A bullish analyst target for Cerebras Systems rests on an unusually large promise: the company reported $25.4 billion of remaining performance obligations at the end of the second quarter of 2026. That contracted visibility supports the view that wafer-scale systems can win a meaningful share of AI inference. It also concentrates the investment case around a small number of customers and years of capacity delivery.
The underlying quarter showed real growth. Cerebras reported GAAP revenue of $180.1 million, up 74% from a year earlier. It also reported "core revenue" of $209.9 million, up 103%, a company-defined measure that excludes certain effects and should not be confused with GAAP sales. Management guided full-year core revenue to $880 million to $890 million.
OpenAI is central to the backlog. The companies announced a multi-year agreement for 750 megawatts of Cerebras inference capacity beginning in 2026. A regulatory filing also disclosed an option for an additional 1.25 gigawatts. Those numbers explain why analysts can model steep future growth, but an option is not an obligation and performance commitments are not revenue already earned.
The valuation argument needs discipline
The source article highlighted a $330 target from UBS and a large gap between the share price and analyst estimates. Price targets are opinions built on assumptions about utilization, margins and delivery timing. They do not resolve the accounting and operating risks.
Cerebras said stock-based compensation weighed heavily on GAAP results. Capacity expansion can also depress margins before contracted customers fully use the infrastructure. More important, OpenAI, G42 and a limited set of partners account for much of expected demand. If deployments slip, customer priorities change or competing GPU systems improve, the same concentration that creates visibility can magnify disappointment.
One claim in the source said a future OpenAI "ultrafast" workload had moved to Nvidia. Current official OpenAI material instead identifies GPT-5.6 Sol ultrafast as running on Cerebras, so the unsupported future-model claim is excluded here rather than reconciled by guesswork.
The next evidence is operational: third-quarter core revenue, gross-margin progression and disclosed capacity coming online. Cerebras has a substantial backlog and credible product demand. Investors still need proof that the company can convert both into GAAP revenue and durable cash generation at the pace implied by aggressive targets.