Sam Altman’s creation stands at a crossroads few companies ever reach. OpenAI commands a staggering $852 billion valuation after its latest funding round. Yet the artificial intelligence leader bleeds cash at a rate that would sink most enterprises. Revenue races ahead at unprecedented speed. Losses mount even faster.
By March 31, 2026, OpenAI had closed a $122 billion funding round. The deal, announced on the company’s own site, set its post-money valuation at $852 billion. Investors piled in. SoftBank co-led alongside Andreessen Horowitz, D.E. Shaw Ventures, MGX, TPG and T. Rowe Price-advised accounts. Strategic backers included Amazon, NVIDIA and Microsoft. The round expanded a revolving credit facility to roughly $4.7 billion, still undrawn at closing. (OpenAI)
Revenue tells one side of the story. OpenAI hit $1 billion in annual revenue within a year of launching ChatGPT. It reached $1 billion per quarter by the end of 2024. Now it generates $2 billion per month. That pace outruns the early growth of Alphabet and Meta by a factor of four. Enterprise customers contribute more than 40 percent of revenue. They stand on track to match consumer revenue by the end of 2026. ChatGPT itself counts over 900 million weekly active users and more than 50 million paid subscribers. (OpenAI)
But the cost side paints a darker picture. Leaked financials for 2025 reveal revenue of $13.07 billion, up from $3.7 billion in 2024. Costs and expenses exploded to $34 billion from $12.48 billion. The operating loss swelled to $20.92 billion from $8.78 billion. After various adjustments, including a massive fair value loss tied to its conversion to a for-profit entity, the net loss attributable to the company reached $38.53 billion. In 2024 that figure stood at $5.09 billion. (Where’s Your Ed At; Fortune)
Those numbers come from audited documents viewed by blogger Ed Zitron and independently verified by the Financial Times. OpenAI paid Microsoft $17.2 billion in 2025 alone across various categories. By year-end the company held assets exceeding $50 billion, roughly half in cash. Still, the burn rate raises eyebrows. Sacra estimates inference costs alone ran $8.4 billion in 2025 and will climb to $14.1 billion this year. Gross margins sit at 33 percent. (Fortune)
Analysts project even steeper cash consumption ahead. Cash burn could hit $27 billion in 2026 and $63 billion in 2027. The company does not expect positive cash flow until 2030. Cumulative losses may top $218 billion by then. That scale dwarfs Uber’s total losses before it turned profitable. And yet OpenAI filed confidentially with the SEC in June 2026 for an initial public offering. It eyes a valuation above $1 trillion. Goldman Sachs and Morgan Stanley are leading the process. Timing remains undecided. (Sacra; Tech Insider)
The growth metrics dazzle. Annualized revenue run rate topped $25 billion by February 2026, up from $20 billion at the end of 2025. Some reports place the figure near $30 billion. API usage has scaled dramatically. The company processes more than 15 billion tokens per minute. Its Codex coding tool counts over 2 million weekly users after fivefold growth in three months. These figures explain why investors keep writing enormous checks. (Sacra; OpenAI)
Yet skepticism grows in some quarters. Certain investors question whether the $852 billion tag holds up as the company shifts emphasis toward enterprise deals and fends off rivals such as Anthropic. That competitor recently surpassed OpenAI in valuation, hitting $965 billion in one report. Revenue leadership has seesawed between the two. OpenAI’s own internal projections once called for $14 billion in losses this year, excluding stock-based compensation. Actual first-quarter 2026 results showed a $6.95 billion non-GAAP operating loss on $5.7 billion in revenue. The operating margin hit negative 122 percent. (The Information; FutureSearch)
Compute demands drive much of the spending. OpenAI’s power consumption grew from 0.2 gigawatts in 2023 to 1.9 gigawatts in 2025. Plans call for hundreds of billions more in infrastructure. The firm earlier scaled back long-term spending expectations from $1.4 trillion to $600 billion through 2030. It relies primarily on NVIDIA chips but spreads risk across Microsoft, Oracle, AWS, Google Cloud and alternatives from AMD, Cerebras and Broadcom. The strategy aims to create a self-reinforcing flywheel. More usage. More data. Better models. More usage. (CNBC)
CFO Sarah Friar called 2026 the year of “practical adoption.” The focus shifts from raw capability to real-world integration. Enterprise traction supports that pivot. So does the subscriber base. But the math remains unforgiving. Revenue tripled from 2024 to 2025. Costs nearly tripled too. The expense-to-revenue ratio improved from $2.37 spent per dollar earned to $1.60. Progress, yes. But still far from breakeven. (Fortune)
Public market investors will soon judge for themselves. The confidential S-1 filing offers a preview of what an OpenAI prospectus might contain. Losses on this scale normally deter traditional IPO buyers. Yet few companies boast this combination of user engagement and revenue velocity. ChatGPT sessions run six times higher than the next leading AI application. Users spend four times more total time with OpenAI products. Those advantages could command a premium. Or they could fade if competitors close the performance gap. (Tech Insider)
Microsoft’s role adds another layer. The software giant holds a long-term partnership. It has invested heavily and receives substantial payments in return. Those flows totaled billions last year. The arrangement gives OpenAI access to vast computing resources while providing Microsoft exclusive access to cutting-edge models. Both sides benefit. Both face risks if the economics shift.
Recent weeks brought fresh signals. Anthropic’s revenue reportedly overtook OpenAI’s in some measurements. Talent flows to new ventures. One former OpenAI executive joined the board of a reusable rocket startup. Investors in older SPVs that bought at $80 billion valuation now sit on paper gains measured against the $850 billion mark, though liquidity remains limited. And a lawsuit against Oracle accused the cloud provider of downplaying OpenAI’s missed internal targets to its own investors. The pressure intensifies. (Morningstar)
OpenAI projects continued rapid expansion. Internal targets once eyed $62 billion in annualized revenue by mid-2027. Outside forecasts land closer to $42 billion. Either path would represent extraordinary scale for a company that generated $28 million as recently as 2022. The question is whether the capital markets will fund the gap until profits arrive. History offers mixed lessons. Amazon burned cash for years. So did Tesla. Both delivered eventual returns that justified the risk. Others did not.
The IPO process itself will test investor appetite. A listing could rank among the largest in history. It might value the company above $1 trillion at debut. But public scrutiny will intensify. Quarterly reports. Margin pressure. Executive compensation tied to stock performance. Altman and his team must convince shareholders that today’s losses buy tomorrow’s dominance. So far, private investors have bought that thesis at ever-higher prices. The public test looms larger.
Enterprise adoption offers the clearest path forward. More than 40 percent of revenue already flows from that segment. Parity with consumer revenue would diversify the base and stabilize growth. Higher pricing, efficiency gains in inference, and new product layers could lift margins. The company talks of building an AI superapp that unifies agent experiences. Success there might accelerate uptake and willingness to pay. Failure could prolong the cash drain.
For now the numbers refuse to reconcile easily. Revenue at $25 billion run rate. Losses measured in tens of billions annually. Valuation at $852 billion. The disconnect captures the unique moment AI enjoys. Excitement over future potential overrides present economics. Whether that equation holds through an IPO and beyond will shape not just OpenAI but the broader technology sector for years. The stakes could hardly run higher. The data could hardly look more extreme.