FULL: Palantir CEO Alex Karp Warns AI Will Redefine Power, War, and Economies at WEF | AI1G
- By
- DRM News
- Length
- 32 min
- Language
- English
- Text from
- the video's captions
- Transcribed
- 13 October 2026
Summary
At the World Economic Forum in Davos, BlackRock's Larry Fink interviews Palantir CEO Alex Karp about AI's impact on defense, business, and economies. Karp explains how battlefield-tested software translates to commercial value, argues that off-the-shelf large language models fail without an ontology layer, and predicts AI will expose the true market value of work and organizations. He warns that Europe has a serious structural tech adoption problem and that AI will reward societies and companies that can bear its load.
Key points
- Karp says Palantir's battlefield work reveals that half of an enterprise often exists only on PowerPoint and fails under real conditions.
- He uses Ukraine as an example of dynamic challenges like data obfuscation, jamming, and operating without connectivity.
- Battlefield lessons translate to commerce: companies should build capabilities no other enterprise has, like in insurance underwriting and hospital intake.
- Buying large language models off the shelf won't work because they aren't precise enough for regulated activities; an ontology layer is needed.
- Palantir barely has a salesforce because in a low-trust AI environment, products that actually work sell themselves.
- Karp believes AI will create more than enough jobs, especially for vocationally trained workers, and questions large-scale immigration without specialized skills.
- He says Europe's tech adoption is a serious structural problem and no political leader has honestly acknowledged it.
- AI acts as a load test: over the next three years it will expose the actual market value of what people and communities do.
Questions it answers
00:00Introduction
Why is Alex Karp being interviewed at Davos?
Larry Fink introduces Karp as a leader at the intersection of technology, national security, and the real economy, noting Palantir's 73% compounded return since going public. The session focuses on what AI means for growth, workers, countries, and national security.
- Fink highlights Palantir's 73% compounded return since IPO.
- The framing question is whether governments are prepared for AI's transformation of society.
02:17AI in defense
How is AI supporting decision-making in defense and security?
Karp explains that military technology historically raised living standards, and building software for harsh battlefield conditions reveals ground truth. A key role of enterprise software is making up for the fact that much of a nation's enterprise only works in a lab or on PowerPoint.
- Battlefield conditions are rough, morally gray, and often disconnected.
- Sovereign nations struggle to objectively rate which tech is better.
- America's battlefield experience let it see what worked and what didn't.
09:01Ukraine lessons
What does the Ukraine battlefield reveal about AI in war?
Karp walks through moving a drone from A to B: synchronizing data, obfuscating it, protecting assets like a general who is a spy, and adapting when Russians jam electronics. Each country also fights differently — Ukraine built proprietary uses on top of Palantir's product.
- Russians are mathematically strong and cobbled together electronic jamming.
- Ukraine's advantage was starting from nothing, so no false enterprise to rediscover.
- Israel reportedly fought intelligence-to-military, unlike typical military-to-military conflicts.
12:28Defense tech to business
How does battlefield technology translate to corporations?
Karp says the lessons are almost one-to-one translatable: enterprises should build capabilities no other enterprise has, like underwriting based on tribal knowledge or hospital intake processed 10-15 times faster. Processing LLMs in an ontology also bolsters civil liberties by showing granularly why someone was processed or rejected.
- Enterprises tend to converge and become like each other, which isn't valuable.
- Faster hospital intake saves lives and makes nurses and doctors happier.
- Companies no longer need to go private to restructure costs.
17:11Adoption barriers
What is hindering AI adoption?
Karp says buying large language models off the shelf won't work because they aren't precise enough for regulated activities. Value comes from building a software layer to orchestrate LLMs in a language the enterprise understands. Palantir barely has a salesforce because working products sell themselves in a low-trust environment.
- Off-the-shelf LLMs are a commodity and not precise enough for underwriting.
- Scarce resource: people with the highest clearance who are also technical.
- Deployment that took a year five years ago can now take a week.
22:48Jobs and aptitude
Will AI create or destroy jobs?
Karp says there will be more than enough jobs for citizens, especially vocationally trained technicians, and questions large-scale immigration without specialized skills. He argues for new ways of testing aptitude, citing a former police officer with a junior college education running high-end targeting for the US Army's Maven system.
- Vocational technicians can become rapidly more valuable, even irreplaceable.
- Many people doing X should be doing Y; aptitude tests miss outlier talent.
- Karp questions large-scale immigration absent specialized skills.
27:24Global divergence
Will AI widen the imbalance between developed and developing economies?
Karp says America and China both make AI work at scale and will accelerate beyond expectations, while Europe has a serious structural tech adoption problem no leader has named. AI acts as a load test: societies that can bear the load thrive, and over the next three years it will expose the actual market value of what people and communities do.
- Europe's tech adoption problem is structural and politically unacknowledged.
- The developing world will see pockets that do very well and very poorly.
- AI will deliver 'market value honesty' across communities within three years.
Notes
Context
- Larry Fink introduces Alex Karp at the World Economic Forum in Davos, noting Palantir's 73% compounded return since going public versus his own 21% at BlackRock.
- The discussion covers AI in defense, its translation to business, jobs, and global economic divergence.
AI in Defense
- Karp describes how military technology historically raised living standards through dual-use products, a link that broke until defense tech startups and Palantir revived it.
- Battlefield conditions are rough, morally gray, and often disconnected from networks, forcing software to work under constraints.
- A key Palantir role is compensating for the fact that much of an enterprise "exists on a PowerPoint" and fails in battle.
- Ukraine example: moving a drone from A to B requires synchronizing data, obfuscating it from adversaries, protecting assets, and then adapting when Russians jam electronics.
Defense to Commerce
- Militaries and intelligence services specialize; similarly, companies should build capabilities no other enterprise has.
- Examples: insurance underwriting based on tribal knowledge, and hospital intake processed 10-15 times faster, saving lives and improving nurse and doctor satisfaction.
- Processing LLMs within an ontology also bolsters civil liberties by showing granularly why someone was processed or rejected.
Adoption
- Buying large language models off the shelf fails; they are not precise enough for regulated work like underwriting.
- Palantir barely has a salesforce because delivered, working products sell themselves in a low-trust AI environment.
- Scarce resource: people with the highest security clearance who are also technical.
Jobs and Economies
- Karp predicts more than enough jobs for citizens, especially vocationally trained technicians, and questions large-scale immigration without specialized skills.
- He calls for new ways of testing aptitude, citing a former police officer running high-end targeting for the US Army's Maven system.
- Europe has a serious, structural tech adoption problem that no political leader has honestly named.
- AI "load tests" societies; over the next three years it will expose the actual market value of what people and communities do.
Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.