Forward Deployed

Five of the world’s biggest software names — Microsoft, Google, Meta, Anthropic, and OpenAI — have spent the past 18 months converging on the same idea: build like Palantir. That means more than selling software into big organizations. It means sending technical people directly into the mess of how those organizations actually work, then turning that mess into something programmable.

Palantir’s bet, made 23 years ago, was that the most valuable data in the economy is not neatly sitting in one database. It’s spread across email, group chats, ERP systems, CRMs, spreadsheets, phone calls, and the heads of the people doing the work. The company built its business around making that institutional knowledge legible, then wrapping software around it. Investors long dismissed that as unglamorous data integration. The market has since stopped laughing: since ChatGPT launched in November 2022, Palantir’s market value has risen more than 20-fold, while its revenue growth has climbed for 12 straight quarters and its operating margin has moved from 2% to 47%.

The AI giants are now acting as if they need the same thing. In the last four months, they have deployed $30B of capital, acquired or backed four ex-Palantir-founded or staffed companies — Tomoro, Fractional, Northslope, and Thrive Holdings — and struck partnerships with large private equity firms including Blackstone, TPG, Bain Capital, and Thoma Bravo. They’re also hiring hard for more than 9,000 forward-deployed engineers, which the source says is roughly the headcount of two Palantirs. Microsoft has gone so far as to name its enterprise deployment platform Foundry, years after Palantir used the same name for its own flagship offering.

That is not a subtle homage. It is a confession. The labs know that raw model power is not enough if the real constraint is inside the customer: local knowledge, permissions, and the right to act on what the software learns. Palantir built a role for that job — the forward-deployed engineer — and made it part salesperson, part builder, part reality check. The labs seem to want the same machine, but without having to admit that their grand theory of scale still needs a very old idea: get close to the institution, or the institution will ignore you.