Earnings Call Highlights
Palantir's Q2 2026 call was a triumphalist affair, with management projecting absolute confidence that the AI market has decisively shifted in their favor after years of warning about the risks of frontier model dependency. The dominant theme was 'sovereign AI' — the idea that enterprises must own their data, logic, weights, and security rather than cede them to frontier labs — and management framed Palantir as the only platform capable of delivering this. Analyst questions were largely confirmatory rather than probing, reflecting a sentiment that has moved from skepticism to capitulation, with the numbers making it difficult to push back. Karp and Sankar were unusually combative toward frontier labs, positioning them as extractive and parasitic, which signals Palantir is actively competing for the same enterprise AI budget that chip and infrastructure vendors depend on. The call's tone suggests Palantir believes it is entering a period of sustained hypergrowth and is actively recruiting partners to scale to meet demand it cannot yet fully capture alone.
- Palantir reported 93% year-over-year revenue growth — its highest ever — with U.S. commercial accelerating to 149% YoY and $2.132 billion in U.S. commercial TCV bookings in a single quarter, nearly $800 million above their prior record, signaling that enterprise AI spending is concentrating rapidly on application-layer platforms rather than raw compute.
- Shyam Sankar introduced the concept of 'benchmaking' over 'benchmaxing,' arguing that customer-specific benchmarks — not frontier model leaderboards — are the new competitive frontier; he disclosed that vanilla Nemotron Ultra beat frontier models on 5 production tasks within 24 hours, directly undermining the narrative that more powerful foundation models automatically win in enterprise deployment.
- Palantir described a head-to-head bake-off where a frontier lab and its deployment team failed to deliver value on a ticketing automation problem while Palantir's FDEs built agent swarms that converted into a $10 million ACV contract — a concrete anecdote that will resonate with any AI chip or infrastructure vendor whose customers are evaluating whether model capability or deployment infrastructure is the binding constraint.
- Karp explicitly stated he is driving the company to sustain U.S. commercial growth rates for the next 18 months and disclosed that Palantir is actively seeking technically capable partners to scale — framing this as a partnership model where occasional competition is acceptable — which opens a potential channel conversation for hardware and infrastructure vendors.
- The company disclosed it is entering the fine-tuning market in classified environments, with models fine-tuned on an NVIDIA stack inside customer security boundaries outperforming frontier models, and that customers are migrating from Foundry-only relationships to the full sovereign AI stack including Ontology — indicating that the AI infrastructure layer is moving deeper into customer operations and that compute partnerships are becoming strategically central.