Cerebras AMD partnership lifts shares as Helios AI systems add its chips
Cerebras shares rose about 4% after AMD said Helios AI systems will use Cerebras wafer-scale chips in data centers.
By Maya Okafor · Markets Writer
· 3 min read
The Cerebras AMD partnership gave Cerebras shares a lift Thursday, with the stock rising about 4% after the AI chipmaker announced a deal tied to Advanced Micro Devices’ Helios AI systems, according to CNBC. For everyday investors, the move adds another data point in the race to build faster AI infrastructure, where speed and power use are becoming key selling points.
Cerebras CEO Andrew Feldman said at AMD’s AI conference in San Francisco that Cerebras chips will be used in AMD Helios AI systems installed in Cerebras data centers beginning later this year, CNBC reported. Server customers will also be able to configure AMD systems with Cerebras’ wafer-scale chips, according to the report.
AMD used the event to discuss new chips and its Helios integrated system, CNBC reported. Helios is AMD’s bundled AI computing platform, designed to combine chips and system components for customers building artificial intelligence infrastructure.
What does the Cerebras AMD partnership mean?
The agreement means AMD can pair its Helios systems with Cerebras chips focused on fast AI responses. Cerebras’ wafer-scale chips are built around a different approach from standard graphics processors: they use a very large chip design intended to speed up AI work, according to CNBC’s description of the company’s technology.
The companies are aiming at what the AI industry calls ultra-low latency. Latency is the delay before a system responds, and in AI services it can affect how quickly a chatbot, search tool or enterprise software assistant starts returning an answer.
CNBC reported that Cerebras-style chips are tuned to deliver the first AI answers as quickly as possible, while giving up some flexibility and total power characteristics compared with other designs. AMD and Cerebras said their system would deliver five times more tokens per second per watt than competing systems, according to CNBC. Tokens per second per watt is a performance measure for how much AI output a system can generate relative to the electricity it uses.
Feldman framed speed as a core requirement for customers using AI systems. “When something’s a necessity, people want to use it, and they want to use it quickly,” he said, according to CNBC.
How has Cerebras stock traded since its IPO?
Cerebras went public in May and has traded sharply in both directions since its debut, CNBC reported. The company priced its IPO at $185, then climbed as high as $386.34 on its first trading day before falling below $161 in late June.
After Thursday’s move, Cerebras shares were trading at $219.80, according to CNBC. That price remains above the IPO price but well below the first-day high, underscoring how volatile newly public AI infrastructure stocks can be when investors are trying to value fast-growing but still-developing markets.
The AMD agreement follows another major Cerebras deal. In January, Cerebras announced an agreement with OpenAI to provide 750 megawatts of computing power through 2028, a contract CNBC said was worth more than $10 billion.
The broader competitive backdrop is also moving quickly. CNBC reported that Nvidia, AMD’s largest rival in AI chips, bought assets from Groq in December for $20 billion to bring Groq’s low-latency technology into Nvidia systems. That deal, along with AMD’s Cerebras partnership, shows that the AI chip fight is expanding beyond raw computing power into response speed and energy efficiency.
This story draws on original reporting from CNBC.
