Analysis
Four venture-backed companies repriced at two to five times their own prior valuation marks in the six weeks since early August, and only one of them is a foundation-model lab:
- Snorkel AI -- $3.5B Series E, 2.7x its own $1.3B Series D mark from 16 months ago. AI training-data infrastructure, not a model lab.
- TEKEVER -- $6.4B Series D, nearly 5x its own July mark. European defense-tech drone maker.
- Enveda Biosciences -- roughly $2B, doubled in under two years. AI-driven natural-compound drug discovery, lead program still Phase 1.
- Numeral -- $100M Series C, its second large round in about a year, taking total funding to $157.5M since a 2023 founding. Enterprise sales-tax compliance -- arguably the least "frontier" business in this set.
How Each Got There
Each of these has its own story. Snorkel AI pivoted from selling data-labeling software to delivering finished training datasets and reinforcement-learning environments directly to frontier labs, and annualized revenue grew 18x in a year to $375 million as a result -- the kind of growth number that justifies a markup on its own. TEKEVER's jump followed a GBP 400 million UK drone-supply deal and rides a broader European defense-tech buildout. Enveda's round drew a public-markets crossover investor in T. Rowe Price, a signal the round is being priced with an eventual IPO in mind. Numeral's growth is more mundane and, in some ways, more telling: total transaction volume grew 327% year over year, and the company expects its tax engine to process more than 80 million transactions, with California's 2027 expansion of sales-and-use tax to prewritten software handing it a regulatory tailwind it didn't have to create itself.
None of these four compete with each other, but they compete for the same investor attention against a common set of benchmarks. Snorkel's rivals -- Scale AI, Labelbox, Surge AI -- are all also raising or reportedly in talks to raise at rising marks. TEKEVER sits well behind Anduril's own $30 billion-plus valuation, underscoring how much room the European defense-tech gap may still have to close. Enveda competes for AI-drug-discovery credibility against Recursion Pharmaceuticals and Isomorphic Labs, neither of which has produced an FDA-approved AI-discovered drug either. And Numeral is still small next to Avalara and Vertex, the two publicly traded incumbents it's trying to underprice, or Anrok, the SaaS-focused rival chasing the same vertical with venture backing of its own.
The aggregate backdrop makes the pattern easier to read. Bessemer Venture Partners closed a large pair of new funds this month on the stated premise that AI-native companies are reaching $100 million in annual recurring revenue faster than any prior technology category -- a growth-speed argument that, on this week's evidence, is no longer confined to companies that train foundation models. Global startup funding already hit a record first-half total in 2026, surpassing all of 2025, according to Crunchbase News -- the macro tide every one of these individual markups is riding.
What The Multiples Miss
For GPs and LPs, the practical read is that "AI-adjacent" has stopped being a narrow label. A company doesn't need to train models to get priced like one is training them, as long as its growth curve looks similar -- which raises the diligence bar on separating genuine AI-driven growth acceleration from a valuation simply chasing the same multiple everyone else in the category is getting.
What the headline multiples miss is that none of these marks has been tested by a down cycle. TEKEVER's valuation leans heavily on wartime-urgency procurement that has historically proven vulnerable to peace negotiations and shifting political budgets; Enveda's lead program is still Phase 1, years from any regulatory decision; and Snorkel's 18x revenue growth, while real, compares against a small base that's easier to multiply fast than sustain. A repriced mark is a bet on a trajectory continuing, not proof that it will.
The number worth watching over the next two quarters isn't the size of the next markup -- it's whether any of these four, or others repricing on the same growth-speed logic, take a flat or down round instead. That would be the first real test of whether September's repricing wave reflects durable operating improvement or a narrative multiple that outran the underlying businesses.