$175M Series B at a $1.8B post-money valuation, September 2025, led by Khosla and Lightspeed, taking total capital to about $202M — a 9x step-up on the prior mark (PR Newswire; Sacra).
Against an estimated ~$31M of ARR in 2025, up from $24M in 2024, on 111 employees (GetLatka) [WEAK]. That is roughly 58x — the highest multiple anywhere in the atlas register, and the reason this company matters more as a test case than as a business.
What it sells
Distyl describes itself as an "AI-native transformation consultant combining software subscriptions with high-touch professional services", signing multi-year enterprise contracts "reaching tens of millions of dollars over several years" (Sacra). It was founded by ex-Palantir people, and the model is recognisably Palantir's: put strong engineers inside the customer, make the platform fit the customer's actual workflow, leave software behind.
The distinction from the marketplace route in Forward-deployed engineering is deliberate. Toptal, Turing and Andela rent engineers and keep a spread — Turing's disclosed 50–55% markup on engineer cost, i.e. a 33–36% take. Distyl hired the people directly and wrapped them in software, trading near-term margin for a reusable asset. That is the "convert field work into IP" play, and it is the only version of this business with a route out of the labour multiple.
Why 58x is the argument, not the anomaly
The mark is the explicit expression of Sequoia's "Services: The New Software" thesis, published 5 March 2026 by Julien Bek (Sequoia). The argument: for every dollar spent on software, six are spent on services; management consulting alone is a $300–400B market; as AI crosses a competency threshold on rule-based "intelligence work", companies that sell the work itself beat companies that sell the tool.
The load-bearing assumption is a margin migration. Services businesses run 30–40% gross; software runs 70–85%. The thesis requires AI to compress delivery cost fast enough to move Distyl from the first band to the second before the multiple has to be defended. If it does, 58x is early. If it does not, 58x is a software price paid for a consultancy.
Palantir is the proven case for escaping the labour multiple — ~82% gross margin on a company built on forward-deployed engineers [UNVERIFIED — widely cited, not sourced in the atlas research notes]. The mechanism, not the margin, is the thing to copy. Palantir's FDEs are a cost of acquisition and installation for a renewing licence: they go in, make the platform fit, and leave behind software that bills every year with no engineer present. In a staffing or agency business the engineers are the product and revenue stops when they stop. Same job title, opposite unit economics. The only question worth asking about Distyl is what renews after the engineers leave — and it has never disclosed the split.
The comparables that should worry you
| Anchor | Multiple | Basis |
|---|---|---|
| Appen (ASX:APX) | ~0.9x revenue | Public, audited, 40.3% gross margin FY2025 |
| Innodata (NASDAQ:INOD) | ~6.2x revenue | Public, profitable, ~40% gross margin |
| Handshake | ~3.2x gross / ~7.8x net | Stale 2025 mark |
| Mercor at $10B | Contractors take 60–70% | |
| Distyl | ~58x ARR | Estimated ARR, undisclosed revenue mix |
Sources: stockanalysis.com for Appen; stockanalysis.com for Innodata; Sacra and Sacra for the private marks. See What the public market pays for labour and GMV is not revenue.
Everything above 6x is a bet that the private companies are structurally different from the public ones. Distyl is the largest such bet in the register.
Risks specific to this position
Buyer concentration. Multi-year, tens-of-millions contracts on ~$31M of ARR arithmetically means very few customers. No concentration figure has been published. Appen was a $4.3B company with 80% of revenue in five clients before Alphabet terminated a ~$83M contract and the shares fell 40% in a day. See One customer is a binary event.
The labs are the competition. AWS, Microsoft, OpenAI and Anthropic are building their own FDE armies — reported at ~$9B of committed implementation spend Jan–Sep 2025 (VC Cafe) [WEAK — no primary source]. They are simultaneously the budget's origin and its most credible substitute.
No exit comp supports the mark. The atlas found no strategic acquirer that pays a software price for a human-supply business. iMerit went to EXL for up to $310M; Segments.ai, Cleanlab, Captiv8, Whalar and WeCP all sold for sums nobody bothered to leak. The only deals paying big are licence-and-hire talent buys (Meta/Scale, Google/Mechanize), and those price ~35 people's judgment, not a services book.
Founding date, headquarters, actual revenue, the split between subscription and services revenue, gross margin, customer count and customer concentration. The ~$31M ARR is a GetLatka estimate, and the atlas's own spot-checks found Latka unreliable enough to list Surge AI — a company doing over $1B — at $330K of revenue and three employees. The 58x that makes this company interesting is a real valuation divided by an unreliable estimate. The subscription-versus-services split is the entire argument, and it is the one number nobody has.