Palantir Wants Enterprises to Stop Renting AI

Palantir's second-quarter results left little doubt that demand for its AI platform is accelerating. Revenue reached $1.935 billion, U.S. commercial revenue surged 149% year over year, and the company raised its full-year guidance to more than $8.15 billion. The company said it delivered its fastest revenue growth on record. 

Yet executives spent surprisingly little time talking about revenue. Much of the discussion focused on why enterprises are rethinking who should own the AI running inside their business.

Much of Palantir's argument centers on ownership. Executives said customers are paying closer attention to where their models run, who controls the underlying infrastructure, and whether they can move between models without rebuilding their AI applications.

"Sovereign AI" Becomes a Business Strategy

Executives returned repeatedly to one topic: AI sovereignty.

Chief Revenue Officer and Chief Legal Officer Ryan Taylor described it as giving enterprises ownership of "the data, logic, actions, and security of their enterprise." Chief Technology Officer Shyam Sankar said businesses should evaluate AI against their own operational goals instead of relying on generic public benchmarks. 

The company recently expanded that strategy by introducing infrastructure for deploying NVIDIA Nemotron open models inside sovereign environments. According to Palantir, customers can switch between AI models, fine-tune them, and operate them inside their own security boundaries without becoming tied to a single provider.

It's also a familiar argument. Palantir has spent much of the past year telling investors that cheaper inference should make enterprise software more valuable, not less.

Chief Executive Officer Alex Karp said customers increasingly want the flexibility to replace AI models without rebuilding the applications around them. That, he argued, reduces dependence on any single AI vendor. 

Beyond Benchmarks, Toward Production AI

Sankar also questioned how the industry measures AI performance.

He described the idea as "benchmaking," using customer-specific evaluations instead of relying on standardized public benchmarks. 

As an example, Sankar said Palantir's internal testing found NVIDIA's Nemotron Ultra outperformed larger commercial models on several real-world workloads despite ranking lower on widely cited benchmarks. He said enterprises should evaluate models against business outcomes instead of public benchmark scores. 

That idea also appears throughout Palantir's Artificial Intelligence Platform (AIP).

Sankar described AIP as an orchestration layer that combines data integration, Ontology, security controls, workflows, evaluation pipelines, and post-training capabilities into a single environment designed to turn AI into measurable business outcomes. 

One recent example is Cleveland-Cliffs, which is using Palantir to integrate AI into manufacturing operations. The steelmaker is deploying AI systems inside production workflows instead of limiting them to pilot projects.

Customers Are Buying More Than Software

The company's largest deals reinforced the point.

Palantir closed 220 deals worth at least $1 million during the quarter, including 98 deals above $5 million and 73 deals above $10 million. U.S. commercial total contract value reached $2.132 billion, while net dollar retention climbed to 157%. Revenue from its top 20 customers increased 67% year over year. 

The company highlighted several deployments that expanded from initial pilots into enterprise-wide agreements. One multinational technology company grew into a nearly $370 million contract spanning multiple business units. A nonprofit health system converted an initial pilot into a three-year partnership. 

Executives credited much of that expansion to Palantir's long-standing practice of embedding forward-deployed engineers alongside customers during deployments. Throughout the earnings call, they said implementation and workflow integration matter as much as model performance. 

Microsoft has also expanded the use of engineering teams working directly inside customer organizations as enterprise AI projects become larger and more operationally complex.

Palantir also raised its full-year revenue outlook to between $8.15 billion and $8.158 billion while increasing its U.S. commercial revenue guidance to more than $3.424 billion. The company expects adjusted free cash flow of $4.5 billion to $4.7 billion and said it will continue investing in technical hiring, AI platform development, and sovereign AI capabilities.