Nvidia Competitor Cerebras Rebounds Nearly 20%. Market Bets It Will Benefit From HBM Shortages, Citi Sees 90% Upside for Stock

TradingKey - On June 29, Eastern Time, Cerebras Systems ( CBRS) once surged over 20%. As of press time, it was up 19.36% at $216.75. The latest portfolio disclosure of Cathie Wood, CEO and CIO of Ark Invest, showed that she added over 20,000 shares of CBRS stock last Friday, reflecting her continued optimism about the AI chip sector.

It is reported that the AI chip company, widely regarded as a competitor to Nvidia, has seen its stock price fall by about 40% since its IPO, hitting an all-time low of $160.81 last Friday.

[Source: TradingView]

The primary reason for the stock price hitting a new low is that in its first quarterly earnings report since going public, the company projected its second-quarter core gross margin would contract sharply to 36%-38% from 46.5% in the first quarter, with the full-year core operating margin expected to range between -28% and -32%. Investors were disappointed by the full-year outlook, sparking a sell-off.

Despite some market concerns over the company's gross margins, the 10 analysts covering the stock have a consensus "Buy" rating. Following the earnings release, Morgan Stanley and UBS raised their price targets to $273 and $320, respectively, both reiterating their bullish stance. Meanwhile, Citi initiated coverage on the stock with a "Buy" rating and set a Wall Street-high price target of $340, representing an upside potential of about 90% from the current share price.

Meanwhile, market analysis suggests that against the backdrop of tightening memory supply, Cerebras is less affected as it utilizes SRAM instead of HBM. This positions the company to achieve incremental performance gains amid industry-wide capacity constraints. From this perspective, worsening memory shortages will only further amplify Cerebras's competitive advantage, helping it capture more market share.

It is understood that what currently limits AI performance is no longer computing power itself, but rather the efficiency of data movement within the memory system. The design of the memory hierarchy architecture has become the ultimate key to system performance.

While mainstream GPU designs currently rely on HBM to rapidly scale up computing power, this has also made them highly dependent on scarce HBM memory stacking and advanced packaging capacity. Consequently, the entire supply chain is experiencing strategic tightness, continuing to serve as the core bottleneck for expanding AI computing capacity.

Cerebras, on the other hand, breaks away from the traditional GPU technical framework by using a wafer-scale architecture to reconstruct the memory layout logic: SRAM is no longer a small, private cache attached to individual computing cores, but is instead evenly distributed across the entire wafer and shared by a global network of computing units, reshaping the execution model of AI computing from the ground up.