AMD and Cerebras Partner to Accelerate AI Inference Chips
AMD teams with Cerebras to deliver ultra-low latency AI inference, while its new EPYC server CPU claims a 20% edge over Nvidia.
AMD unveiled a major partnership with Cerebras on Wednesday, combining its Helios GPU rack with Cerebras' wafer-scale AI chip to build an ultra-low latency AI inference platform aimed squarely at enterprise and hyperscale AI workloads. The joint solution is expected to reach customers through Cerebras Cloud in the second half of 2026, according to AMD CEO Lisa Su, who said the collaboration targets high-speed inference at scale.
Separately, AMD made bold claims about its next-generation EPYC "Venice" server processor, asserting it delivers 20% higher performance than Nvidia's competing CPU offering and maintains a meaningful lead over Arm-based server chips. The Venice processor is already in full production with strong customer demand, and systems built on the chip are slated to begin shipping in the fourth quarter of this year.
Read more ECB Eyes September Rate Hike as Energy Prices Drive Inflation →
The announcements signal AMD's broader ambition to compete across the entire AI technology stack — spanning training infrastructure, inference platforms, and enterprise data centers — rather than relying solely on GPU sales. The strategy pairs AMD's hardware muscle with Cerebras' specialized inference technology in a bid to carve out greater share in an AI infrastructure market that continues to expand rapidly.
Markets greeted the news with a split reaction. AMD shares slipped 3.06% to $535, though they recovered from an intraday low near $525 and climbed back above key 100-hour and 200-hour moving averages, a technically constructive signal. Cerebras shares rose 4.03% to roughly $218, continuing a recovery from a June 24 low of $160 after the stock had plunged from its post-IPO high of $386.34.
Continue reading at Forexlive.
