Public transcript

CNBC's interview with Palantir CEO Alex Karp July 1st 2026 (1080p HD)

By: Global Secrets
Length: 20 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Palantir CEO Alex Karp sits down with CNBC to discuss the company's new NVIDIA partnership, explaining that the deal grew out of enterprise distrust of frontier labs over data control and IP. Karp argues that value comes from compute plus model plus Palantir's Ontology application layer, criticizes AI overselling, and denies the U.S. is in a doomed AI bubble. He also weighs in on China as a peer adversary, his support for Israel, the Iran situation, and briefly, President Trump's remarks.

Key points

Karp says the NVIDIA partnership arose because clients distrust frontier labs over who controls models, weights, and their business's alpha.
Palantir's Ontology application layer makes large language models safe, useful, and precise without touching or caching underlying data.
Karp claims enterprises are 'livid' about paying for tokens that create no value while risking transfer of their IP.
He argues the profitable parts of AI are the application layer and compute, and Palantir can match frontier model performance with open models clients control.
Karp says Palantir has more demand than it can supply and sees 15 to 18 billion dollars in free cash flow two years out.
He identifies America, China, and Israel as the world's relevant tech centers and calls China a peer-to-peer adversary.
Karp describes himself as the most publicly supportive CEO of Israel while also being an effective private critic.
He believes Iran has been degraded and says there are comforting things not in the public space.
The interview ends with Karp being asked about President Trump's remarks at Joint Base Andrews.
chapters_placeholder

Questions it answers

00:00NVIDIA deal origins

Why did Palantir partner with NVIDIA?

Karp says the deal came together because clients across critical infrastructure in America, Ukraine, and Israel distrust frontier labs over who controls models, weights, and business value. The partnership aligns around giving technical customers control over their compute, models, data stack, and alpha.

Everyone using LLMs on the battlefield runs on Palantir's ontology.
Clients want to own the means of production rather than transfer alpha to a third party.

01:20What Ontology does

What is Palantir's Ontology and why does it matter?

Karp describes Ontology as the application layer that takes a large language model and makes it safe, useful, and precise. It prevents the model from touching underlying data, caching data, replicating the business, or transferring IP.

Ontology is now being copied by others.
Critical infrastructure does not run frontier models without an application layer, he says.

04:55Open models and true cost

Can Palantir's approach compete with frontier models?

Karp claims the value comes from model plus application layer plus compute, and that Palantir can get an open model to frontier-level performance while clients control the weights. He argues the true cost includes what a business loses by transferring its alpha.

The profitable parts of AI are the application layer and compute.
Palantir sells an agnostic product that lets customers switch models.

07:01AI bubble debate

Is the AI market a bubble about to burst?

Karp says no: compute plus ontology plus model is genuinely changing the course of history, citing Ukraine, Israel, and the Department of War. But he warns the technology has been irresponsibly oversold, which risks backlash such as a wealth tax.

He says the U.S. does not have to overhype what it has.
Most of it is being built in America, except open models from China and NVIDIA.

09:06Palantir's business

How is Palantir's business performing?

Karp says Palantir has more demand than it can supply and that its financials show 15 to 18 billion dollars in free cash flow two years out. He criticizes investors for judging enterprise value by what they personally use.

He dismisses short-term market fickleness after 20 years in the business.
Enterprise use cases like manufacturing and special forces differ from consumer use.

13:22China and tech centers

How does Karp view the AI race with China?

He names America, China, and Israel as the world's relevant tech centers and calls China a peer-to-peer adversary. He says it is not a foregone conclusion that America wins, and that the U.S. debates AI as if it had no adversaries.

China lacks the deep tech creativity of the U.S. but has other advantages.
He argues the outcome is binary: they win or we win.

15:35Israel and Iran

What does Karp say about Israel and Iran?

Karp calls himself the most publicly supportive CEO of Israel while also being its most effective private critic. He says Iran has been degraded and that there are things not in the public space that would be comforting to people.

He says half of Israel's critics don't believe it should exist.
America and Israel will never be fully aligned, and that's okay.

18:55Trump remarks

What happens at the end of the interview?

As the interview wraps with banter about divergent opinions, the hosts note that President Trump was speaking to reporters at Joint Base Andrews before flying on the Qatari-gifted Air Force One, addressing his financial disclosures and saying everyone is profiting because the stock market is up.

Trump was en route to the opening of Teddy Roosevelt's presidential library in North Dakota.

Notes

The NVIDIA Partnership

Karp frames the deal as a response to enterprise distrust of frontier labs over who controls models, weights, and business value.
Palantir sits on critical infrastructure across America, Ukraine, and Israel; everyone using LLMs on the battlefield runs on top of its ontology.

Ontology and the Application Layer

Karp describes Ontology as the application layer that makes large language models safe, useful, and precise.
It prevents models from touching underlying data, caching data, replicating a business, or transferring IP in classified or clinical contexts.
He notes everyone is now copying the ontology concept.

Enterprise Frustration with Frontier Labs

Karp says enterprises feel they are paying for tokens that create no value while their alpha and IP are transferred elsewhere.
He argues that if a model were truly valuable, vendors would take a percentage of value created rather than charging for tokens.
He calls on viewers to call CEOs privately to verify how 'livid' enterprises are.

The Value Stack

Karp claims the real formula is model plus application layer plus compute, and the profitable pieces are Ontology and compute.
He says Palantir can bring an open model to frontier-level performance in classified or non-classified contexts while letting clients control the weights.
Palantir is agnostic and sells a product allowing customers to switch from model to model.

AI Bubble Question

Karp rejects overselling: compute plus ontology plus model is genuinely changing the course of history, citing Ukraine, Israel, and the Department of War.
He warns that overhype risks policy backlash like a wealth tax that punishes rather than helps.

Financials and Demand

Karp says Palantir has more demand than it can supply and points to 15 to 18 billion dollars in free cash flow visible two years out.
He criticizes investors for assuming what they personally use is what is valuable to enterprises.

Geopolitics

Karp names America, China, and Israel as the world's relevant tech centers and calls China a peer-to-peer adversary.
He describes himself as the most publicly supportive CEO of Israel and an effective private critic, and says Iran has been degraded.
The interview closes with discussion of President Trump speaking at Joint Base Andrews about financial disclosures.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.