OpenAI’s financial documents reveal controversial accounting practices that have drawn sharp criticism from industry experts and accounting professionals. According to documents obtained by The Information, the AI startup is asking investors to exclude billions of dollars spent on training AI models when evaluating its profitability—a move that experts say crosses ethical and practical boundaries.
The company projects $44 billion in total losses from 2023 to 2028, despite revenue growth that could reach $100 billion by 2029. However, using OpenAI’s alternative earnings metric that excludes AI model training costs, much of that massive loss disappears, and the company claims it will achieve profitability by 2026.
This approach has drawn comparisons to failed startups like Groupon and WeWork, both of which attempted to exclude core business expenses from their financial reporting. Groupon’s “adjusted CSOI” excluded marketing expenses before its 2011 IPO, while WeWork promoted “community-adjusted EBITDA” that ignored fundamental operational costs. The SEC objected to Groupon’s accounting, and WeWork ultimately filed for bankruptcy in November 2023.
Francine McKenna, an accounting expert who formerly worked at KPMG Consulting and PwC, called OpenAI’s approach “a bridge too far” and “not kosher.” She emphasized that AI model training represents ongoing product maintenance and development that cannot legitimately be excluded from earnings calculations. “They are going to have to constantly train their AI models on new information,” McKenna explained. “They will never be done.”
A veteran venture capitalist who previously served as a Big Tech executive agreed, calling it “a sign of hubris” that OpenAI would ask investors to ignore fundamental business expenses. The VC noted that while other costs like inference might decline over time, AI model training will remain a large and ongoing expense.
The accounting concerns become especially critical if OpenAI pursues an IPO, which some investors like Brad Gerstner of Altimeter Capital have suggested should be the company’s next move. The SEC heavily scrutinizes alternative profitability measures and typically doesn’t allow companies to exclude ongoing business costs. OpenAI did not respond to requests for comment on these concerns.
Interestingly, SoftBank—known as a bubble chaser—was WeWork’s main investor before its failed IPO and is now one of OpenAI’s biggest backers in its recent $6.6 billion funding round. Meanwhile, Apple reportedly considered investing but backed out at the last minute.
Key Quotes
That is a bridge too far. It’s not kosher.
Francine McKenna, an accounting expert and former KPMG and PwC professional, responded to OpenAI’s practice of excluding AI model training costs from its earnings calculations. Her strong language underscores how unusual and problematic this accounting approach is by professional standards.
They are going to have to constantly train their AI models on new information. They will never be done. I think it’s a sleight of hand to try to exclude these costs.
McKenna explained why AI model training cannot be treated as a one-time expense that can be excluded from earnings. This highlights the fundamental flaw in OpenAI’s accounting approach—the costs they want to exclude are ongoing and central to their business operations.
If they put that in an IPO prospectus and the SEC looked at it, the SEC would probably say you can’t use these as an adjustment to income. To me, that’s more akin to ongoing product maintenance and development.
McKenna outlined the regulatory challenges OpenAI would face if it attempts to use this accounting method in a public offering. This quote emphasizes that what works for private fundraising won’t pass muster with securities regulators.
These costs are real and it’s a sign of hubris that OpenAI is asking investors to ignore what is a fundamental business expense.
An anonymous veteran venture capitalist and former Big Tech executive provided this assessment, characterizing OpenAI’s approach as arrogant and disconnected from financial reality. The source’s credentials lend significant weight to this criticism.
Our Take
OpenAI’s accounting practices reveal a troubling pattern in the AI industry where hype and narrative increasingly trump financial fundamentals. The company’s attempt to exclude AI model training costs—literally the core of what makes OpenAI valuable—suggests either desperation to show profitability or a belief that AI exceptionalism exempts them from standard accounting principles.
The historical parallels are impossible to ignore. Groupon and WeWork both convinced investors to overlook fundamental costs before reality caught up with them. SoftBank’s involvement in both WeWork and OpenAI is particularly concerning, suggesting pattern recognition failures among major investors. Apple’s last-minute withdrawal hints that some sophisticated investors are growing skeptical.
What’s most striking is the $44 billion in projected losses through 2028. Even in the capital-intensive tech world, that’s an extraordinary burn rate. If OpenAI truly believes AI model training costs should be excluded, it raises questions about whether the economics of foundation model development are sustainable at all. The AI industry may be heading toward a reckoning where the enormous costs of training increasingly powerful models collide with the reality of generating actual profits.
Why This Matters
This story reveals potential red flags in AI industry financial practices at a critical moment when artificial intelligence companies are raising unprecedented amounts of capital. OpenAI’s accounting approach could set a dangerous precedent for how AI startups report their financial health to investors and the public.
The implications extend beyond OpenAI to the broader AI investment ecosystem. If leading AI companies can exclude their core operational costs from earnings calculations, it becomes nearly impossible for investors, regulators, and the public to accurately assess the true economics of the AI industry. This matters especially as AI companies consume massive amounts of capital while promising transformative returns.
The comparison to Groupon and WeWork is particularly sobering, as both companies used similar accounting gymnastics before experiencing dramatic failures. WeWork’s bankruptcy and Groupon’s near-collapse demonstrate what can happen when companies prioritize narrative over financial reality. With OpenAI potentially heading toward an IPO, these accounting practices could face SEC scrutiny that forces greater transparency. The involvement of SoftBank, which backed WeWork’s doomed venture, adds another layer of concern about whether lessons from previous tech bubbles have been learned.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- OpenAI CEO Sam Altman Hints at Potential Restructuring in 2024
- OpenAI’s Valuation Soars as AI Race Heats Up
- Wall Street Asks Big Tech: Will AI Ever Make Money?
- The AI Hype Cycle: Reality Check and Future Expectations
- Google’s Gemini: A Potential Game-Changer in the AI Race
Source: https://www.businessinsider.com/openai-ai-adjusted-earnings-groupon-wework-accounting-2024-10