AMD Unveils New Products in Pursuit of $2 Trillion Market

The Advanced Micro Devices headquarters in Santa Clara, California.

Advanced Micro Devices Inc. announced a raft of new products for data centers that it said will outperform those from rival Nvidia Corp., aiming to make gains in a booming market for artificial intelligence computing.

At a company event Thursday, AMD unveiled a new generation of microprocessors, an updated AI accelerator and a server design. The products represent major upgrades to its current lineup and compare favorably with offerings from competitors, the chipmaker said.

AMD is expressing growing confidence that it can be more than a distant rival to Nvidia, whose technology has dominated the massive build-out of AI data centers and generated hundreds of billions in revenue. Both companies agree that customers’ appetite for the technology is insatiable and are racing to offer more powerful chips.

Still, investors took a skeptical view on Thursday, sending shares of AMD down about 3.5% during the company’s presentation. The stock had more than doubled in value this year heading into the event, signaling that expectations were high.

Under Chief Executive Officer Lisa Su, AMD has delivered a remarkable growth run — and Wall Street expects the pace to accelerate. Still, challenging Nvidia won’t be easy. Though AMD is widely seen as a key contender, Nvidia continues to grow faster and gets more revenue in a quarter than AMD does in a year.

The market for AI accelerators — the chips that train and run the artificial intelligence software — will soar to $1.4 trillion by 2030, AMD predicted. The opportunity for the company’s traditional data center product, central processing units, will be worth $220 billion by that date, the company said. Overall, AMD believes it will be competing in a market worth $2 trillion in 2030.

At the event, held in San Francisco, Su was joined on stage by executives from major technology companies that pledged to use AMD products. That included Anthropic PBC and OpenAI, two of the leading AI service providers.

Read More: AMD to Invest Up to $5 Billion in Anthropic, Deploy Chips

AMD showed off its new Helios server rack as part of the presentation. That computer, a design that server makers will adopt and sell themselves, consists of as many as 72 of its MI400 series accelerators. The MI455 version of that lineup will provide a massive increase in performance versus its predecessors when measured by the ability to handle memory and provide results more quickly and cheaply.

OpenAI executive Sachin Katti said his company expects to deploy Helios “at massive scale” as it scrambles to add AI infrastructure.

“If you haven’t heard already, I need more compute more quickly,” he told Su on stage.

The Epyc server processor lineup, meanwhile, is being updated with a new Venice design that will keep it well ahead of Nvidia’s competing Vera product, AMD said. Nvidia has previously claimed that its new offering gives it an edge.

Read More: Nvidia Touts Progress Getting New Rubin Design to Customers

Separately, AMD announced a tie-up with Cerebras Systems Inc. The two companies will offer a combination of Helios and Cerebras servers designed for AI customers who want the fastest responses possible. AMD products will perform the work of deciphering queries, and Cerebras, whose massive chips have huge amounts of memory on board, will handle rapidly providing answers.

The new product will first be deployed in data centers owned by Cerebras starting in the fourth quarter. It will then be offered commercially to other AI companies.

Shares of Cerebras gained 5.6% on the news.

The two will beat Nvidia’s delivery of a similar technology enabled by its recent acquisition of assets from Groq, Cerebras CEO Andrew Feldman said. His company has been working on the combined products for months, and he’s confident it will put them in the lead.

“We are first to market, and our confidence is high,” he said. “This is for the market that cares desperately about speed.”