Cerebras Systems reported Q2 2026 core revenue of $210 million and raised its full-year 2026 guidance to $880–890 million (up from $855–865M), signaling strong demand for AI inference chips despite a 14% post-earnings stock decline. CEO Andrew Feldman said AI demand is 'through the roof' and that companies are paying premium prices for Cerebras's low-latency inference architecture. The company's cloud and services division alone generated $126 million in Q2 revenue.
Cerebras's gross margins are expanding to 38–40% in the current quarter—a key investor concern—because fast-inference products command premium pricing. The company expects revenue to triple in the next fiscal year as it benefits from larger manufacturing scale and improved component pricing. Cerebras has $25.4 billion in remaining performance obligations, which the CEO said reflects 'extraordinary future demand.' The company also announced partnerships with AMD and noted that OpenAI can use its chips for GPT-5.6-Sol inference.
The stock's post-earnings decline, despite raised guidance, reflects investor concern about gross-margin pressure from temporary reliance on rented third-party compute capacity while waiting for Cerebras's own data centers to come online. For architects deploying inference at scale, Cerebras's rapid ramp—moving from $55M in Q1 to a $35B+ annualized revenue run rate—and its premium margins show the market will pay for specialized, low-latency inference silicon. The company's ability to sustain those premiums as competitors (Nvidia, AMD, Groq) expand inference offerings remains the key watch.