All this talk of inflection points…
One day after Nvidia CEO Jensen Huang said at his company’s GTC Paris event that quantum computing is reaching an inflection point, AMD CEO Lisa Su, speaking at her company’s Advancing AI event, said that AI inference in at an inflection point, set to drive AMD’s total addressable market (TAM) for its AI products beyond the $500 billion by 2028 that it previously forecasted.
“Based on everything that we see today, that [TAM] number is going to be even higher, exceeding $500 billion in 2028,” she said during her keynote speech, adding, “We always believed that inference would actually be the driver of AI going forward, and we can now see that inference inflection point. With all the new use cases models, we now expect that inference is going to grow more than 80% a year for the next few years, becoming the largest driver of AI compute, and we expect that high performance GPUs are going to be the vast majority of that market, because they provide the flexibility and programmability that you need as models are continuing to evolve.”
The other trend driving that AI TAM into the stratosphere is that AI is also moving beyond the data center,” Su said. “From intelligent systems at the edge to PC experiences, and we expect to see AI deploy in every single device.”
Agentic AI also is at the heart of the increasing TAM. “[Agentic AI] is always on, constantly accessing data, looking at applications, looking at systems to really make decisions and work autonomously, these agents need high performance GPUs to generate insights in real time, but that's really only part of the story,” Su said. “What we're seeing now is as agentic AI activity increases, all of those agents are now also accessing high-performance GPUs… What we're actually seeing is the equivalent of billions of new virtual users being added to the global compute infrastructure.”
The Advancing AI event also served as the formal launch party for AMD’s much-anticipated Instinct MI350 Series of GPUs, including the MI350X and MI355X. With this series, many company watchers have expected that AMD is finally poised to close the gap between itself and AI GPU market leader Nvidia. Built on the company’s CDNA 4 GPU architecture, it can support “up to 64 GPUs in an air-cooled rack and up to 128 GPUs in a direct liquid-cooled racks delivering up to 2.6 exaFLOPS of FP4/FP6 performance,” AMD stated in a blog post.
AMD also demonstrated an end-to-end, open-standards rack-scale AI infrastructure that is already rolling out with AMD Instinct MI350 Series accelerators, 5th Gen AMD EPYC processors and AMD Pensando Pollara NICs in hyperscaler deployments such as Oracle Cloud Infrastructure (OCI). It is set for broad availability in the second half of this year.
Additionally, AMD previewed its next generation AI rack called “Helios.” It will be built on the next-generation AMD Instinct MI400 Series GPUs, which compared to the previous generation are expected to deliver up to 10x more performance running inference on Mixture of Experts models. Helios also features the Zen 6-based AMD EPYC Venice CPUs and AMD Pensando Vulcano NICs.