Frontier model labs
Soft landing: Lab revenue has more than doubled a year so far, and GPU rents sit near the ~$6.4 break-even.
Hard landing: The labs file nothing, so the course's central number is the one no filing can check.
OpenFoundryPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Operator.ioPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
OsmosisPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
Overstand LabsPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
OwayPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PanelyticPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PanoraPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PapayaPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
ParasmaPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PipeshiftPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PiramidalPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PlaygentPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
PraximPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
ReductoPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
ReflexPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RentFlowPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RevylPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RiftenPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RindlerPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RunAnywherePrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
RunLocal AIPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
SalusPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
SanifuPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
SciloopPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
SecodaPrivate · files nothingAI infrastructure, data & models | — | — | — | 95% | — | — | — | 40% | — | — | — | — |
A private company, and every frontier lab, files nothing, and carries only its mark. Exposure to the build-out is the desk’s estimate, not a disclosure, and so is every figure built on it (build-out revenue and debt). A figure marked * is the course’s typical figure rather than an estimate of this company’s own. Colours are rules of thumb, not ratings.
Course economics and each course’s test are the desk’s, from NBER Working Paper 35865 as drawn in the stack on the indices dashboard; they are schematics of a representative buildout, not a measurement of any one project. A company’s course is read from the layer and role the AI 100 files it under — several companies span more than one course, and each is placed where the bulk of its business sits. Value is market capitalisation for a listed company and the latest primary or secondary mark for a private one. Revenue, debt, cash, EBIT, EBITDA, interest, operating cash flow and capex are filed figures over the trailing twelve months, in US dollars. Exposure to the build-out is the desk’s estimate and nothing more: no filing breaks revenue down by “sold into a datacenter” or tags a bond as raised against one, so revenue exposure is read from segment disclosure and customer concentration, and debt exposure from what the borrowing was raised to pay for. A figure marked * is the typical figure for the course rather than an estimate of that company’s own. Not investment advice.