Unlock Editor's Digest free of charge
Roula Khalaf, editor of the FT, picks her favorite stories on this weekly newsletter.
AMD's chief executive believes the chipmaker is closing the performance gap with Nvidia's market-leading artificial intelligence processors by unveiling latest products aimed toward a market price lots of of billions of dollars.
On Thursday, the Silicon Valley-based group said its MI325X chip will begin shipping to customers within the fourth quarter of this yr, saying it offers “industry-leading” performance in comparison with Nvidia's current generation of H200 AI chips .
AMD's next-generation MI350 chip, designed to compete with Nvidia's latest Blackwell system, is anticipated to ship within the second half of 2025.
The U.S. chipmaker returned from bankruptcy a decade ago when Lisa Su took over as CEO to emerge as the largest challenger to Nvidia's grip on the infrastructure that powers generative AI.
Su said her goal is for the corporate to grow to be an “end-to-end AI leader” in the subsequent 10 years. “You should be extremely ambitious,” Su told the Financial Times. “This is the start, not the tip, of the AI ​​race.”
Nvidia customers are expected to start implementing Blackwell in the present quarter. Microsoft announced this week that it had grow to be the primary cloud provider to supply its customers the newest GB200 chips.
While the so-called “hyperscalers” – Microsoft, Google and Amazon – are also constructing their very own AI chips, AMD has grow to be Nvidia's biggest competitor within the race to supply off-the-shelf AI chips.
However, it stays a distant second. AMD's forecast AI chip sales of $4.5 billion for 2024 are small in comparison with the $26.3 billion in data center AI chip sales that Nvidia posted within the quarter ending July alone .
However, Su is confident that demand will only increase in the approaching years. The company has predicted that the full addressable marketplace for AI chips will reach $400 billion by 2027.
“When we began, that was considered a very big number,” Su said. “And I feel persons are moving toward our large numbers due to the huge demand for AI infrastructure.”
Chips are only a part of the infrastructure required to construct cutting-edge AI systems. AMD on Thursday also announced latest networking technologies and upgrades to its ROCm software toolkit, all aimed toward deploying AI infrastructure quickly and at scale.
“One of the things we’re really constructing is the end-to-end infrastructure for the info center,” Su said. “People need a large cluster (of chips in a server) so you may train the most important language models.”
Su, who has a doctorate in electrical engineering from the Massachusetts Institute of Technology, worked at Texas Instruments, IBM and Freescale Semiconductor before joining AMD in 2012 as senior vice chairman.
When she took over as CEO in 2014, AMD shares were stagnating at around $4. Some analysts predicted the corporate could be bankrupt in a couple of years because it struggled to compete with Intel.
Today, AMD has secured a big share of the server chip market and overtaken Intel in AI as the corporate diversifies from its traditional PC business. AMD's shares closed at $171 on Wednesday before the announcement, giving the corporate a market capitalization of about $275 billion – nearly thrice that of Intel.
Su sees AI as the important thing driver of AMD's next era of growth and is attempting to attract the identical customers as Nvidia.
“People are really open to trying different architectures and finding what works best for his or her workload,” Su said. So far, each Microsoft and Meta have adopted AMD's current generation of MI300 AI graphics processing units (GPUs).
Amazon, which is already a customer for AMD's server CPUs, is more likely to follow suit, Su said: “It's a point-in-time conversation.”
AMD's approach is analogous to what Nvidia is doing with Blackwell, where the corporate goals to sell not individual chips but entire server racks consisting of multiple chips combined with Nvidia's own networking equipment.
To catch up, AMD has pursued an aggressive investment and acquisition strategy, including the recent announcement of its $4.9 billion acquisition of ZT Systems, an organization that makes servers for the small group of AI hyperscalers.
Regarding possible regulatory reviews of the deal, Su said that “our current expectation is the US-EU (checks), and there are a couple of other jurisdictions as well, but we’re currently not exceeding the thresholds for China.”
