Forward Deployed
Short summary
Major AI companies have spent $30 billion in the past 18 months adopting Palantir's business model—deploying engineers directly into institutions to integrate data and turn AI into practical value—but they're copying the structure without understanding the core insight: making powerful computation work inside institutions while preserving their autonomy and incentive to share knowledge.
- Palantir's stock has outperformed every major US company since ChatGPT's launch (20x gain), with revenue growth accelerating to 93% year-over-year and operating margins expanding to 47%, signaling the market's validation of its institutional-focused approach over pure model development.
- The frontier labs (Microsoft, Google, Meta, OpenAI, Anthropic) have independently converged on Palantir's strategy in the last 18 months: hiring 9,000+ forward-deployed engineers, acquiring ex-Palantir companies, and launching enterprise AI platforms—revealing what they actually believe is missing from their AGI ambitions.
- Palantir's real innovation wasn't AI deployment but solving a deeper problem: accessing the operational knowledge trapped in institutions (email, spreadsheets, experienced workers' heads) and making it programmable without destroying institutional sovereignty or incentive to cooperate.
Full summary
Palantir has spent 23 years building something the tech industry largely dismissed as boring—software that integrates fragmented data across institutions—while the Frontier Labs (Microsoft, Google, Meta, Anthropic, OpenAI) chased general AI models. The bet paid off spectacularly. Since ChatGPT launched in November 2022, Palantir's stock rose 20x, crushing Nvidia (13x) and other AI beneficiaries. More striking: the company's revenue growth accelerated for 12 straight quarters, reaching 93% year-over-year by Q2 2026, while operating margins jumped from 2% to 47%. Palantir recognized that the world's most valuable knowledge—how SpaceX achieves $200/kg launch costs, how Airbus coordinates four million parts—sits trapped in email, spreadsheets, phone calls, and workers' heads. Unlike GitHub or the internet, this institutional knowledge won't digitize itself. Someone has to painstakingly extract it, version it, and make it computable.
The Frontier Labs now realize they've been betting on the wrong thing. Raw model scale matters less when you can't access the specific context that makes those models useful inside real organizations. So they're copying Palantir's entire playbook. In the past four months, these five companies deployed $30 billion, acquired four ex-Palantir companies (Tomoro, Fractional, Northslope, Thrive Holdings), partnered with mega PE firms, and launched hiring sprees for 9,000+ Forward-Deployed Engineers—Palantir's signature role. The FDE is someone who understands a customer's actual problem, proposes a real solution, and writes the code to connect their fragmented systems. Microsoft even named its enterprise AI platform "Foundry," copying Palantir's flagship product from a decade ago.
The question now is whether you can actually copy this model or if it's more like a constitution—something that only works in the specific context where it evolved. The Frontier Labs have the capital and talent to hire thousands of FDEs, but institutional trust and deep operational knowledge took Palantir 23 years to build. It's unclear whether that scales by throwing money at it. The broader implication: if general intelligence becomes cheap and abundant through frontier models, what becomes scarce is permission to use it inside institutions and the judgment about when to use it. Palantir understood this from day one. Everyone else is scrambling to catch up.