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

Optimistic that demand for AI inference will continue to grow, Freedom raised the Cerebras (CBRS.US) rating to “buy” wafer-level chips and attracted attention

Zhitongcaijing·10/02/2026 16:09:03
Listen to the news

The Zhitong Finance App learned that Freedom Capital Markets raised the stock rating of artificial intelligence (AI) chip maker Cerebras Systems (CBRS.US) from “hold” to “buy” and gave a target price of $209. The agency believes that although Cerebras was recently sold off by the market due to news that an important partner OpenAI is switching to Nvidia (NVDA.US) GPUs, considering that demand for AI computing power is expected to continue to exceed supply in the next few years, and Cerebras' technical advantage in the field of high-speed AI inference, the recent stock price correction provides an opportunity to re-evaluate the value of its investment.

Cerebras is a company focusing on AI computing chips, and its core product is Wafer Scale Engine (WSE). Unlike traditional GPU architectures, the company provides high-performance computing capabilities for large-scale AI models by integrating a large number of computing resources on the entire wafer, with a particular focus on improving inference speed.

OpenAI's adoption of Nvidia GPUs raises concerns Cerebras shares fall

Freedom analyst Paul Meeks pointed out in a research report released on Friday that Cerebras's recent stock price pressure is mainly related to market concerns about its key partner OpenAI technology choices.

According to reports, OpenAI used Nvidia GPUs instead of Cerebras' wafer-level chips in the “Ultrafast” ultra-high-speed version of its GPT-1 Sol model. This news raised investors' concerns about whether Cerebras will continue to receive orders from OpenAI in the future and the competitiveness of its products in the high-speed AI inference market. Meeks said that related news may be an important reason for the recent cumulative decline in Cerebras stock prices by about 17%.

Cerebras completed its initial public offering (IPO) in mid-May this year at an issue price of $185 per share. After listing, the stock rose rapidly to over $300, but recently, as the market re-evaluated its competitive prospects, the stock price declined markedly.

However, Meeks believes that investors may have overamplified the negative impact of OpenAI's adoption of Nvidia GPUs. According to the agency, the AI computing power market is still in a stage of rapid expansion. If a single customer uses competitor hardware on a certain model, it does not mean that Cerebras' future business opportunities will be fundamentally weakened.

Cerebras is still expected to benefit from AI inference growth

Meeks believes that demand for AI computing may continue to exceed the industry's supply capacity in the next few years, which will provide room for growth for many chip suppliers, including Cerebras.

He said that even if OpenAI chooses Nvidia GPUs for some applications, Cerebras is expected to continue to develop numerous business partnerships with OpenAI. At the same time, Cerebras also has other important partners, including AMD (AMD.US) and AWS, a cloud computing platform owned by Amazon (AMZN.US).

He pointed out that these partnerships are expected to help Cerebras expand market opportunities and reduce investors' concerns about single customer technology choices. More importantly, as the AI industry's computational requirements gradually expand from model training to inference tasks in actual applications, inference speed and operational efficiency are becoming important competitive factors for AI infrastructure.

Unlike the training phase, which mainly focuses on model parameter optimization, the inference phase requires continuous output generation according to user requests. With the popularity of AI applications, how to control costs while improving response speed is becoming an important problem for chip manufacturers to solve.

Meeks notes that Cerebras has competitive solutions in high-speed reasoning and is expected to benefit from this trend.

AI computing architecture or moving towards division of labor cooperation Cerebras and Nvidia have complementary space

In addition to directly competing for AI chip orders, Meeks also sees the possibility that Cerebras and Nvidia will form a complementary relationship in future AI computing architectures. Meeks pointed out that as AI inference tasks become more complex, the industry may gradually adopt the so-called “separation of computing resources” architecture, that is, the use of different types of chips for different computational processes to improve overall efficiency.

Large language model reasoning usually involves two main stages: prefill (prefill) and decode (Decode). The pre-filling stage is mainly responsible for processing prompts and contextual information entered by users, and usually requires strong parallel computing capabilities; the decoding stage is responsible for gradually generating output content, which has different requirements for calculation delays and data processing efficiency.

Meeks believes that in the future, there may be a computing architecture where Nvidia GPUs take on pre-filling tasks, while Cerebras' WSE is responsible for decoding tasks. This division of labor model means that although Cerebras and Nvidia have direct competition in the AI chip market, the two technologies may also complement each other in specific application scenarios.

Meeks even believes that in the future, Cerebras and its main competitor Nvidia will not rule out using their respective strengths in this type of architecture. However, this is still an analysts' judgment on the future direction of technology development; it does not mean that the two companies have announced relevant cooperation.

Overall, Meeks believes that the recent stock price sell-off experienced by Cerebras due to OpenAI's use of Nvidia GPUs has not changed the trend of long-term demand growth in the AI inference market. As the scale of AI applications continues to expand, Cerebras' wafer-level chips are still expected to gain more business opportunities in the field of high-speed inference.

Based on the above judgment, Meeks upgraded the Cerebras rating from “hold” to “buy,” and the target price was set at $209.