Analysis
Three Data Points, One Pattern
Three separate valuation jumps this week point to the same underlying shift, as CNBC's reporting on Databricks and CNBC's reporting on Anthropic's IPO meetings both illustrate: the interval between 'primary' valuation marks for the biggest AI companies is collapsing from years to months, and in Anthropic's case, to no formal round at all.
Two companies show the compression clearly:
“That's a valuation more than doubling with zero transaction attached to confirm it.”
- Databricks — closed its latest $5 billion round at $190 billion, up from $134 billion roughly six months earlier (a 42% increase), on $7 billion of run-rate revenue and 12 straight months of positive adjusted cash flow
- Cognition — the AI coding startup behind Devin is reportedly in talks to jump from the $26 billion valuation it set in May to as much as $40 billion now (a 54% increase in roughly three months), on an annualized revenue run rate nearing $1 billion, about double what it had at its last raise
Anthropic: No Round at All
Anthropic is the most extreme case, because there's no priced round marking the jump at all. The company's last private valuation was reported around $965 billion; the $2 trillion figure now circulating ahead of an October IPO comes purely from investors extrapolating off Anthropic's disclosed $47 billion run-rate revenue, not from a new financing event. That's a valuation more than doubling with zero transaction attached to confirm it.
What's Driving the Compression
The mechanism driving this compression is straightforward: too much capital chasing too few companies with defensible AI revenue growth. Growth-equity funds, sovereign wealth vehicles like MGX, and crossover public investors are all competing for the same limited allocation in Databricks, Cognition, Anthropic, OpenAI and a handful of others, and that competition is what's pushing valuations up faster than fundamentals alone would justify -- similar to what's driving the dual-valuation tranche structures now showing up in deals like Baseten's.
For founders outside this small circle of AI infrastructure leaders, the practical effect is a harder fundraising environment, not an easier one -- capital concentration at the top means less attention and less follow-on interest for companies just below the frontier tier. For LPs, the risk is straightforward: fund marks based on the most recent round are increasingly stale within a single quarter, and any fund holding Databricks, Cognition or Anthropic-adjacent positions should be modeling a wider error band on NAV than the reported numbers suggest.
How Unusual This Is
Historical comparison helps size how unusual this is. In the 2018-2021 late-stage cycle, valuation step-ups of 40-50% between rounds typically took at least a full year, often two, and were themselves considered aggressive by growth-investing standards. Compressing that same step-up into three to six months, as Cognition and Databricks have both done this year, implies either genuinely extraordinary revenue growth -- which in Databricks' case is at least partially verified by its disclosed 80%-plus year-over-year growth -- or investor pricing that's running ahead of fundamentals on the assumption that today's growth rate is durable rather than a temporary AI-adoption spike.
Fact vs. Inference
The distinction between fact and inference matters here: Databricks' revenue growth and cash-flow profile are disclosed and verifiable; Cognition's reported valuation talks and Anthropic's investor-modeled $2 trillion figure are both still unconfirmed by the companies themselves. Treating all three as equally solid data points would be a mistake -- one is a closed transaction with disclosed financials, the other two are reported talks and investor math, respectively.
Watch whether any of these three companies actually closes a formal down-round or flat round in the next 12 months -- that would be the clearest signal the repricing sprint has hit its ceiling.
