Meta is heavily investing in generative AI. The majority of the $72 billion capital expenditures planned for this year are allocated to data centers and servers. The deal shows how much AI companies will pay for data to train AI models.
Zuckerberg promised last year that his company would surpass rivals in 2025 with its models, but Meta’s latest release, Llama 4 has underperformed against independent reasoning and coding standards. Yann LeCun is the chief AI scientist for Meta. He said that the long-term goal has always been “to reach human intelligence and to go beyond it.” This was revealed at the VivaTech Conference in Paris, this week. Many AI companies are aiming to build artificial “general” intelligent technologies, which have the same level of intelligence as humans. A growing number of Silicon Valley companies are also aiming to achieve “superintelligence”a hypothetical scenario in which AI systems surpass human intellect.
Scale’s core business is data-labeling. This is a manual process that ensures images and texts are accurately labeled, classified and used to train AI models.
Wang is well-connected with Silicon Valley’s top investors and technologists including OpenAI’s Sam Altman. Scale AI’s first customers were autonomous vehicle manufacturers, but its $2 billion in revenue this year is expected to come from labeling data used to train massive AI models created by OpenAI and other companies.
This deal will bring a large payout to Scale’s venture capital investors including Accel, Tiger Global Management and Index Ventures. According to a source with knowledge of the situation, Tiger’s $200-million investment is now worth more than $1-billion at the new valuation. Tabby Kinder, San Francisco, contributed to this report
The Financial Times Ltd., 2025 (c),All rights reserved. This document may not be reproduced, copied or modified in any manner.
