Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Without Nvidia’s high-powered chips, there would be no generative artificial intelligence boom. Demand already outstrips supply. The company’s lead over rivals is about to be extended with the release of a new specialist AI chip. If analyst forecasts are correct, annual revenues will more than triple over the next three years to top $98bn.
Chief executive Jensen Huang’s bet on AI grows more bombastic by the day. It will, he says, be “bigger than the internet”. On Tuesday he declared that it represented the biggest expansion of total addressable market for software and hardware. However big the projections are, Huang wants them to be bigger.
If he is right then Nvidia’s lack of competition and wealthy enterprise customers put it in clover. Tech companies with fat cash piles are investing in AI in a bid to improve services and raise productivity. With gross margins at 74 per cent, Nvidia’s own cash pile has grown by $5bn in the fiscal year to date to $18bn. Along with chips, it is selling more products and services devoted to expanding use, including a new kind of Ethernet switch. Its market cap hit a record $1.25tn this week. It has come to represent the market’s collective hopes and dreams for AI.
What might stall future gains? Developers who worry about the speed of AI commercialisation could try to slow the sector down. The start-up market may be cooling, too. CB Insights points out that four AI start-ups reached valuations of $1bn or more in the last quarter, down from seven in the previous quarter.
But the company’s biggest worries centre on China and competitors. China accounted for 21.5 per cent of sales in the past nine months. That is falling. US attempts to ensure chip security have led to restrictions on the country’s access to advanced chips.
This clears the field for Chinese rivals to encroach on Nvidia’s market. At home, cloud customers are busy trying to counter their reliance on the company, too. Along with Intel and AMD, they are developing bespoke chips.
Nvidia’s answer is a new chipset for release next year. The H200 comes with 141 gigabytes of memory and can generate results twice as quickly as the company’s widely used H100 chip. Once again, Nvidia will set the industry standard.
If you are a subscriber and would like to receive alerts when Lex articles are published, just click the button “Add to myFT”, which appears at the top of this page above the headline.
