OpenAI hints it could lose up to $14 billion in 2026, triple its projected losses for this year
TechFlow Selected TechFlow Selected
OpenAI hints it could lose up to $14 billion in 2026, triple its projected losses for this year
According to media analysis of OpenAI's financial documents, the company is not expected to become profitable until 2029, when its revenue will reach $100 billion.
Navigating Web3 tides with focused insightsBy Zhao Yuhe, Wall Street Insights
OpenAI recently completed a $6.6 billion funding round, pushing its valuation above $150 billion. However, according to media analysis of OpenAI's financial documents, the company's losses next year could reach $14 billion—nearly triple this year's projected losses. The company is not expected to become profitable until 2029, when revenues are forecasted to hit $100 billion. This estimate excludes equity compensation, which, despite being one of OpenAI’s largest expenses, is non-cash.
According to The Information, OpenAI’s financial filings reveal that the profitability metrics emphasized to investors exclude certain major costs, such as the billions spent annually on training large language models. If these expenses are excluded, OpenAI expects to achieve profitability by 2026.
Losses Remain High, But Financial Outlook Is Optimistic
The documents, including financial statements and projections, may shift perceptions about OpenAI’s financial outlook toward greater optimism.
First, OpenAI’s cash burn is significantly lower than previously expected. In the first half of this year, the company burned through approximately $340 million in cash, while still holding $1 billion in cash on its balance sheet before fundraising. However, the documents suggest cash burn could sharply increase in coming years.
Second, there is a substantial gap between OpenAI’s cash flow and its reported losses, reflecting different accounting treatments under standard GAAP for key expenses like equity compensation and computing credits. In the first half of this year, OpenAI reported a net loss of $3 billion.
OpenAI projects total spending from now through 2029 will exceed $200 billion, excluding equity compensation costs. Of that, 60% to 80% annually will be allocated to model training or inference.
Document analysis indicates OpenAI anticipates cumulative losses from 2023 to 2028 (excluding equity compensation) will amount to $44 billion. The same analysis suggests the company expects profits of $14 billion on that basis by 2029.
In the first half of this year, OpenAI reported $1.5 billion in equity compensation, potentially equivalent to its revenue during that period.
The documents also indicate Microsoft takes a 20% share of OpenAI’s revenue—higher than previously anticipated.
OpenAI expects computing costs for model training to rise sharply in the coming years, potentially reaching $9.5 billion annually by 2026. This is in addition to amortizing upfront research costs for large language models; the company spreads research compute costs over multiple years. That figure is also rising rapidly—from an expected $1 billion this year to more than $5 billion by 2026.
Some of OpenAI’s computing costs are not paid in cash. As part of Microsoft’s $10 billion investment last year, it prepaid computing credits for OpenAI. According to the documents and sources familiar with the matter, Microsoft covered around $500 million in data center leasing fees on OpenAI’s behalf in the first half of this year.
It remains unclear how much computing credit OpenAI has left. However, if the company increases compute spending as projected, it may need to draw more heavily on its own funds. Previous reports indicate OpenAI is also discussing borrowing to build data centers faster than Microsoft can provide them.
Still, analysts note that if future models prove more durable—due to competitors’ inability to catch up quickly—or if breakthroughs reduce training costs, OpenAI could scale back compute spending, reducing pressure on its cash reserves.
Moreover, if ChatGPT continues growing as expected and new products generate strong revenue, investors may overlook high spending. By 2029, after a decade of operating a profitable business, OpenAI is projected to generate annual revenue comparable to what Nvidia and Tesla have achieved over the past 12 months.
ChatGPT Will Dominate Sales
OpenAI believes ChatGPT will continue to account for the vast majority of its revenue in the coming years, far surpassing API sales to developers. The company also expects new products to surpass API revenue by the end of 2025, reaching nearly $2 billion that year.
While the exact nature of these new products is unclear, insiders told media the company is currently developing agent-like tools capable of handling complex and repetitive computer tasks on users’ behalf, as well as research assistants.
In addition, the company has discussed raising subscription prices for its most advanced AI technologies. Other products not yet fully launched include the Sora video generator, a more direct competitor to Google Search, and software for robotics developers.
OpenAI also expects API sales growth to slow significantly. The reason is unclear, though competition in this space is intensifying from companies like Anthropic, Microsoft, and Google.
Nonetheless, these forecasts likely assume OpenAI maintains its lead in AI development, despite increasing competition and ongoing employee departures.
Gross Margin Expected to Improve
OpenAI projects its gross margin—a profitability metric reflecting the percentage of direct business costs relative to revenue—will reach about 41% this year. This is well below the typical 65% to 70% range seen among cloud software startups. Higher costs are primarily due to the computing power required to run existing models (known as inference computing), which is expected to consume $1.8 billion of this year’s projected $3.7 billion in revenue.
Currently, OpenAI’s direct business cost ratio is slightly higher than that of Uber during its loss-making phase in 2016, three years before going public.
OpenAI says its business model will improve as revenue grows faster than compute costs, projecting gross margins of 49% next year and 67% by 2028.
Altimeter Capital, one of OpenAI’s supporters, attributes this improvement mainly to declining inference costs. Citing OpenAI data, Altimeter notes that developer pricing for GPT-4 dropped 89% between March 2023 and August 2024.
Staff Growth, Data Spending Declines
OpenAI’s largest operating expense is personnel costs, projected at around $700 million this year, excluding stock-based compensation. The company expects headcount to grow next year, with salary expenses nearly tripling to $2 billion. Afterward, however, growth in these costs is expected to slow.
Another major operating cost—data expenses—is projected to swell to about $500 million this year but then gradually decline to $200 million in subsequent years. This suggests OpenAI believes future data spending for model training will be less intensive than this year, during which it signed numerous licensing deals with media companies.
Financial projections also show OpenAI does not plan to spend heavily on sales and marketing to drive revenue growth, expecting these expenses to represent only 5% to 7% of revenue. This is lower than the typical sales and marketing ratios seen at popular consumer subscription services like Netflix and Spotify.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
