Cerebras Raises 2026 Targets As AI Chip Demand Keeps Building

Cerebras has raised its 2026 targets, giving investors another sign that demand for AI chips and fast inference infrastructure remains strong even outside Nvidia’s GPU empire.

The AI chip company reported Q2 results after the market closed on Wednesday. According to its earnings release, core revenue reached $209.9 million, up 103 percent year-on-year, while GAAP revenue rose to $180.1 million. Core cloud revenue grew 281 percent, showing how much of the story is now moving toward cloud access to Cerebras systems rather than only hardware sales.

Cerebras now expects full-year 2026 core revenue between $880 million and $890 million, above its prior forecast of $855 million to $865 million. It also lifted its core gross-margin outlook to 41 percent to 43 percent, up from 38 percent to 41 percent previously. Reuters framed the upgrade as a response to strong AI chip demand.

The company is still losing money, and that matters. Cerebras reported a GAAP net loss of $450.5 million for the quarter, or $2.98 per share. The market reaction was negative in after-hours trading, partly because investors are watching margins, losses and how quickly the company can convert demand into profitable revenue.

The long-term pitch remains Cerebras’ wafer-scale approach. Its Wafer-Scale Engine is a single giant chip designed to reduce the bottlenecks that come from linking many smaller processors together. The company argues that this design is particularly useful for fast inference, long outputs and workloads where latency matters.

That is why Cerebras has become more relevant as the AI market shifts from training models to serving them. Training grabs headlines, but inference is what happens every time users ask a model to write, code, reason, translate, summarize or generate output. If demand for AI applications keeps rising, inference capacity becomes a huge market.

OpenAI has already tied itself to Cerebras through a major infrastructure partnership. The OpenAI-Cerebras deal added 750MW of ultra-low-latency AI compute to OpenAI’s platform, and OpenAI described Cerebras’ systems as purpose-built for accelerating long outputs. That connection gives Cerebras a stronger credibility story in a market where many AI chip challengers are still trying to prove they can win major customers.

The company also says it has a large remaining performance obligation and plans to more than triple revenue in 2027. That ambition is exactly what investors want to hear, but it also raises the execution bar. Cerebras has to scale data-centre capacity, manufacturing, cloud delivery, customer support and margins at the same time.

This fits the wider AI infrastructure story we have been tracking, from Nvidia’s AI compute financing push to Cisco’s surge in AI infrastructure orders. The AI boom is becoming a full supply chain: chips, servers, networking, data centres, power and financing.

For Cerebras, the opportunity is clear. The market wants alternatives to Nvidia, especially for inference workloads where speed and cost matter. The risk is also clear. Strong revenue growth is useful, but investors will keep asking how quickly the company can turn demand into sustainable margins and cash flow. Raising targets is a good signal. Proving the economics will be the harder part.