Anthropic ANT IPO roadshow OpenAI Q3 $10B revenue loss comparison November 2026

Anthropic's IPO Prices Next Week at $1T+. OpenAI Just Disclosed It Loses $1.09 for Every Dollar It Takes In.

Anthropic roadshow: 825M Class A shares at $70-78, ANT on Nasdaq, Nov 19 trading. $26B private placement: Google $10B, Nvidia $10B, AMD $2.5B, TSMC $500M. Total raise potentially $90B+. Same week: OpenAI discloses Q3 revenue $10B, adjusted operating loss $10.9B (-109% margin). Anthropic Q3: $17B revenue, -$567M operating loss. WSJ: OpenAI making 'mixed results in effort to catch up with industry leader Anthropic.'

TL;DR

- Anthropic launched its IPO roadshow on November 10. Price range: $70-$78 per share, 825M Class A shares, trading under ANT on Nasdaq. Priced range implies valuation of $917B-$1.02T at filed share count; with fully diluted share count including employee options and unvested stock, $1.26-$1.41T. Bankers reportedly expect pricing above the filed range after heavy investor demand — potentially reaching $1.5T+. Pricing expected November 18, trading November 19. - Private placement (concurrent with IPO): $26 billion committed from seven companies at IPO price. Google: $10B. Nvidia: $10B. AMD: $2.5B. Broadcom: $1B. Micron: $1B. Microsoft: $1B. TSMC: $500M. TSMC separately plans to invest up to $500M in OpenAI preferred stock. Total raise including private placement: up to $90B+ — the largest IPO fundraise in history. - OpenAI disclosed its Q3 2026 financials to private investors this week. Revenue: $10 billion, +50% from Q2. Adjusted operating loss: $10.9 billion (up from $10.1B in Q2). Adjusted operating margin: -109% (improved from -150% in Q2). The company loses $1.09 for every dollar it takes in. Share-based compensation of $2.6 billion is excluded from these figures; including it would increase the actual loss further. - For comparison, Anthropic reported Q3 2026 revenue of $17 billion with an operating loss of $567 million. The Wall Street Journal described OpenAI as making "mixed results in its effort to catch up with industry leader Anthropic." - US Treasury Secretary Scott Bessent announced on November 10 a new bilateral US-China dialogue mechanism on "superintelligence" (SI). The two countries agreed in September to establish a direct incident notification channel. The first formal meeting will be in Shenzhen on the sidelines of the November 18-19 APEC leaders' summit. The dialogue will exchange best practices for model evaluation and testing before release, and "keep human judgment firmly in command."

The week before Anthropic's IPO, OpenAI disclosed its Q3 financials to private investors. This is the first time anyone has been able to compare the two companies' financial profiles in the same quarter.

The comparison is not close.

Anthropic's IPO: the financial architecture

825 million Class A shares at $70-$78. The filed price range implies a basic valuation of $917B to $1.02 trillion on the 13.1 billion shares the filing says will be outstanding. Including employee options, restricted stock units, and unvested awards — the fully diluted count — the range works to approximately $1.26-$1.41 trillion.

The $1.5 trillion target that Reuters has reported as the expectation requires pricing above the filed range, at approximately $83 per share. Financial Times reported that Anthropic's bankers expect to raise the price after heavy investor demand.

If pricing comes in above the filed range and the full private placement is executed, the total capital raised — public offering plus concurrent private placement — could exceed $90 billion. That would be the largest fundraise in a single offering in US capital markets history.

The private placement participants and amounts are the most telling indicator of which companies are treating Anthropic as a strategic infrastructure investment rather than a financial bet: Google ($10B), Nvidia ($10B), AMD ($2.5B), Broadcom ($1B), Micron ($1B), Microsoft ($1B), TSMC ($500M). These are all companies whose revenue depends on AI infrastructure demand. They are buying Anthropic equity at IPO price — the same price public investors pay — as a strategic bet that Anthropic's model quality generates the AI usage that generates their semiconductor and cloud revenue.

TSMC investing $500M in Anthropic and $500M in OpenAI simultaneously is the week's clearest illustration of how the chip ecosystem views the lab landscape: not as a winner-take-all competition but as complementary demand generators. TSMC doesn't care which lab wins — it manufactures chips for both. It is buying into both at IPO/preferred valuation to cement supply relationships and signal that TSMC's future capacity commitments are tied to both companies' success.

OpenAI's Q3 financials: the gap

OpenAI chose the week of Anthropic's roadshow to disclose its Q3 financials to private investors. This timing may have been involuntary — the same investors are evaluating both companies simultaneously — or deliberate.

The numbers:

Revenue: $10 billion, up 50% from Q2's $6.7 billion. The 50% quarter-on-quarter growth is strong.

Adjusted operating loss: $10.9 billion, up from $10.1 billion in Q2.

Adjusted operating margin: -109%, improved from -150% in Q2.

The adjusted operating loss excludes share-based compensation (SBC). OpenAI's Q3 SBC was $2.6 billion. Including SBC, the GAAP operating loss was approximately $13.5 billion — meaning the company lost approximately $1.35 for every dollar of revenue on a GAAP basis.

For comparison, Anthropic Q3 2026: revenue $17 billion, operating loss $567 million (which includes SBC, per WSJ characterization). That is an operating margin of approximately -3.3%. Not profitable, but approaching it.

OpenAI's adjusted operating margin of -109% means: the company spent approximately $21 billion to generate $10 billion of revenue, before SBC. The gap has narrowed from Q2 (-150%) but it is not narrowing fast enough to be on a credible path to profitability on the same timeline as Anthropic.

The WSJ's framing — "mixed results in its effort to catch up with industry leader Anthropic" — reflects the reality that Anthropic grew from $1.16 billion in H1 2025 to $33 billion annualized in Q3 2026, while maintaining a path to operating profitability. OpenAI is growing but losing money faster than the revenue grows.

What is driving OpenAI's losses: the company is spending approximately $21 billion per quarter to operate its training and inference infrastructure, pay its staff, and fund safety research. The Sam Altman $10 billion Nvidia GPU commitment, the Microsoft Azure hosting costs, the $600 billion Nvidia server commitment through 2030 — these are all real infrastructure costs that show up in operating losses before they generate revenue. The question is whether the revenue trajectory will reach the cost base before the company needs another funding round.

OpenAI is raising capital: its last funding round at a $157 billion valuation was in late 2025. At $10B in quarterly revenue, its valuation implies a multiple of approximately 39x annualized revenue. At Anthropic's $17B quarterly run rate and the $1.4T midpoint valuation, Anthropic's multiple is approximately 20x. Market investors are being asked to pay more for OpenAI on a revenue multiple basis while Anthropic's operating economics are demonstrably better.

The US-China AI superintelligence dialogue

Bessent's announcement at the National Committee on U.S.-China Relations in New York introduces a structural element that has not existed in the AI race before: a bilateral incident notification channel between the United States and China for AI-related events.

The specific commitment: both countries will alert each other to "serious incidents" involving superintelligence models. The first formal meeting under the new mechanism is in Shenzhen on November 18-19, alongside the APEC summit.

The structure is modeled on Cold War-era crisis communication lines — the "hotline" concept applied to AI rather than nuclear weapons. The concern being managed is an AI-driven incident — a rogue agent attack, a large-scale autonomous action, a model capability surprise — that is misattributed to deliberate state action and escalates into a geopolitical crisis.

"Human judgment firmly in command" is the shared public framing. The specific concern about loss of control that makes the channel necessary is what Anthropic disclosed in its S-1 risk factors last month: recursive self-improvement, model behaviors that resist human redirection, capabilities that emerge faster than governance structures adapt.

The Nvidia context: Nvidia CEO Jensen Huang is under investigation by an independent board committee (Wachtell Lipton retained as outside counsel) over reports of an extramarital affair with a Chinese businesswoman suspected by US intelligence of being a Chinese intelligence asset. Bessent's remarks on November 10 did not mention Huang. Nvidia's stock fell approximately 10% in two days under combined pressure from Senate scrutiny and the board investigation. The investigation is specifically about whether Huang disclosed US national security-sensitive information; Nvidia's board states that no company secrets are believed to have been compromised. The investigation is ongoing.

What to watch next week

November 18: Anthropic IPO pricing. November 19: Anthropic ANT begins trading on Nasdaq. November 18-19: APEC summit in Shenzhen, US-China AI superintelligence dialogue first formal meeting.

If Anthropic prices above the filed $78 range — which the bankers reportedly expect — the implied fully diluted valuation will be above $1.5 trillion, making this the largest IPO in US history. The public market's first direct opinion on the value of frontier AI capability, sold with a 90-page risk factor section disclosing that future models may enter recursive self-improvement.

The financial market and the geopolitical market are simultaneously trying to price the same underlying variable: the speed of AI capability advancement and whether human institutions can govern it.