PALANTIR: War, Surveillance, Artificial Intelligence

Part 4: Digital Transformation of Enterprise and the Public Sector

Chapter 8. Business Model and Financial Structure

Attorney Kyungjin Kim

A. Balancing Government and Commercial Business

(1) Long-Term Contracts with the U.S. Department of Defense, Intelligence Agencies, and Allied Governments

On July 31, 2025, the U.S. Army issued a press release just two paragraphs long. But the number contained in that short document sent shockwaves through the entire defense industry. Ten billion dollars. A ten-year contract. Palantir Technologies.

The contract's official name was the Enterprise Service Agreement. Its core purpose was to consolidate 75 separate data and software contracts that the Army had been managing individually. Of those, 15 had Palantir as the prime contractor and 60 had it as a subcontractor. Years of fragmented agreements were being funneled into a single point of contact.

Alex Karp reportedly sat in silence for a long time in his office at the Denver headquarters when he heard the news. The company that had started 17 years earlier with $2 million from In-Q-Tel, the CIA's venture investment arm, had just won the largest defense contract ever awarded to a software company in American history.

There was a catch to the $10 billion figure. It was a ceiling, not a guaranteed amount. The Army had no obligation to spend the full sum. But the contract mattered for a different reason. Palantir had previously delivered its products through middleman resellers. Now those arrangements were converting to direct contracts, eliminating commissions and speeding up delivery. One thick layer of bureaucracy had been peeled away.

Government business had been the backbone of Palantir since its founding. As of Q3 2025, the government segment accounted for roughly 55% of total revenue. U.S. government revenue alone approached 80% of the total government figure. CIA, FBI, NSA, the Department of Defense, Army, Navy, Air Force. Nearly every American intelligence agency and military branch appeared on Palantir's client list.

Allied nations were no exception. Britain's GCHQ, France's DGSI, the Danish police, Australia's Department of Defence. Five Eyes allies and NATO member states analyzed intelligence and conducted operations on Palantir's platform. In December 2025, France's DGSI extended its contract with Palantir by another three years. A partnership spanning nearly a decade would continue.

But heavy dependence on government was a double-edged sword. Government budgets could be cut at any time depending on political decisions. When administrations changed, priorities shifted. By late 2025, a U.S. federal government shutdown had dragged into its fourth week, delaying approval of new contracts. Wall Street analysts worried about this concentration risk.

Karp was well aware of the concerns. So he decided to fire up a second engine.

(2) Commercial Expansion into Finance, Healthcare, and Manufacturing

In early 2024, during Palantir's earnings conference call, an analyst posed a sharp question. "Commercial segment growth, particularly outside the U.S., is slowing down. Could you explain what's happening?"

Karp's answer was brief. "We are focused on the United States. The U.S. is our most important market."

He wasn't bluffing. In Q3 2025, U.S. commercial revenue grew 121% year over year. $397 million. That was a 29% increase from the previous quarter as well. During the same period, international commercial revenue grew only 10%. In Q2 it had actually declined 3%. Palantir was deliberately concentrating its resources on the American market.

The number of commercial customers was surging. From roughly 169 in 2021, the customer count exceeded 1,500 by 2025. More important was the size of the deals. In Q3 2025 alone, 204 contracts exceeded $1 million, and 53 exceeded $10 million. The company was landing big fish.

The client roster was impressive. Airbus used Palantir's Foundry for aircraft maintenance and parts management under the name Skywise. BP adopted it for refinery operations. Merck brought it in for drug development data analysis. Lowe's, the largest home improvement retailer in the United States, used Palantir and Nvidia technology together to build a digital twin of its global supply chain network.

Growth in healthcare was also notable. Multiple hospitals and health systems across the U.S. began adopting Palantir's platform. AI was used to integrate patient data, optimize bed management, and improve staff allocation. The model proven at Britain's NHS had crossed the Atlantic.

In manufacturing, an operating system called Warp Speed drew attention. It was a solution that managed everything from factory production lines to supply chain management and inventory optimization on a single platform. Karp called it the software backbone of American re-industrialization. Finance was another important vertical. Investment banks, hedge funds, and insurance companies used Palantir's platform to analyze market data and manage risk. Ontology-based data integration also proved powerful for regulatory compliance work.

The rapid growth of the commercial segment sent an important message to investors. Palantir was no longer a government-only shop. It was a company running two engines, government and commercial. If one faltered, the other would hold. That was the business model Karp had designed.

As of Q3 2025, the revenue split between government and commercial was 55 to 45. That ratio had held steady for four consecutive quarters. The company had reached an equilibrium. But looking at growth rates alone, the commercial segment was dominant. U.S. commercial at 121% versus U.S. government at 52%. Forecasts emerged that a reversal might not be far off.

B. The Evolution of the Revenue Model

(1) The Mix of Licenses, Subscriptions, and Services

Palantir's early business model was straightforward. Deploy engineers to a client's site, integrate their data, and build a custom solution. These field-based engineers, called Forward Deployed Engineers (FDEs), were practically synonymous with Palantir. They lived in clients' offices, identified problems, and coded solutions.

The model was effective but had scalability limits. People are finite. Every new customer required hiring and training additional engineers. Revenue grew, but margins didn't improve. Wall Street hammered this point relentlessly. "Is Palantir a software company or a consulting firm?"

Karp agonized over this question for years. Then he found the answer. Apollo.

Apollo was a software deployment platform. It allowed Palantir's software to be installed and updated anywhere, from the cloud to a client's on-premise servers, even to edge environments on battlefields with no internet connection. Once Apollo arrived, the FDE role changed. Previously, engineers had to install and maintain software on site. Now Apollo automated much of that work.

The revenue model evolved too. In the early days, many contracts were one-off, project-based deals. Clients paid a large upfront build fee, followed by maintenance charges. But the company gradually shifted to a subscription model. Customers paid a fixed amount annually or monthly to use the platform. It was the standard SaaS approach.

The launch of AIP (Artificial Intelligence Platform) accelerated this transition. AIP was a platform that let enterprises apply generative AI in their operations. It controlled the hallucination problem of large language models through ontology-based data structures and embedded AI into real business processes. Customers paid additional fees on top of their existing Foundry or Gotham contracts to use AIP.

In Q3 2025, Palantir's revenue reached $1.18 billion. That was 63% growth year over year. Even more striking was profitability. Net income was $475.6 million, nearly triple the year-ago figure. A 40% net margin. Adjusted operating margin oscillated between 46% and 51%. Gross margin held between 81% and 84%.

There is a metric called the Rule of 40. It measures the health of a software company by adding revenue growth rate to profit margin; a score above 40 qualifies a company as high-quality. In Q2 2025, Palantir's Rule of 40 score was 94. By Q3 it had soared to 114. Numbers like these were virtually unprecedented in enterprise software.

Adjusted free cash flow was equally strong. In Q3 2025 alone, $540 million in cash flowed in. The trailing twelve-month total exceeded $2 billion. Annual guidance was between $1.9 billion and $2.1 billion. Palantir was no longer a cash-burning startup. It had become a cash-generating machine.

(2) Reaching a $300 Trillion Won Market Cap and Overtaking Legacy Defense Contractors

On November 3, 2025, Palantir's stock hit an all-time high of $207.52. Market capitalization surpassed $490 billion. Roughly 680 trillion Korean won. More than South Korea's entire annual national budget.

What the number meant was clear. Palantir had surpassed Lockheed Martin, Raytheon, and Northrop Grumman, the legacy defense giants. Of course, by revenue alone, there was no comparison. Lockheed Martin's annual revenue exceeded $70 billion.

Palantir's projected 2025 revenue was around $4.4 billion. A 16x gap. Yet the market assigned Palantir a higher valuation.

Why? The answer lay in growth rates. Lockheed Martin's revenue grew 5-6% per year. Palantir grew 50-60%. Investors were buying the future. They bet not on today's revenue but on tomorrow's potential.

Criticism of this rich valuation was fierce. In August 2025, Britain's The Economist called Palantir "the most overvalued company in history." Citron Research, the well-known short seller, calculated Palantir's fair value at $40 per share, less than 75% below the market price at the time. The stock plunged 25% right after Citron's report was published.

Valuation metrics were extreme. As of late 2025, Palantir's price-to-earnings ratio stood at 387x. Price-to-sales was 109x. Legacy defense firms traded at roughly 2x P/S, a gap of more than 50-fold. From a Warren Buffett-style value investing perspective, the numbers were impossible to justify.

Wall Street analysts were divided. Of 26 analysts, 17 rated the stock Hold, and 3 recommended Sell or Strong Sell. The average price target was $155, about 18% below the market price. Major investors were exiting. Legendary hedge fund manager Stanley Druckenmiller sold his entire Palantir position. Cathie Wood's ARK Invest also sold approximately 9 million shares.

Karp responded to the criticism in his characteristic way. In a letter to shareholders after the Q3 2025 earnings release, he wrote: "The critics are confused. We have delivered venture-capital-level returns to retail investors." It was the kind of sentence that invites charges of arrogance. But the numbers backed him up. Since its 2020 IPO, Palantir's market cap had grown from $15.6 billion to over $400 billion. A compound annual growth rate of 86%. That was only possible because of the fervent support of retail investors.

In September 2025, Goldman Sachs data showed that retail investors poured $1.2 billion into Palantir stock in a single month. They ignored the warnings of Wall Street analysts. To them, Palantir was not just a stock. It was a symbol of the AI era.

C. Global Partnership Strategy

(1) Partnerships with Nvidia and Microsoft

October 28, 2025. A convention center in Washington, D.C. On the stage at Nvidia's annual GTC technology conference, Jensen Huang addressed the audience.

"Palantir's Ontology is probably the most important enterprise software stack in the world."

That single sentence from the CEO of the world's largest AI semiconductor company sent ripples through the tech industry. It was not just praise. It was a signal announcing a new partnership between the two companies.

Here was the announcement: Nvidia's GPU-accelerated computing, its CUDA-X data science libraries, and its open-source AI model Nemotron would be integrated with Palantir's Ontology framework. This would allow enterprises and government agencies to build real-time decision intelligence systems.

In more technical detail, it worked like this.

Palantir's AIP (Artificial Intelligence Platform) gained access to Nvidia's Nemotron Super model, an inference model with 49 billion parameters. The 9-billion-parameter Nemotron Nano 2 and the 8-billion-parameter vision model Llama Nemotron Nano VL were also loaded onto the Palantir platform.

A decision-optimization tool called cuOpt was integrated as well. It solved complex problems like supply chain management and logistics optimization using AI. A technology called NeMo Retriever helped companies build AI agents grounded in ontology data.

The first major customer was Lowe's. Lowe's used the combined Palantir-Nvidia technology stack to build a digital twin of its global supply chain network. Where individual nodes had previously been optimized once a week, continuous and dynamic optimization at a global scale was now possible.

Karp said at the announcement: "We are fusing AI-powered decision intelligence systems with the world's best AI infrastructure." In December 2025, the collaboration expanded further. A project called Chain Reaction was announced. It was essentially an operating system to accelerate the buildout of America's AI infrastructure. The goal was to manage the complex supply chains spanning power generation, transmission, and data center construction using Palantir's AIP and Ontology together with Nvidia's AI models. CenterPoint Energy, a major Houston-based energy company, joined as a founding partner.

The collaboration with Microsoft also deepened. The core of a partnership announced in August 2024 was this: Palantir's full product suite, Foundry, Gotham, Apollo, and AIP, would be deployed on Microsoft Azure Government Cloud. This included both Secret and Top Secret classified cloud environments.

More significant was the integration with Azure OpenAI Service. Large language models including GPT-4 became usable within the Palantir platform, and in classified network environments at that. A path opened for analysts at the Department of Defense and intelligence agencies to use generative AI in their work with security guaranteed.

Palantir CTO Shyam Sankar described the partnership this way: "Bringing the combined capabilities of Palantir and Microsoft to national security organizations is a fundamental shift in how we support the defense and intelligence community."

The two partnerships, Nvidia and Microsoft, carried a clear message. Palantir was no longer fighting alone. It was joining hands with the key infrastructure companies of the AI era, positioning itself at the center of the ecosystem.

(2) Entering the Korean Market: Alliances with HD Hyundai, KT, and Samsung

The morning of October 14, 2025. KT's headquarters near Gwanghwamun, Seoul. Alex Karp walked into the building. His visit to Korea had been kept strictly confidential. It was not on any public schedule.

KT CEO Kim Young-shub greeted him. It was their first meeting on Korean soil since signing a strategic partnership in the U.S. in March. In the conference room, they reviewed the results of seven months of collaboration and discussed strategy for spreading Palantir's technology across Korean industry.

That same afternoon, KT hosted the AX Leader Summit. AX stood for AI Transformation. On the surface, it was a forum for sharing enterprise innovation strategies using AI. But the event's real purpose lay elsewhere.

Four executives at the chairman or vice-chairman level of major Korean conglomerates attended, each personally selected by Karp. Korean Air Vice Chairman Woo Ki-hong, Meritz Financial Group Vice Chairman Kim Yong-beom, LS Electric Chairman Koo Ja-kyun, and POSCO Holdings CEO Lee Ju-tae. Each had a 25-minute one-on-one meeting with Karp. Each session took place in a separate room, and attendees did not know the contents of each other's meetings. A person familiar with the matter said: "Every meeting was kept secret. The goal was to prevent each company's AI transformation strategy from being exposed."

This secretive approach was typical Palantir. They handled clients' data and strategies. Trust and confidentiality were central to the business.

The partnership with KT was already producing tangible results. At the time of the March signing, KT became the first Korean company to officially join Palantir's Worldwide Partner Ecosystem. Since then, KT had built its own cloud-based work environment and piloted Foundry and AIP in select business units. The results were good. KT now planned to offer the platform as a service to financial and public sector clients.

Karp spoke about the KT collaboration: "We value the partnership with KT highly. This cooperation is an important step toward secure cloud-based data use and industry-specific innovation."

Palantir's moves in Korea did not stop at KT.

In April 2024, a memorandum of understanding was signed with HD Hyundai Heavy Industries at Palantir's Washington, D.C. office. The two companies agreed to jointly develop an unmanned surface vessel (USV).

The USV was named Tenebris, Latin for "darkness." A mid-size vessel 17 meters long and weighing 14 tons, it targeted a speed of 50 knots and an operating range of over 1,000 nautical miles. It was designed to operate in rough seas up to Sea State 6.

Autonomous navigation technology developed by Avikus, an HD Hyundai subsidiary, would be combined with Palantir's AI-based mission autonomy system. The plan called for developing a reconnaissance USV first by 2026, then expanding to combat variants. In May, a scale model of Tenebris was unveiled at an AI expo in Washington. South Korea's Ambassador to the U.S., Cho Hyun-dong, visited Palantir's booth for a briefing.

HD Hyundai Heavy Industries CEO Joo Won-ho said of the collaboration: "The unmanned vessel market is a new blue ocean where advanced technology determines the winner. Building on the achievements and trust both companies have accumulated, we will pioneer this field." Cooperation with Samsung was also underway. Applications of Palantir's platform in supply chain management and manufacturing were being discussed. Partnerships with other Korean companies, including LIG Nex1, DL E&C, Kolon Benit, and Waiker, were expanding as well. Palantir's interest in the Korean market was serious. During his October visit, Karp told a reporter: "Korea is the most exciting market outside the United States."

A Palantir pop-up store even opened in Seoul's Seongsu-dong neighborhood. It was a marketing effort to raise brand awareness. A Silicon Valley defense-AI company running a pop-up store in one of Seoul's trendiest districts. Ten years ago, that scene would have been hard to imagine.

Korea was important to Palantir in several ways. World-class IT infrastructure, a corporate culture eager to use data, and a security environment defined by the North Korean threat. The conditions for Palantir's technology were in place in both the government and private sectors. The global rise of K-defense also broadened opportunities for collaboration.

Karp's Seoul visit lasted just one day. But the seeds he planted had the potential to reshape Korea's AI landscape. If major Korean conglomerates begin making decisions on Palantir's platform, it would mean not just adopting new software but a fundamental change in how they run their businesses.