Inference Hits $23.3B in 2026: Cerebras vs. SpaceX — Who Wins AI Compute?

The AI community should focus on one number: $23.3 billion. That's Gartner's projected 2026 AI inference spend, overtaking the $19 billion training market for the first time. Cerebras is architecturally positioned for that shift with wafer-scale systems optimized for production workloads, while SpaceX pairs Grok with AI infrastructure funded by Starlink. The question is whether a pure-play inference vendor can outcompete a diversified AI stack over the next five years.

The 2026 public-market calendar produced two of the most consequential AI infrastructure listings in recent memory. Cerebras Systems listed on Nasdaq under the ticker CBRS on May 14, and Space Exploration Technologies — SpaceX — followed with the ticker SPCX on June 12. The two companies represent opposing bets on how value will accrue in AI over the next five years: Cerebras is a pure-play manufacturer of wafer-scale AI systems that sells cloud access to its computing power, while SpaceX is a diversified operator spanning reusable rockets, the Starlink satellite network, and an AI segment anchored by the Grok large language model and AI computing infrastructure.

“

The company also raised its full-year core revenue outlook to a range of $880 million to $890 million, up from a prior range of $855 million to $865 million.

The macroeconomic backdrop is defined by a structural shift in AI spending. For nearly a decade, the industry's capital has flowed overwhelmingly toward training large models. That dynamic is changing. Gartner projects that global spending on AI inference will reach $23.3 billion in 2026, surpassing the $19 billion spent on training for the first time. By 2027, inference is expected to account for 59% of AI-optimized cloud infrastructure spending. This pivot matters because inference — running models in production — favors different hardware economics than training, and it is where Cerebras's wafer-scale architecture is designed to compete.

Cerebras's most recent results support the thesis. In the second quarter of fiscal 2026, ending June 30, non-GAAP cloud and services revenue jumped 287% year over year to $127.7 million. Total non-GAAP core revenue rose 103.3% to $209.9 million, ahead of management's guidance of approximately $194 million. The company also raised its full-year core revenue outlook to a range of $880 million to $890 million, up from a prior range of $855 million to $865 million. That growth rate is striking for a company of Cerebras's size — it is much smaller than SpaceX in market capitalization — but it comes with caveats. The company faces higher customer concentration and execution risk relative to its larger peer.

SpaceX brings a very different financial profile to the same comparison. The company generated nearly $2.6 billion in revenue from its AI business in the second quarter of fiscal 2026 alone, and it has a profitable Starlink-driven Connectivity business that can fund AI expansion without the same capital-market pressure a pure-play faces. That combination — a scaled, cash-generating satellite internet business plus a fast-growing AI segment — gives SpaceX a financial cushion Cerebras does not have. Yet SpaceX's size also means its AI segment's percentage growth, while substantial, is layered on top of an already-large base.