Ampere Computing unveiled 256-core processor targeting data center deployments.

Ampere Computing has announced plans for a faster, more efficient 256-core server processor to address data center power demands. The startup chipmaker, which launched in 2018 under CEO Renee James—previously Intel's president—revealed that its forthcoming 256-core AmpereOne chip, available next year, will deliver 40% better performance than any CPU on the market today while using the same amount of power as its 192-core AmpereOne chip released last year. To expand its reach in AI applications, Ampere is also developing a joint solution featuring its Arm-based CPUs with Qualcomm Cloud AI 100 Ultra accelerators for AI inferencing, with Supermicro selling a server powered by both chips. As Chief Product Officer Jeff Wittich explained, "Data centers are increasingly consuming more power, and AI is a big catalyst for this. We can come in and help with a more efficient solution, whether it's the most gigantic models with Qualcomm or smaller models that just run on CPUs."

Ampere's cloud provider customers include Oracle Cloud, Google Cloud, Equinix Metal, and Tencent Cloud, and the company is pursuing enterprise on-premises data centers with servers from Hewlett Packard Enterprise and Supermicro. However, the company faces growing competition from hyperscalers building their own Arm chips—Google's Axion and Microsoft's Cobalt—which could impact Ampere's position in that segment. According to market data from Omdia, Arm CPUs have captured only 9% of the market as of 2023, while x86 chips dominate with Intel holding 61% market share and AMD reaching 27%. Other companies producing Arm CPUs include Amazon Web Services and Nvidia.

Ampere's CPUs alone can run eight billion to 13 billion parameter large language models. In April, Oracle Cloud Infrastructure announced it was running Meta's eight billion parameter Llama 3 on Ampere CPUs, with benchmarks showing the same performance as an Nvidia A10 GPU with an x86 CPU while using just one-third of the power and costing 28% less. The joint Ampere-Qualcomm solution targets larger models, with Wittich noting that "when you get to hundreds of billions of parameters or a trillion-parameter model, that's a specialized enough type of workload that you might want to scale out across something that's really specialized to do that task—and that's where the Qualcomm solution comes in." Ampere is the second company to partner with Qualcomm on AI inferencing, following AI hardware startup Cerebras, which collaborates with Qualcomm to optimize models trained on Cerebras hardware for inference on Qualcomm's Cloud AI 100 Ultra accelerator.