I spent some time this week discussing with others about the recent private equity partnership announcements from OpenAI and Anthropic, and what it means for AI adoption. [1]
I think the most interesting question is not just distribution or increasing AI usage. It’s whether PE-backed companies become one of the first real-world test cases for AI adoption.
PE investors are disciplined underwriters. They typically need to see a path to measurable value creation within a 5 – 7 year hold period.
After an acquisition, the PE value creation playbook is to invest in three areas: (1) people, (2) processes, and (3) software. AI has the potential to affect all of these areas:
People: higher productivity per employee, especially when paired with high-agency teams willing to adopt new tools
Processes: faster, more automated workflows with fewer manual handoffs
Software: new systems of record and execution that don’t just track work, but increasingly perform it
It’s unclear based on the public press releases how the money is being spent in the PE + AI partnerships, but assuming the portfolio companies bear the operating costs, it will result in a real-world test case of ROI on AI adoption. That is where token pricing becomes strategically important.
When Krishna Rao, Anthropic CFO, was asked by Patrick O’Shaughnessy why Anthropic doesn’t raise its token pricing when demand is so high, I thought his reply was spot on and probably overlooked by many. [2]
(As a sidenote, this is something that Bill Gurley has spoken about at length. In a properly functioning marketplace, you raise prices to meet demand when supply is constrained. Makes total sense, otherwise someone — aka VCs — is subsidizing the demand.) [3]
However, Krishna responded that by keeping token prices low, it leads to Jevons Paradox, where usage actually increases as new use cases emerge, opening up new TAMs that would otherwise be locked out. In other words, as an AI lab, low token unit pricing can expand the addressable market, while usage-based pricing still allows revenue to scale with consumption. For PE investors, that means AI adoption may not simply be a cost-reduction story; it may also increase usage-driven software and compute spend where the ROI comes from productivity gains, workflow automation, and new revenue opportunities.
Over the next 12-24 months, as more compute comes online, and the PE + AI partnerships unfold, we’ll see whether the market is simply tokenmaxxing or whether real-world value is actually being created.
Notes
[1] OpenAI’s press release: https://openai.com/index/openai-launches-the-deployment-company/
Anthropic’s press release: https://www.anthropic.com/news/enterprise-ai-services-company
[3] https://x.com/bgurley/status/2047362384562700385?s=20
Thanks to Da Yu for reading drafts of this.