As the Financial Times (FT) reported Sunday (Dec. 28), it’s part of an effort by these companies to protect themselves in case the boom in artificial intelligence (AI) investment fizzles next year.
The report, citing data from Pitchbook, shows that this year’s funding levels surpassed the previous record of $92 billion set in 2021, with companies like OpenAI and Anthropic attracting strong interest from investors.
Venture capitalists and industry experts told the FT that this funding will fuel growth while also offering companies a cushion in case there’s an investment downturn caused by worries about heavy spending on AI infrastructure.
“You should make hay while the sun is shining,” said Lucas Swisher, a partner at Coatue who has invested in OpenAI, Databricks and SpaceX. “2026 might bring something unexpected . . . when the market is providing the option, build a fortress balance sheet.”
Funding levels for 2025 have been elevated by some record rounds, the FT said: $41 billion for OpenAI, $13 billion for Anthropic, and Meta’s $14 billion bet on data-labelling start-up Scale AI.
Advertisement: Scroll to Continue
The report adds that cost pressures have led to more frequent funding rounds, particularly for companies working on “frontier” AI models that require massive amounts of computing power and costly chips.
OpenAI’s revenues for this year are about $13 billion, sources close to the company told the FT. However, the startup is losing billions of dollars each year as it develops its models, products and infrastructure.
In other AI news, PYMNTS spoke Monday with Adam Hiatt, vice president of fraud strategy at payments orchestration platform Spreedly, about the role of the technology in preventing fraud.
Although AI has rightfully gotten lots of attention for the part it has played in democratizing fraud, the report said, fraudsters don’t have a monopoly on artificial intelligence.
All the same, the advent of AI and its applications across popular and increasingly industrialized fraud schemes have compressed timelines and raised the level of abstraction at which humans must operate, that report added.
“Distinguishing between the good and the bad is turning into something that even good manual review isn’t able to do,” Hiatt said. “It used to be you could throw people at the problem, but that’s becoming tougher.”
