🥛 The most important AI IPO? 👀

Cerebras Systems IPO valuation

Despite the engineering marvel and oversubscription frenzy, Cerebras trades at 91x trailing revenue, 86% of revenue is from UAE-tied entities, and the $20B OpenAI contract is conditional—making the valuation extremely risky.

AI industry shift from training to inference

The phase of building the biggest AI brain is largely done; the inference market growing from $106B to $255B by 2030 is the whole game, and Cerebras arrived at the IPO window right as this shift accelerates.

Micron (MU) as AI infrastructure trade

The author took a massive position in MU about a month ago and it has already gone up over 100%, as memory is becoming one of the real bottlenecks of the AI economy.

Sam Altman's governance role at Cerebras

Sam Altman being simultaneously a personal investor in Cerebras, CEO of its largest customer, and connected as a creditor is exactly the kind of relationship structure that makes institutional investors and compliance teams nervous.

Rising HBM memory prices benefiting Cerebras pitch

As MU and SK Hynix go higher, NVIDIA-based infrastructure gets more expensive to build and operate, making the Cerebras pitch—a chip that sidesteps HBM costs entirely—easier to make to every CFO.

Wafer-Scale Chip: Don't Cut the Silicon

Cerebras Systems refused to follow 60 years of semiconductor convention by not dicing silicon wafers into individual chips, instead building one massive 46,225mm² chip that eliminates memory bandwidth bottlenecks critical for AI inference.

Every major chip company—Nvidia, AMD, Intel—dices wafers; Cerebras bets that keeping the entire wafer intact creates a fundamentally superior memory architecture for inference workloads that no conventional chipmaker can match.