvidia says the technical improvements made to its massive AI division will be applied to gaming graphics cards.
“Even though we are now an AI company as well as a gaming company, our gaming side still benefits tremendously from the fact that we are an AI company,” Walker said.
The Blackwell GPU architecture and core design that powers the 50-series chips debuted in the company's AI accelerators, which have been announced in March and started shipping late last year. Nvidia says they are designed and optimized to support the neural networks used by ChatGPT and OpenAI Google Twins.
The new desktop and laptop chips will be available in several different configurations. The company says the most expensive and powerful of the chips, the RTX 5090, will sell individually for $1,999 and is twice as fast as its predecessor, the RTX 4090. Nvidia says it has 92 billion transistors compared to 208 billion transistors in the company's processor. B200 for servers.
Nvidia says the chips will be optimized to run AI models and computer graphics, not just to run the latest games. The chips will be powerful enough that some game developers will be able to integrate generative artificial intelligence into their characters in games like “PUBG: Battlegrounds.”
The new processors will also be powerful enough to support large language models and image generation models from companies such as: MetaMistral and Stability AI, Nvidia said.
TO WATCH: Bernstein's Rasgon says investors need Nvidia's exposure
