Welcome to Eye on AI. Emily Forlini right here, filling in for Jeremy one final time as his trip involves an in depth. In as we speak’s problem:
- Juicy particulars OpenAI doesn’t need you to see in its new report
- Anthropic reportedly plans a $2 trillion IPO in October—the biggest ever
- OpenAI replaces its chief income officer after lower than a yr
- Google pronounces Sam Altman lifeless—for 41 minutes
It actually sunk in for me this week simply how a lot cash is flowing within the AI trade.
I spoke with two former OpenAI staff who made about $10 million in a day by promoting shares in an inner tender provide, which Bloomberg experiences totaled $7 billion throughout the employees. Then, this morning, on Fortune‘s weekly AI podcast, my coworker Beatrice Nolan and I interviewed the CEO of Lovable. This week, the previous Stockholm-based agency, which is barely three years previous, doubled its valuation to $13.3 billion.
A pair million, 100 billion, a trillion (or two, within the case of Anthropic’s upcoming IPO)—what’s the distinction at this level? There’s only one huge downside looming within the background: The ROI of AI adoption continues to be not clear for firms.
OpenAI grapples with this existential query in a 69-page report revealed on August 11 on the enterprise adoption of ChatGPT. On its face, the report tells the a narrative of exponential AI utilization progress throughout all seniority ranges and job capabilities, highlighting what it calls a “frontier hole,” during which firms who’re utilizing AI are pulling forward of those that aren’t. In different phrases, for those who’re not utilizing AI—particularly brokers you may delegate duties to—you’re shedding.
However the positive print tells a unique story.
Extra AI doesn’t imply more cash
In a single small desk on web page 35, the researchers report no statistically important correlation between the income per worker, and the way a lot these staff use AI, measured in messages despatched and tokens used.
“Income per worker isn’t meaningfully related to output tokens per worker or messages per lively consumer as soon as different controls are included,” the report explains.
It says that firms that have already got larger income per worker are usually early ChatGPT adopters. Additionally, firms that use the tech extra are inclined to have larger income per worker usually. In different phrases, huge, profitable firms usually tend to have hopped on the AI prepare. Nevertheless, the examine doesn’t clearly set up that the extra AI they use, the more cash they make.
Executives are utilizing AI the least
Executives will not be greatest geared up to gauge ROI as a result of they’re utilizing it the least—one other nugget buried within the report. It’s not simply that firms have fewer executives than they do basic staff. However what’s fascinating concerning the graph on web page 29 is that almost all senior staff are utilizing it much less intensely, with the least weekly messages per consumer.
Early profession staff have by far probably the most utilization, a degree OpenAI CFO Sarah Friar highlighted in her LinkedIn submit concerning the report: “For leaders, that’s a reminder that aggressive benefit comes from the individuals closest to the work. Hearken to them, be taught from them, and assist the remainder of the group catch up.”
OpenAI’s enterprise gross sales had a tough This fall’2025
Surprisingly, OpenAI’s total utilization inside enterprises utterly flatlined from about October 2025 to December 2025. In a graph (web page 26) depicting output token progress, the black line representing “complete” progress is nearly completely flat for that point interval. Throughout this time, Anthropic’s Claude Code was taking the company world by a storm, changing into the go-to platform at many locations.
To the corporate’s credit score, in January 2026, the road thrusts upward into an exponential curve. As one VC advised me yesterday, “OpenAI’s run fee in 2026 has been fairly unimaginable.” OpenAI attributes the expansion not solely to including new purchasers, but in addition additionally to its present purchasers deepening their use. We additionally know CEO Sam Altman has been reorganizing the corporate round enterprise gross sales, and killing what the corporate referred to as “facet quests,” such because the video app Sora.
In a dash to speed up this line—Or, perhaps to get it going once more? Who is aware of, the graph ends at March 2026—OpenAI as we speak introduced it employed a brand new Chief Income Officer, Dali Rajic, who will change Denise Dresser. It’s an aggressive transfer; Dresser held the function for lower than one yr. Rajic’s focus can be accelerating buyer adoption and serving to companies measure impression as the corporate sprints in the direction of its IPO.
OpenAI paid the teachers who contributed to the report
Lastly, two of the 5 authors are teachers that OpenAI paid to assist with the report. The opposite three are OpenAI staff. Together with teachers in a paper like this sometimes implies better credibility and the impartiality of an outdoor analysis establishment, however the waters are a little bit muddier right here.
On the primary web page, David Holtz and Prasanna Tambe are affiliated with Columbia Enterprise Faculty and Wharton on the College of Pennsylvania, respectively. However a footnote specifies that each “contributed to this work of their capability as paid contractors for OpenAI.”
Did the researchers discover extra that they didn’t publish, as they sometimes would for a tutorial paper? We’ll by no means know, nevertheless it’s one other reminder of what has all the time been true: You’ll need to measure AI’s impression primarily based by yourself first-hand expertise—not the hype.
With that, right here’s extra AI information.
Emily Forlini
emily.forlini@fortune.com
@emilyforlini
FORTUNE ON AI
Microsoft begins to merge client and enterprise Copilot apps in push for tremendous app — by Sebastian Herrera
CoreWeave CEO Michael Intrator cites ‘offered out’ capability as income greater than doubles and backlog swells to $104 billion — by Amanda Gerut
When OpenAI staff have an issue, they e-mail this particular deal with to see if Sam Altman will remedy it instantly — by Emily Forlini
CIOs and CTOs spent years lauding AI. Now, with prices rising, they’re placing limits on the way it’s used — by John Kell
Neglect the gala, these Silicon Valley colleges run their very own enterprise capital funds — by Amanda Gerut
AI IN THE NEWS
Anthropic plans $2 trillion IPO, probably in October. It will be the biggest ever on document, eclipsing SpaceX, however the valuation and IPO timing should not but finalized. Buyers consider the AI lab’s annual income run fee can be between $100 billion and $120 billion by the tip of the yr. Extra in The Monetary Occasions.
Lovable raises $400 million, doubles valuation to $13.3 billion. The vibe-coding startup is barely three years previous. It plans to make use of the funds for growth, together with rising its workforce and increasing its Latin American buyer base. It additionally plans to implement robust security and safety. Extra in The Wall Road Journal.
OpenAI’s chief income officer is out after lower than a yr. The corporate has employed Dali Rajic to switch Denise Dresser, who says she left on her personal accord to pursue different ventures. Many interpret the transfer as coming from the rising presence of co-founder and president Greg Brockman as the corporate gears up for a possible IPO. Extra from OpenAI.
EYE ON AI NUMBERS
41
Lifespan of OpenAI CEO Sam Altman, in accordance with Google’s botched algorithm. On Wednesday, the search large proclaimed Altman lifeless, including a demise date of August 12, 2026 to the actual fact panel it displayed on the prime of its search outcomes. That very same panel mentioned Altman was nonetheless the present CEO of OpenAI. Laborious to do from the grave, however we have seen crazier issues in AI. Possibly it is a premonition for future CEOs who keep on by way of AI avatar reproductions of themselves.
Somebody had apparently vandalized Altman’s Wikipedia web page, claiming he was assassinated in Seattle. The malefactor re-wrote all the introduction to Altman’s Wiki entry prior to now tense, in accordance with Gadget Overview. Wikipedia’s automated instruments flagged the change, and a human editor mounted it 41 minutes later—one minute for every of Altman’s supposed years of life.
Google responded on X after the incident, saying, “When individuals vandalize public data sources, this could have an effect on the data that seems in Search.” It is a foolish, however stark, reminder of how fragile the web’s data ecosystem is. Some blame Wikipedia for permitting rogue editors, whereas others say Google’s on the hook for selecting to show inaccurate data. It is unclear if Google would have caught the difficulty by itself, in 41 minutes or much less.
AI CALENDAR
Nov. 16-17: Fortune 500 Innovation Discussion board, Detroit. Apply right here to attend.
Dec. 6-12: Neural Info Processing Programs (Neurips) convention. Sydney, Australia.
Dec. 7-8: Fortune Brainstorm AI, San Francisco. Apply right here to attend.
