Palantir (PLTR): What This Company Actually Does

How Palantir makes money

Founded in 2003 to help U.S. intelligence agencies analyze counterterrorism data, Palantir now sells four core products — Gotham (defense/intelligence), Foundry (enterprise data operations), Apollo (deployment infrastructure), and AIP (generative AI) — to governments and large companies drowning in disconnected data. Customers pay ongoing usage fees under multi-year contracts, and existing customers tend to expand their usage sharply over time (a "land and expand" pattern), rather than most growth coming from brand-new logos.

R&D reinvestment (about $420M/year) continues to strengthen the software, feeding back into the same customer base. Source: 10-K FY2025, p.4 (business overview), p.6 (customer/segment revenue mix), p.68 (top-20-customer revenue).

Where the revenue comes from

Source: 10-K FY2025, p.113 (Note 13).

The U.S. share is dominant — currency exposure is limited, but reliance on U.S. government budget and policy decisions is correspondingly high. Source: 10-K FY2025, p.113 (Note 13).

Customers and competitors

A B2G (government) and B2B (large enterprise) mix with a broad customer base — no single customer accounted for 10% or more of total revenue in 2023, 2024, or 2025. The top 20 customers' average trailing-twelve-month revenue reached $93.9M, up 45% from $64.6M the prior year — a sign of fast usage expansion within existing large accounts, more than a widening customer base alone.

• C3.ai — Palantir's most frequently compared rival, but at roughly one-tenth the revenue, with much weaker government/defense trust and no equivalent to Palantir's Ontology data-integration layer.
• Snowflake / Databricks — strong in data warehousing and lakehouse infrastructure (storing and processing data), but don't build the decision-making applications on top that Palantir does.
• Booz Allen Hamilton / Leidos — long-established government-contracting and systems-integration firms, but built around people-intensive consulting rather than scalable software.

Note: Palantir's 10-K doesn't name competitors directly, describing them only as "internally developed solutions, large enterprise software vendors, and government systems integrators" (10-K FY2025, p.9). Source: 10-K FY2025, p.91 (customer concentration), p.68 (top-20 customers); web search — datagrom.com 2026 competitor analysis.

The metric that matters most in this sector

Total Remaining Deal Value (TRDV) — the sum of everything already contracted to be paid in the future — is a leading indicator. When it grows faster than current revenue, future revenue acceleration tends to follow.

TRDV's 105% jump in 2025 far outpaced revenue growth (+56%) that same year — a strong signal that revenue growth has more room to keep accelerating from contracts already signed.

Source: 10-K FY2025, p.69; 10-K FY2023, p.65–67.

Leadership and ownership

CEO Alexander Karp co-founded the company, joined the board in 2003, and has been CEO since 2005. All three co-founders — Karp, Stephen Cohen, and Peter Thiel — remain board members. A triple-class share structure (Class A: 1 vote, Class B: 10 votes, Class F: special voting rights) means economic ownership is low (Karp under 1%, Thiel 3.0%, Cohen under 1%) but voting power is concentrated (Karp 11.4%, Thiel 7.9%, Cohen 3.0%) — the founder group controls the company through a small economic stake.

Source: DEF 14A 2026, p.51–54 (ownership and voting tables), p.1 (titles).

Capital returns

Dividend: none, ever, with no stated plans to start. Buybacks: a $1 billion repurchase program was approved in August 2023, but only $75 million (600,000 shares) was actually repurchased in 2025 — and the entire program was terminated in January 2026.

Source: 10-K FY2025, p.65 (dividend/repurchase program), p.85 (shares outstanding), p.89 (stock-based compensation), p.103 (repurchase program detail).

How this company could fail

• Government dependence — 54% of revenue comes from government agencies. U.S. federal contract options generally can't be exercised more than a year in advance, making revenue harder to predict, and budget cuts or a change in political priorities could delay or cancel contracts.
• An extremely high valuation — as of August 2026, the company's roughly $414.7 billion market cap traded at about 68 times FY2025 revenue. Any slowdown in growth could hit the stock hard from that starting point.
• Dilution from stock-based compensation — 2025 SBC exceeded $684 million, more than 15% of revenue, and wasn't offset by buybacks (see above). If this pattern continues, per-share value keeps eroding even as the business grows.

Source: 10-K FY2025, p.12 (risk summary), p.69; web search — stockanalysis.com (market cap, Aug 2026); 10-K FY2025, p.89 (SBC).

Five-year financials

Revenue growth has accelerated every single year (23.6% → 16.7% → 28.8% → 56.2%). Operating income turned positive in 2023 and has improved sharply since, while free cash flow has grown even faster than operating income — a sign of high-quality earnings, not an accounting artifact.

Source: 10-K FY2025, p.86 (revenue/operating income), p.89 (cash flow); 10-K FY2023, p.86, p.89; 10-K FY2025, p.85 (balance sheet).

What we still don't know

• Management's specific full-year 2026 revenue guidance isn't captured in this card — the latest earnings call is the place to check.
• AIP's exact share of total revenue isn't disclosed separately.
• What growth rate would be needed to justify the current roughly 68x revenue multiple can't be calculated from this data alone — that's exactly what a reverse DCF is for.