and Google. All three use chips from Nvidia and AMD in their data centers to power their booming cloud-computing businesses. Meta’s ambitions will thus expand an already lucrative market for the two chip makers.
AMD said last week that Meta is a new customer for its Epyc server processors. And
estimates that about $5 billion to $9 billion of Meta’s additional capital spending will go toward artificial intelligence, “for which Nvidia is likely to be the largest beneficiary.”
Nvidia, which reports fiscal third-quarter results on Wednesday afternoon, already has a data-center business generating about $8.2 billion in revenue annually. That is expected to surpass the $10 billion mark by the end of the company’s fiscal year in January, representing a fivefold increase in four years.
Success in data centers has helped remake the fortunes of a company once known primarily for personal-computer game chips. With a market capitalization of more than $750 billion, Nvidia is now the most valuable company in the semiconductor space and the seventh-most-valued on the S&P.
—which has more than 10 times the annual revenue—is valued around $637 billion.
AMD is smaller, but its run has been no less striking. Trailing 12-month revenue as of the quarter ended Sept. 25 was nearly $14.9 billion—more than double the level of two years ago and more than tripling over five years. The company has chipped away at Intel’s longtime lock on the market for central processor, or CPU, chips for servers.
Mercury Research estimates AMD had a 10.2% share of the server CPU market in the third quarter, up nearly 4 percentage points from the same period last year. AMD’s market value has surged ninefold over the past three years and—at its current level of just under $180 billion—is just 14% below Intel’s, which generates more than five times as much revenue.
The two have cut relatively separate paths to their current positions, with Nvidia focusing on graphics processors, or GPUs, used to accelerate artificial-intelligence capabilities in data centers. But they will likely end up competing more directly in the months and years ahead. AMD announced the second generation of its own data-center GPU chip last week, which it can integrate with its server CPU chips to optimize performance for the two. Nvidia likewise has ambitions beyond the GPU slot, with the company unveiling plans earlier this year to field a server CPU chip by 2023.
Both chip makers have plenty of addressable market ahead, even if Facebook’s metaverse dreams come to naught. IDC projects world-wide spending on cloud-computing services and related components will average nearly 17% annual growth to hit $1.3 trillion by 2025.
But Nvidia and AMD are now carrying relatively pristine valuations for semiconductor companies at 68 times and 47 times forward earnings, respectively. At those levels, neither one can afford to leave any chip unturned.