OpenAI at $1 Trillion: Here Is Why I Would Wait Before Buying

STREETWISE ECONOMICS

Weekly Column   ·  22 May 2026

By Isaac Jonas  |  Economist, Streetwise Economics  |  May 2026

When the most consequential private company of the artificial intelligence era prepares to list on public markets at a targeted valuation approaching one trillion dollars, the question is not whether it deserves attention. It demands attention. The question is whether the price being asked of public investors leaves any room for the investor to be right and still be rewarded.

That company is OpenAI. After its $122 billion private funding round closed in March 2026 at a post-money valuation of $852 billion, reports indicate the company is preparing for a public listing in the fourth quarter of 2026 at a valuation of up to $1 trillion. I have run OpenAI through the same economist’s framework I applied to SpaceX last month — revenue analysis, valuation multiples, scenario modelling, and historical precedent. The conclusion is different. Where I argued the SpaceX bet was expensive but defensible, my view on OpenAI is that the asymmetry has moved against the buyer at IPO. Here is my reasoning.

What OpenAI Actually Is

The first analytical step is to understand what investors are actually paying for. OpenAI is, in 2026, no longer simply the company that makes ChatGPT. It is a multi-product platform built on five distinct revenue lines, each with a different growth profile and a different margin structure.

Consumer subscriptions remain the largest revenue contributor. ChatGPT has approximately 900 million weekly active users, with around 50 million of those paying across the company’s subscription tiers — Free, ChatGPT Go at $5 to $8 per month, Plus at $20 per month, Pro at $200 per month, and Team and Enterprise seats at higher levels. Notably, OpenAI’s own internal projections show the company expects its $20-per-month Plus tier to shrink dramatically in 2026, offset by an ambitious push to grow the cheaper, ad-supported Go tier from roughly 3 million to over 100 million subscribers. The average revenue per user is expected to fall from approximately $23 to below $12. This is a significant strategic shift that compresses unit economics even as headline subscriber numbers grow.

The enterprise business is the second pillar. OpenAI surpassed one million business customers in late 2025, with more than 9 million paying business seats and a fast-growing ChatGPT for Work product. The API and developer platform, including Codex for software development, is the third pillar — and the one most directly exposed to competitive pricing pressure from Anthropic, Google, and open-source alternatives such as DeepSeek.

The fourth line is advertising, which the company began rolling out only recently. Early indicators suggest the ad business is generating annualised revenue in the low hundreds of millions of dollars, with internal projections of $2.4 to $2.5 billion in 2026 and aspirations of approaching $100 billion annually by 2030. The fifth line is Frontier, OpenAI’s newly unveiled enterprise platform for managing AI agents, distributed through Amazon Web Services as the exclusive third-party cloud partner under the February 2026 Amazon deal.

Pulling these lines together, OpenAI reported approximately $13.1 billion in full-year 2025 revenue and exited the year at a run rate above $20 billion. By February 2026, the annualised run rate had reached approximately $25 billion, equivalent to roughly $2 billion in monthly revenue. That trajectory — from $3.7 billion in revenue in 2024 to a $25 billion run rate within fourteen months — is one of the fastest commercial scaling stories in corporate history.

The Valuation Question

Now comes the harder part.

At a $852 billion post-money valuation on a $25 billion run rate, OpenAI trades at approximately 34 times annualised revenue. If the company lists at $1 trillion, that multiple rises to roughly 40 times. For comparison, Nvidia, the defining hardware company of the AI era and a company with positive operating margins, trades at approximately 30 times revenue. Microsoft, which holds a 27% stake in OpenAI worth approximately $135 billion, trades at roughly 13 times revenue. Amazon’s cloud division AWS trades at approximately 15 times. By any traditional benchmark, OpenAI is asking public markets to pay an extraordinary premium for growth that has not yet translated into operating profit.

This is the issue I cannot escape. SpaceX, at its filed valuation of $1.75 trillion, is asking for a price-to-revenue multiple of approximately 113 times — but the underlying businesses, Falcon launch services and Starlink, are operationally profitable and generating cash. OpenAI is not. According to internal projections reported by The Wall Street Journal and The Information, OpenAI generated approximately $13.1 billion in revenue in 2025 against approximately $22 billion in spending, producing a net loss of roughly $9 billion. The company spent approximately $1.69 for every dollar of revenue it earned. Projections for 2026 show losses widening to approximately $14 billion. The company’s own internal forecasts indicate that cumulative losses between 2023 and 2028 will reach approximately $44 billion, with profitability not arriving until 2029 or 2030 at the earliest. HSBC analysts have suggested OpenAI may need an additional $207 billion in capital through 2030 to fund its growth and infrastructure plans.

OpenAI is asking public markets to pay roughly 40 times revenue for a business that loses approximately $1.69 for every dollar of revenue it earns. The growth is real. The path to making that growth profitable is not yet proven.

The structural issue is compute. OpenAI has committed to approximately $250 billion in Azure spending under its agreement with Microsoft, an incremental $100 billion in AWS commitments over eight years following the February 2026 Amazon deal, and reportedly more than $600 billion in total infrastructure obligations across the next several years. Independent analysts have placed the total compute commitment closer to $1.15 trillion over five years. These are not marketing numbers. These are contractual commitments the company must service regardless of whether ChatGPT user growth or enterprise revenue meets expectations.

What History Teaches Us — And Where the Parallel Breaks

The bull case for paying any price for OpenAI rests on the same historical reasoning that justified Amazon at its 1997 IPO and Google at its 2004 listing. Both companies were richly valued at the time. Both went on to deliver returns measured in tens of thousands of percent. The argument is that OpenAI, as the defining company of the artificial intelligence era, will follow the same pattern.

I find this analogy seductive but incomplete. Amazon listed at a $438 million valuation. Google listed at $23 billion. Both companies were given the gift of being underestimated by the market. OpenAI, by contrast, may list at $1 trillion. The market is not underestimating OpenAI. The market is, if anything, pricing it for near-perfect execution across multiple unproven business lines, while ignoring substantial financial and structural risks. You are not getting in before the world recognises the company. You are getting in after the world has already concluded that this is the most consequential AI company on earth — and priced it accordingly.

Amazon listed at $438 million. Google listed at $23 billion. OpenAI may list at $1 trillion. The historical parallels assume you are buying before the market understands the story. With OpenAI, the story is in the price.

There is also a fundamental structural difference. Amazon and Google built their businesses on infrastructure that became cheaper over time, with widening operating margins as they scaled. OpenAI’s core cost — frontier AI compute — has not yet demonstrated the same trajectory. Inference costs are reportedly dropping, but training costs for next-generation models are rising even faster. According to leaked Microsoft data referenced in industry reporting, OpenAI still burns approximately $2 in inference cost for every $1 of inference revenue it generates. The path to operating leverage exists in theory. It has not yet shown up in the income statement.

The Risks Investors Cannot Ignore

Beyond the valuation, there are four risks that must inform any IPO entry decision.

First, competition has materially intensified. Anthropic, OpenAI’s closest pure-play rival, reached an estimated $19 billion annualised revenue run rate by early 2026, with reported active fundraising at a $900 billion valuation. According to Ramp’s enterprise spending data, Anthropic now wins approximately 70 percent of head-to-head matchups against OpenAI among businesses purchasing AI services for the first time — a complete reversal from 2025. Google’s Gemini family, supported by its own custom silicon and integration into search, launched an aggressive pricing campaign in May 2026 that pressures the entire model layer. Open-source models from DeepSeek and others continue to compress the moat. The duopoly narrative many investors held in 2023 has become a four- or five-way race by 2026.

Second, the Microsoft relationship has been fundamentally restructured. In April 2026, OpenAI and Microsoft amended their partnership to remove the so-called AGI clause and end Microsoft’s exclusivity. OpenAI is now free to distribute through Amazon Web Services, Google Cloud, and other partners, and Microsoft’s IP licence runs through 2032 on a non-exclusive basis. OpenAI continues to pay Microsoft a capped revenue share through 2030. This is a significant net change for the financial model. It opens distribution but removes the single-cloud advantage that underpinned a portion of the original investment thesis.

Third, governance and key-person risk remain genuine concerns. CEO Sam Altman is reportedly the subject of a Congressional review of personal investments and conflicts of interest. The November 2023 board removal episode demonstrated that OpenAI’s governance structure can produce sudden and material outcomes. CFO Sarah Friar is reportedly advocating for a 2027 IPO rather than Altman’s preferred 2026 timeline, citing concerns about whether the company can meet the reporting and operational standards required of a public entity. This is unusual public disagreement and it deserves weight.

Fourth, the litigation overhang from Elon Musk’s lawsuit was substantially resolved on May 18, 2026, when a federal jury in Oakland unanimously found that Musk’s claims against OpenAI and Altman were filed outside the statute of limitations. Musk has indicated he will appeal to the Ninth Circuit. The most material near-term risk from the litigation has now been removed, but the matter is not closed.

My Three-Scenario Framework

Running a simplified scenario model to 2030 produces a wider range of outcomes than the SpaceX exercise, reflecting OpenAI’s greater dependence on competitive position, cost structure, and a still-unproven path to profitability.

In a bear scenario — where competition continues to compress pricing, the path to profitability slips to 2031 or beyond, compute commitments overshoot revenue growth, and the market de-rates the revenue multiple to 12 times — implied 2030 revenue of approximately $100 billion would produce a value of $1.2 trillion. Against a $1 trillion IPO price, that is a gain of approximately 20 percent over four years, or roughly 5 percent annualised. Modest, and well below the cost of capital for most investors. In the worst version of this scenario, if revenue lands closer to $70 billion and the multiple compresses to 10 times, the implied value falls to $700 billion — a 30 percent loss.

In a base scenario — where OpenAI hits roughly two-thirds of its $280 billion 2030 revenue target, the multiple holds at 18 times, and operating margins approach the high teens — the implied value reaches approximately $3.4 trillion. That is a gain of approximately 240 percent over four years, or roughly 36 percent annualised. This is the scenario that justifies the IPO entry — provided you believe the company can execute it.

In a bull scenario — where OpenAI reaches or exceeds its $280 billion 2030 revenue target, achieves operating leverage, and the market values the equity at 25 times revenue — the implied value approaches $7 trillion. That would be a gain of approximately 600 percent over four years.

The base and bull cases are large. The bear case is genuinely modest. The asymmetry, in my view, is therefore weaker than the SpaceX case — because the bear case for SpaceX still leaves a recognisable business with positive cash flows and a launch monopoly, while the bear case for OpenAI involves a business that may not have reached profitability and is competing against three other well-capitalised giants.

My Personal Conditions

Based on this analysis, my position is to wait — not to buy at IPO. The reasoning is that the price already reflects the bull case, the company has not yet demonstrated unit economics that support its valuation, and the most attractive historical pattern for buying frontier technology companies has not been at IPO but during the post-lock-up window when early employees and investors become eligible to sell. For Amazon in 1997, that window arrived. For Facebook in 2012, it arrived. For most large IPOs in market history, it arrives. Patience is the rational response to a high price.

If I were to consider an entry, the conditions would be specific. The IPO price would have to be at or below $1 trillion, not above. The first-half 2026 financials disclosed in the S-1 filing would need to show the loss rate stabilising rather than accelerating. OpenAI’s enterprise market share, as tracked by Ramp and other independent data sources, would need to stop declining relative to Anthropic. And ideally, I would want to see at least two quarterly earnings reports as a public company before committing capital, to understand how Wall Street values the business when it is forced to disclose what it has been free to obscure as a private entity.

Beyond IPO conditions, my general investment principles apply with particular force here. I would treat this as a minimum 10-year hold. I would not use leverage or borrowed money. I would size any position at no more than what I would normally allocate to a single high-risk investment — and most likely below that, given the absence of operating profit. If the first-day price surges significantly above the IPO range, I would wait. The post-lock-up window, which typically opens 90 to 180 days after listing, has historically offered the best entry into transformative technology companies. Discipline at that moment matters more than enthusiasm at the bell.

The Closing Argument

OpenAI is one of the most important companies in the world. It has shaped the public understanding of artificial intelligence more than any other entity. It commands roughly 900 million weekly users, generates $25 billion in annualised revenue, and is growing at a rate that has no clean historical comparison. The company has done genuinely remarkable things and may continue to do so for a long time.

But importance is not the same as investability. A company can be transformative and still be the wrong price at the wrong moment for the wrong investor. My judgement is that the IPO, if it prices anywhere near the reported $1 trillion target, leaves too little room for the buyer to be right and still be rewarded — and almost no room for the buyer to be wrong and avoid a loss.

There is a reason patient investors have historically earned the highest returns. They have been willing to be the second buyer rather than the first. They have been willing to wait until the price reflects the risks as well as the opportunities. With OpenAI at $1 trillion, I do not yet see that balance.

The question is not whether OpenAI is a great company. It clearly is. The question is whether the IPO price leaves enough room for the buyer to be right and still be rewarded — and right now, my view is that it does not.

I will watch the IPO. I will study the S-1. I will read the first earnings report. And I will remain patient. There is rarely a penalty in financial markets for being right too late. There is almost always one for being wrong too early.

Isaac Jonas is an economist and founder of Streetwise Economics. He holds dual Master’s degrees from the University of British Columbia. His weekly market analysis is published at www.streetwiseeconomics.com and on the Streetwise Economics YouTube channel. This article does not constitute financial advice.