Palantir: Worst nightmare for OpenAI, Anthropic?

Palantir is actively campaigning against foundation model companies, faux forward deployed engineers, tokenomics and benchmark chasing.

During Palantir's second quarter earnings call, CTO Shyam Sankar served as the warm-up act for CEO Alex Karp. The common theme is that foundation model giants, OpenAI and Anthropic, are bad for business.

Sure, Sankar were talking Palantir's game, but it's hard to argue with a Rule of 40 score of 155%. It's also worth noting that Palantir is echoing what enterprise CIOs have been saying: Tokenomics doesn't work and it's time to push back with open models, local AI deployments and home-grown models focused on domain-specific use cases.

Here are the developments we've been tracking for months and vendors are just starting to address. Palantir is front and center now.

Choice quotes from Sankar:

• "Tokens are the new coal and AIP is the train. Now our customers can build their own locomotives. AIP is where your sovereign AI is built, deployed, and compounded."
• "We're excited for a new era, not one of bench maxing, but of bench making. The assumption that the frontier is actually the best performing is just not borne out in practice. Within 24 hours of bringing (Nvidia) Nemotron Ultra into our stacks, we found five production tasks where a standard Nemotron Ultra model without post-training beat frontier models."
• "A customer-specific benchmark is not just a scorecard. It's a hill to climb. It's the normative orientation that directs your entire operational workflows and post-training pipeline. It defines what better means on your terms for your business with your strategy, and then everything optimizes against it continuously. And because the benchmark is yours, the trade-offs become yours too."
• "Only Palantir has FDEs; everyone else has sparkling sales engineers."
• “We've been beholden to a small number of benchmarks that people have been designing models to, and then releasing models, saying like, ‘Look how well it does on this benchmark.’ But the benchmark has actually almost nothing to do with your business. How do you figure out how to make the benchmark that represents your reality, what you're trying to succeed at, what you're trying to get better at, and then see what model makes sense?”

Karp built on his shareholder letter and targeted the AI industry. He said:

• "You are paying for the right for them to migrate your IP, your know-how, your expertise to their model, so that they can build a competitive business that doesn't require your business or people. And why are they doing it? It's actually being done for what they believe are moral reasons. They are superior to you. They deserve to colonize your enterprise. You deserve to be colonized."
• "The models actually fine-tuned by us in our enterprise on Nvidia stack outperform frontier models. And you own the links, you own the alpha, you own everything. Every single enterprise in this country is going to either look at doing this, find ways of doing it, or at least avoid the alternative of unprotected interaction with frontier models."
• "We cannot progress to a thin philosophical model where only a small group of people who think very differently than most of us actually absorb all of the revenue and value in our business and transfer all the dangers to us."
• "People understand that they need a way of controlling their alpha. They understand broadly that token maxxing is at their own cost, and they certainly understand that token maxxing is leading to them transferring their data, their prompts, the way they run their business, their expertise to a third party. They need education on what they can do about that with us, preferably, but also without us."

More

• Palantir, Nvidia bring Nemotron open models to sovereign AI
• Palantir's US commercial, government deals surge in Q1
• Palantir and Nvidia build sovereign, on-premises AI reference architecture