Chamath Palihapitiya points out that many large U.S. companies struggle to exceed a capital cost of 8-11% in returns
Chamath Palihapitiya points out that many large U.S. companies struggle to exceed a capital cost of 8-11% in returns, while globally, one-seventh of companies have long-term returns of only 1-5%. In this environment, companies are massively opening up data, workflows, and processes to AI labs.
He warns that transmitting proprietary information such as prompts, customer data, and pricing logic through APIs is akin to handing over core competitive advantages.
Chamath cites Alex Karp's views and examples like Microsoft prohibiting employees from using Anthropic Claude Fable 5 due to data retention policies, emphasizing the importance of data sovereignty.
Source: Public Information
ABAB AI Insight
Chamath's remarks focus on the hidden costs and risks of corporate AI adoption. Under pressure for capital returns, large companies are eager to enhance efficiency but overlook the potential for data leakage that could be used by AI vendors for training or vertical competition.
In terms of capital pathways, companies turning to open-source weight models can significantly reduce costs (Deutsche Bank estimates a 65-fold difference between cutting-edge models and open-source). Chamath personally tested that the cost of open-source solutions for enterprise code migration tasks was only 1/16.4 of that of cutting-edge models, strategically encouraging self-control over data and computing power.
Similar to Palantir's emphasis on data sovereignty or warnings from Mistral's founders, the current deployment of corporate AI is shifting from reliance on external APIs to open-source and self-controlled infrastructure, driven by dual considerations of cost and risk.
Essentially, this is a data capital game, where cutting-edge closed-source models enhance their capabilities through corporate data, while open-source solutions allow companies to retain advantages. Capital is concentrating on companies that control data sovereignty and low-cost deployment.
ABAB News · Cognitive Law
Data leakage equates to loss of competitive advantage.
Open-source weight models offer far better cost-performance than cutting-edge closed-source models, with cost differences determining adoption pathways.
The premise for corporate AI dividends is that data control remains in-house.