Analysis · Enterprise Software & IT

Palantir CEO criticises AI model data fees

By Startup Enthusiast · 15 July 2026

What they do

Palantir sells enterprise AI software for data analysis

CEO Alex Karp called token fees a 'wealth tax' on proprietary data

Open-weight models are gaining share due to cost and control advantages

Palantir CEO Alex Karp said enterprise CEOs are angry at AI model developers for collecting proprietary data via token fees.
He called the practice a 'wealth tax' on companies that provide the raw material for foundation models.
Palantir shares rose about 9% on the day of the CNBC interview.
The interview was tied to a partnership announcement; Palantir sells an alternative approach.
Key enterprise AI procurement questions include data ownership, cache location, weight control, and training data usage.
Customer proprietary data and process knowledge are raw material that improves foundation models.
Any AI product requiring customers to lose visibility into data handling has a ceiling on market it can win.

India's RBI has proposed the FREE-AI framework, preferring indigenous sector-specific AI models over generic third-party LLMs.
The EU AI Act becomes fully applicable on 2 August 2026, adding regulatory pressure.
European data residency rules are driving a measurable move toward self-hosting.
Open-weight models now account for over 45% of traffic on OpenRouter, up from under 2% a year ago.
Prosus reports inference cost reductions of up to 26x from running open models vs proprietary ones.

Palantir sells enterprise AI and data analytics software to governments and large corporations.
Its platforms help organisations integrate, manage, and analyse large datasets.
The company offers both cloud-based and on-premise deployment options.
Palantir generates revenue through software licenses and professional services.
It operates globally, with significant business in the US, Europe, and other markets.