Analysis
Two of this week's largest private rounds -- Valar Atomics at a $6 billion valuation building factory-produced nuclear microreactors, and OLIX at $3.3 billion building photonic AI inference chips -- share a structural problem neither company has to solve today but will eventually: there's no clean public comp set for either business model.
No Clean Public Comps
Existing public nuclear names are utilities and uranium producers, not factory-manufactured microreactor developers selling directly to data center operators. Existing public chip names -- Nvidia, AMD, and now Cerebras -- are built around GPU architectures and HBM-based memory, not the SRAM-based approach OLIX is betting will bypass that exact supply constraint. Public investors pricing either company at IPO will be doing genuinely novel valuation work, not applying an existing multiple to a new logo.
“Public investors pricing either company at IPO will be doing genuinely novel valuation work, not applying an existing multiple to a new logo.”
That comp vacuum matters because whichever of these two adjacent categories -- AI-driven nuclear power, or non-GPU AI chip architectures -- produces the first real IPO will effectively set the valuation framework the rest of the category gets measured against, the same dynamic playing out right now between Anthropic and OpenAI on the AI-lab side.
Both categories also require a genuinely different kind of investor patience than most AI application IPOs. Neither Valar nor OLIX will ship a commercial product for one to two more years at minimum, meaning any IPO conversation for either company remains purely theoretical for now -- but the capital intensity of both businesses means an eventual public listing, to fund the next stage of manufacturing scale-up, is a more likely outcome than a pure M&A exit.
What to watch: whether either company signals IPO intent within the next 12-18 months as manufacturing milestones firm up, and whether public investors show any early appetite for infrastructure-AI stories without an existing comp set to lean on.
