For the past two years, the AI race has largely been framed as a battle between foundation model makers like OpenAI and Anthropic. But an interview published Thursday suggests Palantir Technologies Inc. (PLTR) CEO Alex Karp believes the industry's next competitive moat may lie somewhere else entirely.
Karp isn't trying to build the next frontier AI model. Instead, he's making the case that the most valuable part of the AI stack could ultimately sit above it—a software layer that lets enterprises switch between models without giving up control of their data, workflows or intellectual property.
Speaking to The Information after his CNBC appearance, Karp said businesses are becoming increasingly concerned that relying too heavily on proprietary AI providers could leave them vulnerable if those companies optimize models using customer insights or eventually compete against them.
“There's just very deep frustration around…are they gonna optimize the models for me, or are they gonna take the alpha of my business, transfer in their weights, and compete against me?” Karp said.
Palantir Bets on the AI Application Layer
That philosophy is increasingly shaping Palantir's AI strategy.
Earlier this week, the company launched a platform designed to help U.S. government agencies securely deploy and customize Nvidia Corp (NVDA)'s open-source Nemotron models through Palantir's software.
Karp also told The Information that some U.S. government customers had recently switched from proprietary AI models developed by companies such as Anthropic to Nvidia's open-source alternatives, although he declined to identify the agencies involved.
Rather than persuading customers to commit to a single AI model, Palantir is positioning itself as the software layer that manages whichever model an enterprise chooses. Its Evolve platform already routes workloads across multiple AI models based on customer priorities such as performance, cost or security.
Why the AI Moat Could Be Shifting
The strategy reflects a broader shift emerging across enterprise AI.
As more open-source models reach competitive performance, businesses are increasingly looking for flexibility rather than vendor lock-in. If enterprises can switch between OpenAI, Anthropic, Nvidia's Nemotron and future models without disrupting their applications, the value may increasingly reside in the software that orchestrates those models instead of the models themselves.
That doesn't necessarily diminish the importance of OpenAI or Anthropic, whose proprietary models continue to lead many industry benchmarks. But it does suggest that enterprise customers may ultimately place a higher premium on governance, security and interoperability than exclusive access to any one model.
What Investors Should Watch
For investors, Karp's comments point to a broader debate unfolding across enterprise AI: whether long-term pricing power will remain with foundation model developers or migrate to the companies helping businesses manage them.
If that thesis plays out, companies building the enterprise AI orchestration layer—including Palantir, alongside platforms such as Snowflake Inc. (SNOW) and Databricks—could become some of the biggest beneficiaries of AI adoption, regardless of which foundation model ultimately leads the benchmarks.
