Palantir Technologies delivered record-breaking results for its second fiscal quarter of 2026, powered by a staggering 149% surge in its U.S. commercial business and a precise bet on "sovereign AI." During its earnings call, the company launched a sharp critique of what it calls the current "burn game" in the large language model industry.

The company reported adjusted earnings of $0.41 per share for the quarter ended in early August, easily surpassing the $0.35 consensus estimate. Revenue soared 93% year-over-year to $1.94 billion, significantly beating the expected $1.8 billion.

On the call, Chief Financial Officer David Glazer raised the midpoint of the full-year 2026 revenue guidance to $8.154 billion. This represents an approximately 11-percentage-point increase from the previous quarter's forecast, marking the largest upward revision to annual guidance in the company's history.

Palantir's U.S. operations accounted for over 81% of total quarterly revenue, acting as the primary engine for the explosive performance. Chief Revenue Officer Ryan Taylor described the result as "unprecedented but entirely expected." Taylor stated that the commitment and urgency from clients to deepen their partnerships with Palantir is unlike anything he has ever witnessed.

The company declared the official arrival of the "Enterprise AI Sovereignty" era. The core logic behind this quarter's performance is the concept of helping enterprises control their own data, model weights, and business logic, rather than relying on large frontier model providers.

Ryan Taylor painted the current market landscape in stark terms, arguing that companies not using Palantir are burning money on valueless output while simultaneously handing their most important trade secrets to model providers, turning them into training data for future models.

U.S. Market Acts as the Strongest Engine, U.S. Government Business Continues to Expand

The central growth story for the quarter remains centered on the U.S. market. Financial metrics show that the U.S. business now accounts for over 81% of Palantir's total revenue, growing 115% year-over-year in the second quarter. Within that, the most closely watched U.S. commercial revenue growth accelerated further to 149% year-over-year, increasing 28% quarter-over-quarter to reach $764 million.

The U.S. government business generated $809 million in quarterly revenue, a 90% year-over-year increase, which the company described as an "extraordinary source of strength." Chief Technology Officer Shyam Sankar disclosed that the Maven intelligent warfare platform has completed its first government program of record go-live. This program selected Maven as its operating platform, leveraging its foundational capabilities like open data standards, Ontology, and developer toolchains.

Currently, over 25,000 developers are on the Maven platform, including active-duty military personnel, civilians, contractors, and enterprises. Notably, Sankar pointed out that the company's revenue from the U.S. Department of Defense over the past twelve months still represents less than 25 basis points of the Pentagon's budget, hinting at massive potential growth headroom.

On the international government side, revenue grew 42% year-over-year to $181 million. International commercial revenue increased 26% year-over-year to $182 million, showing significantly slower growth compared to the U.S. market.

Sharp Critique of the LLM 'Parasitic Model,' Enterprises Paying for Worthless 'Slop'

Addressing the persistently fervent AI model market, Palantir's management used the call to highlight the pain points and dilemmas many enterprises face in their AI adoption. Chief Revenue Officer Ryan Taylor pointed out that their second-quarter performance was unprecedented but completely expected, as the shift in the LLM market landscape they had warned about for years has now arrived.

Taylor emphasized that companies not using Palantir are watching their token meters spin endlessly, only to receive worthless "digital slop." He argued that the current model of charging purely by compute token might work for labs but fails for everyone else. This not only drains enterprise budgets but, more critically, Taylor stated that companies are paying to give away their most important secrets, which are the very foundation of their competitive advantage. As their secrets become training data for all future models, this will ultimately lead to the commoditization of their own businesses.

Palantir CEO Alex Karp also took aim at Silicon Valley's traditional software business models. He said the company rejects the "parasitic model" of tricking customers into giving them money to create dependency without delivering real value. Karp warned that the risk of solely relying on external frontier models is immense. He argued that people are paying for the "self-gratification" of tokens, but the real cost is allowing these providers to migrate your intellectual property, know-how, and expertise into their models. This enables them to build a competitive business that no longer needs your enterprise or your employees. Unprotected interaction with frontier models, he said, is extremely dangerous.

From Benchmaxing to Benchmaking, Turning Compute into Real Alpha

Having rejected the singular pursuit of large models, Palantir's prescribed antidote is "Sovereign AI," which gives enterprises complete ownership and control over the definition of their data, logic, actions, and security. Chief Technology Officer Shyam Sankar used a vivid analogy to summarize the current AI landscape, calling tokens the new coal and the AI Platform (AIP) the train. Now, he said, their customers can build their own locomotives.

Sankar pointed out that the market is creating far more "intelligence" than is being converted into actual "value," and more powerful models alone won't solve this problem. He declared that the era of blindly chasing general benchmarks is over. The age of "benchmaxing" has ended, and a new epoch of "benchmaking" has begun.

He revealed that in Palantir's testing, even a standard open-source model without fine-tuning, when run within Palantir's AIP architecture for a specific task, can outperform the most advanced frontier models. Sankar stated that a customer-specific benchmark is more than a scorecard; it is a mountain to climb that defines what "better" means based on your business strategy.

CEO Karp expressed extreme confidence in future growth. He presented an aggressive forecast, stating he is pushing the company to grow at or above the rate currently seen in its U.S. commercial business, which was 149% this quarter, over the next 18 months. He called this a very high target but one that is achievable because the company is fully aligned with the real needs of enterprises. For the first time, he said, people truly believe in them.