-+ 0.00%
-+ 0.00%
-+ 0.00%

Cerebras (CBRS.US) is favored by Wedbush: OpenAI cooperation continues to deepen, and 750MW computing power orders unleash capacity expansion potential

Zhitongcaijing·08/17/2026 13:41:14
Listen to the news

The Zhitong Finance App learned that on Monday, Cerebras Systems (CBRS.US) became the focus of the market. Wedbush Securities released a research report stating that after OpenAI revealed that its GPT-5.6 Sol Ultrafast model will be supported by Cerebras, Wedbush believes that the existing partnership between the two sides is making “positive progress.” Boosted by this news, Cerebras shares rose 2.2% in the premarket.

In a report to clients, Wedbush analyst Matt Bryson said, “We believe this news marks the continued advancement of Cerebras' partnership with OpenAI. Although OpenAI has committed to procuring 750 megawatts of Cerebras computing power capacity by 2028, we note that the agreement structure allows OpenAI to extend the contract period related to this capacity and further expand the capacity until 2030.”

The processing speed of the newly launched UltraFast mode is up to 14 times faster than the OpenAI standard model, and is currently being rolled out in batches for specific customer groups with high requirements for low latency. OpenAI said it has observed “encouraging” test feedback in application scenarios such as incident response and reliability, financial research and security, customer support and voice interaction, commercial and real-time research and experiments. Early GPT-5.6 Sol users in the UltraFast model included quantitative trading giant Jane Street, AI lead generation company Podium, and financial companies Basis and Rogo.

Cerebras and OpenAI reached a $20 billion cooperation agreement in April this year. Under the agreement, Cerebras will provide OpenAI with servers driven by its self-developed chip over the next three years. Meanwhile, OpenAI promised to provide Cerebras with approximately $1 billion in funding to support the data center development needed to run its AI products.