What is Cerebras' key product and how does it differ from Nvidia's approach?

Cerebras' flagship wafer-scale engine (WSE) is a single chip the size of a dinner plate containing trillions of transistors, with memory placed directly on the chip. This design reduces data-transfer delays and is more efficient than Nvidia's approach of connecting thousands of smaller graphics processors together, each relying on separate high-bandwidth memory.

What is the $20 billion contract with OpenAI?

Cerebras has a $20 billion multi-year agreement to provide AI compute to OpenAI, which the company views as key to justifying its valuation. The company is racing to expand chip production to support this deal.

By how much did second-quarter sales grow?

Second-quarter sales rose 74.3% to $180.11 million, and core cloud and services revenue—which reflects the OpenAI ramp—nearly quadrupled to $127.73 million in the second quarter.

AI Stocks & MarketsAI Business & IndustryYahoo Finance AI1h agoAug 13, 2026, 06:00 JST2 min read

Cerebras raises 2026 revenue guidance on AI chip demand surge

Send on LINE

Key takeaway

Cerebras Systems, a chip designer based in Sunnyvale, California, raised its 2026 revenue and profit-margin forecasts on Wednesday, citing strong customer demand for its AI processors as companies invest heavily in data-center capacity.
The company's wafer-scale engine chip places memory directly on the processor to speed up inference tasks and reduce costs, positioning it to compete with Nvidia in a market where component shortages have driven up competitor pricing.
The company is backed by a $20 billion multi-year contract with OpenAI and plans to more than triple revenue in 2027.

3 Key Points

What happened
Cerebras Systems raised its 2026 adjusted revenue forecast to $880 million–$890 million (from $855 million–$865 million) and boosted its annual adjusted gross margin outlook to 41%–43% (from 38%–41%), citing strong demand for its chips from companies expanding data-center capacity for AI services.
Why it matters
The company is betting on inference—the computational work behind chatbot queries—as a path to challenge Nvidia's dominance in AI processors. Its wafer-scale engine (WSE) chip places memory directly on a single large processor, reducing data-transfer delays and insulating it from soaring memory-component costs that have pushed Nvidia's prices higher, according to CEO Andrew Feldman.
What to watch
Cerebras plans to more than triple revenue in 2027 and is racing to expand production to support a $20 billion multi-year deal to provide AI compute to OpenAI—a contract the company views as key to justifying its valuation. Despite raising guidance, the stock fell more than 14% in extended trading after a 11.6% gain in the regular session.

Summaries like this, in your inbox every morning.

Get news like this every morning — free

FAQ

What is Cerebras' key product and how does it differ from Nvidia's approach?: Cerebras' flagship wafer-scale engine (WSE) is a single chip the size of a dinner plate containing trillions of transistors, with memory placed directly on the chip. This design reduces data-transfer delays and is more efficient than Nvidia's approach of connecting thousands of smaller graphics processors together, each relying on separate high-bandwidth memory.
What is the $20 billion contract with OpenAI?: Cerebras has a $20 billion multi-year agreement to provide AI compute to OpenAI, which the company views as key to justifying its valuation. The company is racing to expand chip production to support this deal.
By how much did second-quarter sales grow?: Second-quarter sales rose 74.3% to $180.11 million, and core cloud and services revenue—which reflects the OpenAI ramp—nearly quadrupled to $127.73 million in the second quarter.

Yahoo Finance AIRead Original Article