Cerebras Sells Off After Earnings: Is This a Market Disconnection?

Key Points

• Cerebras shares fell roughly 16% after Q2 results, but its expanding role in the fast-growing AI inference market remains largely intact.
• The company's disaggregated inference partnership with AMD's Helios racks, plus deals with AWS and OpenAI, helped push remaining performance obligations above $25 billion.
• Analysts maintain a Moderate Buy rating with a roughly $302 average price target, citing a strong balance sheet and manageable short interest despite near-term volatility.

Cerebras' NASDAQ: CBRS Q2 results provided a catalyst for sellers, initially sending shares down roughly 16% in extended trading following the report. The reality, however, is that this market’s near-term focus is disconnecting it from reality.

While one-offs, including non-cash impairment and share-based compensation, cut into profitability, and revenue growth was tepid relative to expectations, the fundamental story continues to strengthen.

That story centers on inference and Cerebras' ultra-large semiconductor products. They provide lightning-fast output and are quickly becoming critical to inference markets.

Expected to grow at a solid, double-digit compound annual growth rate for the next 10 years, inference infrastructure is forecasted to double the value of the training market.

The Market May Be Underestimating Inference Demand

Inference is going to be a big market; if you think it's already growing, just wait until Advanced Micro Devices’ NASDAQ: AMD Helios racks are widely deployed. Deals with AMD, Amazon's NASDAQ: AMZN AWS, and OpenAI underscore its importance, while the OpenAI agreement represents a significant portion of the company’s remaining performance obligations (RPO). Where does CBRS fit into the AI architecture? Disaggregated inference.

Disaggregated inference splits the two phases of AI, preload and output, into two distinct jobs. On one hand, AMD’s Helios racks will act as the brain, decoding the queries, loading the data, and understanding context. Then the CBRS inference engine, which is natively integrated into the Helios system, takes the data and context and uses them to spit out an answer. The benefits to hyperscalers include speed and cost, two drivers of revenue and earnings.

What the market is getting wrong is that AMD’s MI450 Series, including the MI455X GPUs powering Helios, has only recently entered production, and the inference boom is only beginning. In this scenario, Cerebras' headwinds, which include accelerated capital expenditures (CapEx) to ramp capacity to meet demand, will fade over time as it ramps capacity. As it stands, the company is renting back some of its own systems from cloud customers to meet strong demand for inference while it executes its strategy.

This year presents numerous catalysts, including hyperscaler ramps, production ramps, and converting its backlog into revenue. RPO is a critical detail, as it reflects the strength of demand without any near-term noise. RPO expanded to over $25 billion in Q2, providing substantial visibility into future revenue. That backlog supports the company’s 2027 goal of more than tripling revenue.

Cerebras Bypasses More Than One AI Bottleneck

Cerebras' utility lies in its ability to bypass AI bottlenecks, including the memory bottleneck at multiple levels of the process. Not only does its large-wafer design enable it to operate without high-bandwidth memory (HBM), but it also bypasses the advanced packaging processes that go with it. Meanwhile, it has secured sufficient capacity with Taiwan Semiconductor Manufacturing Company NYSE: TSM to see it through the end of next year and beyond, which greatly derisks the outlook.

Cerebras' biggest risk is its customer concentration. Centered in only a few large entities, this sets it up for lumpiness in upcoming results. The offset, however, is the MI450 Series rollout, which supports Cerebras’ disaggregated inference opportunity as the market increasingly shifts toward inference. The likely outcome is that hyperscale and enterprise clients expand in numbers while deepening product use. Additionally, a high price multiple and ongoing losses set the stage for volatility and knee-jerk price pullbacks, as is the case now.

Cerebras’ post-earnings pullback isn’t as bearish as it may seem. The move comes within months of the IPO, with the market still in post-IPO discovery and plenty of catalysts in play. The critical details are that market action remains well above its post-IPO low, and the analysts are overlooking the near-term weakness. The group is responding with reaffirmed targets, highlighting core growth, CapEx focus, and plans to scale capacity 10x by year’s end. They rate the stock a Moderate Buy, with 10 of 12 analysts rating it Buy or Strong Buy and an average price target of about $302.

Short sellers are present but not a significant risk, as of mid-August. While short interest has increased from its immediate post-IPO levels, it remains tepid, near 6% of the float, with a days-to-cover ratio of 2.2. Signs short sellers may choose to exit this market include the balance sheet, which reveals a healthy position. The company has no significant debt, more than $8.5 billion in cash, equivalents, and investments, a robust net cash position relative to total liabilities, and is capable of executing its production ramp.

Continue following MarketBeat

Add MarketBeat as your preferred source on Google to see our latest stories in your feed.

Add As Preferred Source