Positive Outlook on Growing AI Inference Demand: Freedom Upgrades Cerebras (CBRS.US) Rating to "Buy", Wafer-Level Chip Advantages Gain Attention

According to Zhitong Finance APP, Freedom Capital Markets has upgraded the stock rating of artificial intelligence (AI) chip manufacturer Cerebras Systems (CBRS.US) from "Hold" to "Buy," assigning a target price of $209. The institution believes that despite the recent market sell-off following the news that important partner OpenAI has switched to using Nvidia (NVDA.US) GPUs, the projected continued excess demand for AI computing power over supply in the coming years, along with Cerebras’ technological advantages in high-speed AI inference, means the recent stock price pullback offers an opportunity to reassess its investment value.

Cerebras is a company focused on AI computing chips, with its core product being the Wafer Scale Engine (WSE). Unlike traditional GPU architectures, the company integrates massive amounts of computing resources onto a single wafer, providing high-performance computing capabilities for large AI models, with a primary emphasis on improving inference speeds.

OpenAI’s Adoption of Nvidia GPUs Raises Concerns, Cerebras Shares Decline

Freedom analyst Paul Meeks noted in a research report released on Friday that Cerebras’ recent share price pressure is mainly related to market concerns about its key partner OpenAI’s choice of technology.

It is reported that OpenAI adopted Nvidia GPUs rather than Cerebras' wafer-scale chips in the "Ultrafast" high-speed version of its GPT-1 Sol model. This news has prompted investor doubts regarding whether Cerebras can continue to secure OpenAI orders in the future, as well as concerns over the competitiveness of its products in the high-speed AI inference market. Meeks pointed out that this development may be an important reason for Cerebras’ roughly 17% cumulative decline in share price recently.

Cerebras completed its initial public offering (IPO) in mid-May this year at an issue price of $185 per share. After listing, the stock quickly surged above $300, but with the market reassessing its competitive prospects, the price has now pulled back significantly.

However, Meeks believes that investors may have overemphasized the negative impact of OpenAI’s adoption of Nvidia GPUs. According to the institution, the AI computing power market remains in a rapid expansion phase, and a single client using a competitor’s hardware for a specific model does not mean Cerebras’ future business opportunities will be fundamentally impaired.

Cerebras Still Poised to Benefit from AI Inference Growth

Meeks believes that over the next several years, demand for AI computing will likely continue to outpace industry supply, providing growth opportunities for multiple chip vendors, including Cerebras.

He stated that even if OpenAI chooses Nvidia GPUs for certain applications, Cerebras is still likely to maintain significant business cooperation with OpenAI. At the same time, Cerebras has other important partners, including AMD (AMD.US) and Amazon’s (AMZN.US) cloud platform AWS.

He pointed out that these partnerships will help Cerebras expand its market opportunities and reduce investor concerns related to technology decisions by any single client. More importantly, as the computing needs of the AI industry shift from model training to inference tasks in real-world applications, inference speed and operating efficiency are becoming crucial competitive factors for AI infrastructure.

Unlike the training phase, which mainly focuses on optimizing model parameters, the inference phase needs to continuously generate outputs in response to user requests. As AI applications proliferate, improving response speed while controlling costs is a key challenge chip manufacturers must address.

Meeks pointed out that Cerebras provides competitive solutions in high-speed inference and stands to benefit from this trend.

AI Computing Architectures May Shift Toward Divided Labor; Complementary Room for Cerebras and Nvidia

Aside from directly competing for AI chip orders, Meeks also sees the potential for a complementary relationship between Cerebras and Nvidia within future AI computing architectures. As Meeks notes, as AI inference tasks grow more complex, the industry may gradually adopt what’s termed a “separated computing resource” architecture—using different types of chips for different computation stages to improve overall efficiency.

Large language model inference usually involves two main stages: prefill and decode. The prefill stage mainly handles users' input prompts and contextual information and generally requires strong parallel computing capability; the decode stage is responsible for generating outputs step-by-step, which has different demands for computational latency and data processing efficiency.

Meeks believes a future architecture may emerge where Nvidia GPUs take on the prefill tasks, while Cerebras’ WSE handles decode tasks. This division of labor means that, although Cerebras and Nvidia compete directly in the AI chip market, their technologies could complement each other in specific application scenarios.

Meeks even suggests that Cerebras in the future may not rule out cooperation with its chief competitor Nvidia in such architectures. However, this remains an analyst’s forecast on future technology development and does not represent any officially announced collaboration between the two companies.

Overall, Meeks believes the recent sell-off in Cerebras shares due to OpenAI’s use of Nvidia GPUs does not change the long-term trend of rising demand in the AI inference market. As AI application scale continues to expand, Cerebras’ wafer-scale chips are still likely to secure more business opportunities in the high-speed inference domain.

On this basis, Meeks has upgraded Cerebras’ rating from “Hold” to “Buy,” with a target price set at $209.