One-sentence summary
When a company losing $10.9 billion in a quarter at a -109% adjusted operating margin raises $15 billion at a $1.25 trillion pre-money valuation, the anchor is no longer profit — it is future share of the inference market plus the scarcity of the listing window. Understanding that pricing means laying out four variables, not applying a P/E.
1. Aligning the two sets of numbers
Two sets of numbers appeared this week.
The first came from financing: on November 12, 2026 (US Eastern), CNBC reported, citing sources, that OpenAI had raised $15 billion in its latest private round at a $1.25 trillion pre-money valuation. The round has not formally closed, is expected to close in coming weeks, and was described as an opportunistic raise ahead of a potential 2027 IPO.
The second came from operations: the Wall Street Journal reported Q3 2026 revenue of about $10 billion, up nearly 50% sequentially, an operating loss of about $10.9 billion versus about $10.1 billion in Q2, and an adjusted operating margin improving from about -150% to about -109%.
Put the two together and the ratio defies conventional financial sense: a company losing $10.9 billion in a quarter valued at $1.25 trillion. Annualizing revenue at $40 billion and losses at $43.6 billion, the valuation can be explained neither by P/E nor by P/S — price-to-sales is about 31x, and the loss is 1.09x revenue.
| Dimension | Figure | Basis and source | Note |
|---|---|---|---|
| Round size | $15 billion | CNBC, citing sources | Above the ~$11B previously reported |
| Pre-money valuation | $1.25 trillion | Same | Explicitly pre-money |
| Simple post-money | About $1.265 trillion | Pre-money plus round size | Ignores instrument structure and preferences |
| Q3 revenue | About $10 billion | WSJ via CNBC | Up nearly 50% sequentially |
| Q3 operating loss | About $10.9 billion | Same | Above Q2's ~$10.1B |
| Q2 adjusted operating margin | About -150% | Same | — |
| Q3 adjusted operating margin | About -109% | Same | 41pt improvement |
| Annualized revenue (rough) | About $40 billion | Quarter × 4 | This article's calculation |
| Price-to-sales (rough) | About 31x | $1.25T ÷ $40B | This article's calculation |
| Anthropic IPO valuation target | Up to about $1.4 trillion | Prior Reuters reporting | Nasdaq debut planned Nov 19 |
| Anthropic adjusted margin | About +15% | CNBC program commentary | Different basis from OpenAI |
So the question is not whether the valuation is reasonable, but what it is pricing.
2. Pricing variable one: share, not profit
The first variable supporting a trillion-dollar valuation on losses is expected market share.
AI model economics follow a clear pattern: training costs come first, inference revenue comes later. A company in the top tier of model capability does not earn current profit; it earns the distribution rights on inference requests for years to come. Those rights are entry-level across search, office productivity, coding and customer service.
A blunter analogy: this is capitalizing the toll on infrastructure. The road is built and traffic direction is set, even if the toll has not been fully collected.
That is why a widening loss does not automatically lower the valuation: if the loss is spent on compute and R&D, it buys future toll rights; if it is spent on customer subsidies, the character is different. Separating the two requires a cost breakdown, which OpenAI does not publish.
3. Pricing variable two: the scarcity of the listing window
The second variable is time.
Anthropic plans to list on Nasdaq on November 19 with an IPO valuation target of up to about $1.4 trillion. That is the first genuine public anchor.
For OpenAI this means two things:
First, a reference point is about to exist. Until now every judgment about what an AI company is worth has been a private-market figure without comparables. After Anthropic lists, the market gets statements under one disclosure regime, giving OpenAI's $1.25 trillion a relative coordinate.
Second, the window carries a time cost. If Anthropic trades strongly, OpenAI's 2027 pricing room widens; if weakly, the window may tighten. Facing that uncertainty, banking $15 billion at $1.25 trillion trades price for certainty.
That is the most notable feature of this round: the price did not rise, but the size rose about 40% (from roughly $11 billion reported earlier to $15 billion). A company at the very top of the field chose to trade on allocation rather than on price.
4. Pricing variable three: the structure of the cash
The third variable is rarely discussed but directly determines valuation quality: what kind of money the $15 billion is.
If it is common equity, post-money is $1.25 trillion plus $15 billion, about $1.265 trillion. If it includes convertible instruments or preferred-style structures, dilution and liquidation preferences create a gap between paper valuation and common-share value.
Public information currently stops at "pre-money $1.25 trillion, $15 billion raised." The instrument structure is not disclosed, which means the comparability of the $1.25 trillion figure depends on the securities terms behind it — and in private markets terms often say more about true pricing than the headline number.
5. Pricing variable four: the composition of the loss, not its size
The fourth variable is loss quality.
A $10.9 billion quarterly operating loss can be built from three very different things:
- Compute and data center spend: the expensed portion of capital investment, buying capacity;
- R&D spend: buying future model capability;
- Customer acquisition and subsidies: buying current users.
The valuation implication differs completely. The first maps to infrastructure assets and can be compared on EBITDA multiples; the second maps to a research pipeline and can be valued as pipeline; the third resembles consumer-internet subsidy wars, with the weakest valuation logic.
OpenAI does not publish a cost breakdown. Industry common sense suggests the bulk is the first two, which is the premise behind a trillion-dollar valuation. If that premise is falsified, the valuation logic changes immediately.
6. Why Anthropic's +15% cannot be compared directly
CNBC's program noted Anthropic's adjusted margin at about +15% against OpenAI's -109%. The 124-point gap looks striking, but direct comparison fails for three reasons.
First, different adjustments. The two companies adjust for stock-based compensation, charitable contributions and other items differently. Adjusted margin is a company-defined metric, not an accounting standard. CNBC's reporter flagged this explicitly on air.
Second, different business mix. Anthropic's revenue is more concentrated in APIs and enterprise contracts; OpenAI's includes large consumer subscription and advertising-style businesses. Cost structures, gross margins and growth curves all differ.
Third, different disclosure completeness. Anthropic is in an IPO process with fuller disclosure; OpenAI remains private and outside observers receive fragments via media.
The conclusion: these two margins are not evidence of which company is healthier, only a signal that the two are at different commercialization stages. Real comparability arrives when both report under the same rules.
7. Three testable implications
If the framework above holds, three observable signals should appear in coming months.
Signal one: Anthropic's IPO price and first-day performance. If it prices at $1.4 trillion and the market accepts it, public markets endorse the "losses for share" narrative; if the range moves down or the stock breaks issue, OpenAI's 2027 window gets repriced.
Signal two: whether OpenAI raises again. The round is described as "not formally closed." A final size above $15 billion would indicate demand stronger than expected; below would indicate resistance to the price.
Signal three: the direction of the loss ratio. From -150% to -109% is a clear improvement. If it keeps moving toward -80%, scale effects are appearing; if it stalls near -109%, the cost structure still lacks leverage.
8. Boundaries and inferences not to draw
- Round size and valuation come from CNBC citing sources; OpenAI has not confirmed publicly, and final closing figures may change;
- $1.25 trillion is pre-money; post-money is about $1.265 trillion by simple addition, ignoring possible instrument structures and preferences;
- Q3 revenue of $10 billion, loss of $10.9 billion and the -109% margin come from the Wall Street Journal via media relay, not company statements;
- "Price-to-sales about 31x" and "loss at 1.09x revenue" are simple calculations by this site from public figures, not company-disclosed metrics, and ignore seasonality;
- OpenAI does not publish a cost breakdown; the assessment of loss composition is industry inference, not fact;
- Anthropic's +15% and OpenAI's -109% use different bases and cannot be compared directly;
- This article makes no judgment on valuation fairness, forecasts no IPO price, and is not investment advice.
9. Sources
- CNBC original report: Kate Rooney, "OpenAI raises $15B at a $1.25T valuation, sources say" (November 12, 2026, US Eastern);
- Caixin: OpenAI raises $15 billion to prepare for a 2027 IPO (November 13, 2026)
- Related on this site: OpenAI's latest funding round and valuation data (November 3, 2026), Anthropic vs OpenAI: IPO valuation and funding comparison (November 3, 2026), How to read AI funding and valuation rounds (November 4, 2026)