$150 Billion in AI IPOs. Not a Dollar of Profit.

Wall Street rolls out the red carpet for two companies that have never turned a profit

The two most valuable artificial intelligence companies on the planet are racing to sell shares to the public before the year ends, and neither one has ever posted a profitable quarter. OpenAI, the maker of ChatGPT, is targeting a Nasdaq listing at a valuation approaching $1 trillion, while Anthropic, the maker of Claude, is in active discussions with Goldman Sachs and JPMorgan Chase for an IPO as early as October 2026 that could raise more than $60 billion. Together, the two companies could collectively raise upwards of $150 billion in what would be the largest capital markets event in the history of the technology industry. The largest tech IPO ever, Alibaba’s 2014 listing, raised $21.8 billion. OpenAI and Anthropic are proposing to raise roughly seven times that amount — combined — in a single quarter.

The numbers are staggering on both sides of the ledger. OpenAI crossed $25 billion in annualized revenue at the end of February, generating $2 billion per month from ChatGPT subscriptions, API access, and enterprise contracts. Anthropic’s annualized revenue has surged to $19 billion as of March, more than doubling from $9 billion at the end of 2025 — growth of 1,167 percent year-over-year that makes even the fastest-scaling SaaS companies in history look pedestrian. But revenue is not profit. OpenAI generated $13.1 billion in revenue in 2025 and spent approximately $22 billion to do it, burning $1.69 for every dollar it earned. Internal projections show a $14 billion loss in 2026 and cash burn reaching $57 billion annually by 2027. Anthropic’s finances are less dire but still sobering: the company burned roughly $3 billion in cash in 2025 and has estimated $80 billion in cloud infrastructure costs through 2029.

Wall Street, apparently, does not mind. The banks are jockeying for position. Anthropic has engaged Wilson Sonsini as lead IPO counsel and Goldman Sachs and JPMorgan as lead underwriters, with Morgan Stanley in preliminary conversations. OpenAI hired Cynthia Gaylor, the former CFO of DocuSign, as its first head of investor relations to sharpen governance and investor messaging. The race to list first is not academic — it is strategic. Institutional allocation budgets for new listings are finite, and whichever company captures the bulk of that capital first gains a structural advantage. Bloomberg has reported that SpaceX, valued at $1.5 to $1.75 trillion, intends to list ahead of both AI companies — potentially as early as June — in a deliberate attempt to beat them to market and absorb available institutional capital before the AI IPO wave begins. If SpaceX, OpenAI, and Anthropic all go public in 2026, they could become the three biggest venture-backed IPOs of all time, with a combined private valuation exceeding $2.9 trillion and a collective capital raise that would eclipse every VC-backed IPO since 2000.

The structural transformation that made this possible happened quietly. In October 2025, OpenAI completed a recapitalization that converted its for-profit subsidiary into OpenAI Group PBC — a public benefit corporation — while the nonprofit was renamed the OpenAI Foundation and retained oversight. The Foundation’s equity stake was valued at approximately $130 billion; Microsoft’s stake in the PBC was valued at roughly $135 billion, representing about 27 percent on an as-converted diluted basis. That restructuring eliminated the original capped-profit structure — investors’ returns had been limited to 100 times their investment — and created a conventional equity instrument that public market investors can buy and sell. Without that conversion, neither the $122 billion private raise nor a future IPO would have been legally possible.

The question that hangs over all of this is deceptively simple: can two companies that have never demonstrated they can make money convince public market investors to give them a combined $1.5 trillion in equity value? The answer will define whether 2026 becomes the year that AI proved it could build enduring businesses or the year that the most expensive technology bubble in history found its ceiling.

The revenue engines and why the math still does not work

The bull case for these IPOs starts with revenue growth that borders on the absurd. OpenAI’s trajectory speaks for itself: from near-zero revenue in 2022 to $3.7 billion in 2024 to $13.1 billion in 2025 to an annualized $25 billion run rate in early 2026. The company now generates $2 billion per month, a figure that puts it among the fastest-growing technology companies in history by any measure. ChatGPT has become the default consumer AI interface for hundreds of millions of users, and the enterprise API business has attracted virtually every Fortune 500 company. OpenAI’s February 2026 funding round — a staggering $122 billion raise at an $840 billion post-money valuation — was the largest private technology financing in history.

Anthropic’s revenue story is arguably even more impressive on a percentage-growth basis. The company went from roughly $1 billion in annualized revenue in early 2025 to $9 billion by year-end and then to $19 billion by March 2026. Claude Code, Anthropic’s developer-focused coding tool, has emerged as a breakout product, reaching $2.5 billion in annualized revenue in February — a figure that more than doubled since the start of the year. The enterprise flywheel is spinning: Anthropic now serves over 300,000 business customers, with eight of the Fortune 10 among them and more than 500 customers spending over $1 million annually, up from a dozen two years ago. Corporate customers account for approximately 80 percent of revenue, giving Anthropic a more defensible revenue base than OpenAI’s consumer-heavy mix.

But here is where the numbers turn uncomfortable. OpenAI’s cost structure is brutal. In 2025, the company spent roughly $22 billion — on compute, research, salaries, and infrastructure — to generate $13.1 billion in revenue. That ratio implies the company is spending $1.69 for every dollar it earns. Inference costs quadrupled in 2025 as demand for ChatGPT and API services outpaced available compute capacity, forcing OpenAI to purchase computing resources at emergency prices. The projection for 2026 is a $14 billion net loss. Cumulative losses before profitability could reach $143 billion by 2029. OpenAI itself has acknowledged it does not expect to reach breakeven until 2030, four full years from the date it plans to ask public market investors to assign it a trillion-dollar valuation.

Anthropic’s economics are slightly better but far from comfortable. The company forecasts dropping its cash burn to roughly one-third of revenue in 2026, with internal projections showing cash-flow breakeven in 2028. But inference costs surged 23 percent more than expected in 2025, torching ten percentage points of gross margin and leaving it at approximately 40 percent. The company has projected $80 billion in cloud infrastructure costs through 2029 — a figure that is staggering even by big tech standards. Anthropic projects generating $2.10 in revenue per dollar of compute cost by 2028, compared to OpenAI’s projected ratio of $1.60, which suggests a plausible path to unit economics. But plausible and proven are separated by billions of dollars of execution risk.

The combined picture yields an original calculation that no single press release contains: OpenAI and Anthropic generated a combined $44 billion in annualized revenue in early 2026 while collectively burning through an estimated $17 billion or more in annual cash. The AI industry’s two flagship companies are growing faster than almost any businesses in history — and spending even faster than they grow. The IPO question is not whether these companies generate demand. The question is whether the economics of intelligence-as-a-service can ever support the valuations that Wall Street is preparing to stamp on them.

Three fault lines that could crack the trillion-dollar thesis

The skeptic’s case against these IPOs does not require you to believe that AI is a fad. It requires only that you ask whether the current competitive dynamics, cost structures, and market conditions can sustain the implied valuations. There are at least three structural risks that deserve serious scrutiny.

The first is the profitability paradox. Neither company has demonstrated that selling AI inference at scale can be a profitable business. OpenAI burns more than $150 million every day. Anthropic’s 40 percent gross margins are respectable for a young company but perilous for one seeking a $400 billion to $500 billion public valuation — for context, Microsoft’s gross margins are 69 percent, and Google’s are 57 percent. The fundamental challenge is that training frontier models costs billions and the resulting capabilities become commoditized within months. GPT-5.4, Gemini 3.1 Pro, and Claude Sonnet 4.6 are all world-class, and choosing between them increasingly comes down to workflow fit and cost rather than raw capability differences. When the product is approaching parity, pricing power erodes. When pricing power erodes, the path from $44 billion in combined revenue to sustainable profit margins becomes significantly harder.

The second fault line is market absorption. Can the public markets actually digest $150 billion in AI IPOs — plus SpaceX’s expected $75 billion raise — in a single quarter without triggering a liquidity crunch? The total IPO market in 2024 was approximately $130 billion globally. OpenAI and Anthropic alone are proposing to raise more in one quarter than the entire world’s IPO market produced in the previous year. Institutional investors who participate in these offerings will necessarily have less capital to allocate elsewhere, which means every other tech company contemplating a 2026 or 2027 IPO faces the prospect of a severely compressed market. The mega-IPO wave could create its own gravity, pulling capital away from the broader technology ecosystem and creating a feast-or-famine dynamic where two companies absorb the oxygen that hundreds of others need to breathe.

The third risk is the structural dependency on a single assumption: that AI demand will continue accelerating indefinitely. Both companies’ valuations are built on the premise that the total addressable market for AI inference will expand fast enough to support their growth trajectories while margins improve. But there are early cracks in that assumption. The MIT study from 2025 that found 95 percent of organizations getting zero measurable return from generative AI has become a reference point for enterprise CIOs questioning their AI budgets. Cloud providers are reporting that many enterprise AI projects never make it past the proof-of-concept stage. If enterprise AI adoption plateaus — even temporarily — the revenue growth rates that justify these valuations will slow, and the losses that Wall Street is currently willing to overlook will become impossible to ignore. History offers a cautionary precedent: the WeWork IPO in 2019 was pulled after investors questioned whether a company losing $1.9 billion per year on $3.5 billion in revenue deserved a $47 billion valuation. OpenAI is projecting losses that are seven times larger on revenue that is less than twice as large. The question is not whether AI is more transformative than coworking spaces — it obviously is. The question is whether transformation alone justifies any price.

Investors may also be paying for these IPOs later in the company lifecycle than they realize. As The Conversation noted, today’s mega-IPOs are fundamentally different from the Amazon and Apple debuts of decades past. These companies are going public at valuations that already reflect many years of private market appreciation, meaning early employees and venture investors are cashing out, and public market investors may be left holding shares with far less upside than they expect. Amazon went public at a $438 million valuation. OpenAI wants to go public at roughly 2,300 times that amount. The appreciation curve is not starting from the same baseline, and the risk-return calculus is fundamentally different.

The trillion-dollar fog and an operator’s survival guide

Despite the risks, the bull case for these IPOs rests on something more powerful than financial models: narrative gravity. AI is the defining technology of the decade, and OpenAI and Anthropic are its two most prominent standard-bearers. If these companies go public successfully, they will validate AI as a permanent infrastructure layer — the equivalent of the internet’s mid-1990s transition from curiosity to utility. Institutional investors who miss the offering risk being left out of the most important technology cycle since cloud computing. That fear of missing out, combined with genuine revenue traction that dwarfs anything the dot-com era produced, creates a powerful cocktail of demand.

The sequencing matters enormously. SpaceX is positioning itself to list first, potentially in June, at a valuation of $1.5 to $1.75 trillion. Bloomberg has reported that Elon Musk specifically wants to beat the AI companies to market, capturing the lion’s share of institutional allocation budgets before OpenAI and Anthropic absorb available capital. If SpaceX executes cleanly, it will set the tone for the AI IPOs. If SpaceX stumbles, or if broader market conditions deteriorate between June and October, the AI IPO window could narrow or close entirely. Timing, as much as fundamentals, will determine whether 2026 becomes the year of the trillion-dollar debut or the year the music stopped.

The competitive dynamics between OpenAI and Anthropic add another layer of complexity. OpenAI executives have privately expressed concern about Anthropic getting to market first, and Anthropic’s October target date puts pressure on OpenAI to accelerate its own timeline. The race to list first is not vanity — it is a capital strategy. The first mover captures the freshest allocation budgets, sets the valuation benchmark, and forces the second mover to accept whatever the market has left. In a scenario where both companies list within weeks of each other, the competition for institutional dollars could create a zero-sum dynamic that depresses both offerings below their targets.

Anthropic may hold a subtle structural advantage in this race. Its revenue mix — 80 percent enterprise, driven by products like Claude Code that have clear productivity ROI — tells a cleaner story to institutional investors than OpenAI’s consumer-heavy model. Anthropic’s projected path to cash-flow breakeven in 2028 is two years ahead of OpenAI’s 2030 target. And Anthropic’s $19 billion ARR at a $380 billion private valuation implies a more reasonable revenue multiple than OpenAI’s $25 billion ARR at $840 billion. The market may reward the company that can articulate a credible path to profitability over the one that offers the biggest top-line number and the most impressive brand recognition.

For operators, investors, and builders watching this unfold, the calculus requires navigating what amounts to a trillion-dollar fog — a market where the signal-to-noise ratio is historically low and the stakes are historically high. Here is a practical framework for reading the room:

• Watch the SpaceX listing. If SpaceX prices above its range and trades up on Day One, the AI IPO window is wide open. If SpaceX prices at or below its range, the AI companies will likely delay or downsize.
• Track the revenue multiples, not the revenue. A company growing at triple digits deserves a premium, but OpenAI at roughly 34x forward revenue and Anthropic at 20-26x are pricing in perfection. Any deceleration in quarterly revenue growth will compress those multiples violently.
• Follow the lockup expirations. Early investors and employees typically face 90- to 180-day lockups after an IPO. The real test of conviction comes when insiders are free to sell. If Sequoia, Andreessen Horowitz, and the other early backers liquidate aggressively, it signals that the people closest to these businesses believe the upside has been captured.
• Monitor enterprise retention rates. The single most important metric for both companies post-IPO will be net revenue retention among enterprise customers. If the 80 percent of Anthropic’s revenue that comes from businesses proves sticky — expanding within accounts, not churning — the premium valuation holds. If enterprises start consolidating AI vendors or cutting budgets, the thesis unravels quickly.
• Do not confuse the IPO with the investment. The IPO is a liquidity event for insiders, not necessarily an optimal entry point for new investors. The most asymmetric opportunities in AI may come not from buying OpenAI or Anthropic at peak hype, but from the infrastructure companies, vertical AI startups, and second-tier model providers that benefit from the ecosystem’s growth without carrying the burden of trillion-dollar expectations.

What is certain is that the next six months will determine whether the AI industry’s financial architecture stands on bedrock or sand. OpenAI and Anthropic have built extraordinary technology and captured extraordinary demand. Whether that demand translates into extraordinary returns for public market investors is the $150 billion question that 2026 will answer.

In other news

Tennessee signs AI therapy bot ban into law — Governor Bill Lee signed SB 1580 this week, prohibiting the deployment of any AI system that represents itself as a qualified mental health professional. The bill passed the Senate 32-0 and the House 94-0, reflecting rare bipartisan consensus on restricting AI in clinical mental health contexts. Nebraska and Oregon are advancing similar legislation.

Mercor confirms $10 billion AI startup breached via LiteLLM supply chain attack — Mercor, the AI recruiting startup valued at $10 billion with clients including Anthropic, OpenAI, and Meta, confirmed a cyberattack linked to a compromised version of the open-source LiteLLM project. Extortion group Lapsus$ claimed it stole 4 TB of data including 939 GB of source code, internal Slack messages, and 3 TB of contractor video interviews and identity documents.

Kleiner Perkins raises $3.5 billion for AI-focused funds — The storied venture firm closed $3.5 billion across two funds — $1 billion for early-stage bets and $2.5 billion for growth-stage AI companies — bringing total assets under management above $21 billion. Kleiner Perkins holds early stakes in Anthropic, Together AI, and Harvey and is positioning for the very IPO wave this article describes.

AI chatbot safety bills surge across red and blue states — Following Oregon and Washington’s passage of chatbot safety laws, at least six more states are advancing similar legislation. Nebraska’s LB 1185 has been attached to the Agricultural Data Privacy Act and appears headed for passage, while Georgia and New York have active bills targeting AI disclosure requirements and minor protections.