PHASE 01 · 2003-2015
Military & Security
CIA · NSA · U.S. military counterterrorism intelligence analysis. Built irreplaceable intelligence analysis capabilities, and accumulated the foundational context layer through this vertical depth.
The first step to understanding Palantir is to drop the question of "whether it's an AI company."
It is a company that builds context infrastructure—AI is just a new component on top of this infrastructure. The four-layer architecture constitutes its strategic depth:
Intelligence analysis platform. Originally served the CIA, NSA, and U.S. military counterterrorism. Core capability is connecting data scattered across different systems into comprehensible context.
Enterprise data operating system. Democratizes Gotham's capabilities. The moat lies in deep API integration with customer business systems, industry Domain Models, and compliance knowledge bases.
Launched in 2023. Lets LLMs run within Palantir's existing context layer—models see not raw data, but industry knowledge graphs refined over two decades.
Continuous delivery platform. Solves the problem of continuous software updates in highly secure environments (military networks, air-gapped systems).
Models are replaceable (using Claude today, can switch to Gemini tomorrow),
but the context layer is irreplaceable.
Integrating LLMs and AI-based products is a
perfect use case for the role. — PRAGMATIC ENGINEER · ON FORWARD DEPLOYED ENGINEERS
Traditional enterprise AI deployment has two extremes: pure product (standardized SaaS, no customization), or pure consulting (Accenture-style, every project from scratch). The former lacks depth of engagement; the latter has uncontrollable costs. Palantir's FDSE model takes a third path:
FDSEs are engineers embedded at customer sites, with both technical implementation skills and industry understanding. The key distinction: all customization work is based on Palantir's standard tool stack—FDSEs don't write new code; they assemble solutions customers need using standard components from Foundry/AIP.
This means: deliverables from every customer project accumulate as part of the platform's capabilities. A supply chain optimization solution built by an FDSE at an oil company can, after abstraction and productization, be reused for other energy customers. Internal tool productization, not external selling—this is the key to Palantir's POC cycle of just 1 month (vs. C3.ai's 4-5 months).
This model is being replicated across the industry. OpenAI has hired many FDSE roles, driven by its former Chief Research Officer Bob McGrew (who came from Palantir). Companies like Ramp and Commure are also following suit.
CIA · NSA · U.S. military counterterrorism intelligence analysis. Built irreplaceable intelligence analysis capabilities, and accumulated the foundational context layer through this vertical depth.
Brought military-grade data integration into industrial supply chains. Energy companies use it to optimize supply chains; manufacturers use it to predict equipment failures.
After AIP's launch, rapidly reached a broad range of commercial customers. Stock price surged from $15 to $111 (7×), and rose 40% against the trend during the DeepSeek valuation shock.
The strategy of "penetrate vertically, replicate horizontally" is mutually validated by C3.ai (energy → manufacturing/military → finance), and is the most worthy expansion paradigm for B2B AI companies to reference.
In March 2026, Palantir made headlines again in a disturbing way: U.S. Central Command used Claude through Palantir's Maven Smart System to assist in identifying strike targets during military operations against Iran—over 1,000 targets in the first 24 hours.
Palantir's CEO publicly discussed this involvement, claiming to execute a "higher mission," backed by 430% annual growth and a $360 billion valuation. External commentary was sharp but accurate: the modern military-industrial complex builds algorithms, not tanks.
Meanwhile, New York hospitals dropped Palantir, and the UK raised privacy concerns about it—a reminder that:
The power of the Context Layer is a double-edged sword. The same data integration capability,
applied to enterprise supply chains, is an efficiency tool;
applied to military intelligence, is a targeting system;
applied to law enforcement, is a surveillance infrastructure.
The deeper the moat, the heavier the ethical responsibility. This is not a problem that can be solved with "usage policy documents."
Model capability is not the moat; the context layer is. Core investment should go toward accumulating industry Domain Models, API integrations, and compliance knowledge bases—not chasing the latest foundation models.
The FDSE model is worth serious consideration. But all customization must be based on a standardized tool stack, otherwise you'll fall into the cost trap of pure consulting.
"Vertical first, then horizontal" is a proven expansion path. Find a high-barrier vertical domain to penetrate, build an irreplaceable context layer, then replicate horizontally—rather than pursuing a general-purpose platform from the start.
Ethical frameworks need to be thought through in advance. When your context layer is powerful enough, customers (including military and government) will find ways to use your capabilities, and the boundaries you can control are far narrower than you imagine.