“Buy out the big AI labs, OpenAI and Anthropic, altogether, for two cents on the dollar — which is generous, since they are both mostly money-losing ventures in an industry that has yet to convincingly establish profitability.”
Our reading:
OpenAI and Anthropic are mostly money-losing ventures in an industry that has not convincingly established profitability.
Assessment: Unresolved
What happened
Was it true for the period described?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The records are consistent with Marcus's description of the two companies' recent overall losses. Anthropic's profitable operating quarter and positive adjusted first-half income qualify that picture, but do not erase its net losses or automatically disprove 'mostly'. The broader industry clause remains unsettled: losses at two companies and incomplete disclosure across other labs do not establish an industry-wide result. The supplied accounting-basis rule also prevents treating these records as exact-scope decisive evidence. This is not a finding that Marcus's ordinary-language claim is false or uncheckable.
Caveats
- The reporting window and industry boundaries are evaluator interpretations.
- Substantial noncash charges complicate the relationship between reported losses and operating economics.
- OpenAI evidence relies on reporting about audited statements.
- Unknown accounting basis limits formal scoring; it is not evidence against the speaker.
Evidence (4 items)
Context only: Anthropic, PBC / U.S. Securities and Exchange Commission,
net losses
Why it matters: The filing reports net losses of $41.965 billion for 2025 and $80.748 billion for the first half of 2026, consistent with the company-loss portion of the claim. It also reports $974 million of operating profit in the June quarter and substantial noncash financing revaluations, so the net losses must not be described as equivalent cash losses or proof that every quarter's operations lost money.
- Publisher
- Anthropic, PBC / U.S. Securities and Exchange Commission
- Kind of source
- Regulatory filing, unaudited
- Measures
- Anthropic, Net loss, USD billions, gaap, 2026-01-01 to 2026-06-30
- About events on
- Published
Context only: Ars Technica,
nearly $39 billion
Why it matters: Reporting on audited statements indicates substantial company-wide losses in 2025; the article also reports that removing a large restructuring charge still leaves a loss. This is reputable secondary reporting, not a directly obtained OpenAI filing.
- Publisher
- Ars Technica
- Kind of source
- News report, audited
- Measures
- OpenAI, Reported net loss, USD billions, company reported, 2025-01-01 to 2025-12-31
- About events on
- Published
Context only: Helmbeam,
Adjusted net profit of $2.671 billion
Why it matters: Anthropic's positive first-half adjusted net income qualifies a blanket description of its economics; the amount and adjustments were checked against its S-1. It is a different accounting measure from reported net loss and does not by itself overturn a claim about a predominantly loss-making recent history.
- Publisher
- Helmbeam
- Kind of source
- Secondary summary
- Measures
- Anthropic, Adjusted net income, USD billions, non gaap, 2026-01-01 to 2026-06-30
- About events on
- Published
Context only: Banca d'Italia,
The segment of AI labs remains particularly opaque.
Why it matters: The paper explains that private labs and diversified companies often lack separately disclosed AI financial results. This establishes a measurement gap, not proof that the entire industry lacks profitable businesses.
- Publisher
- Banca d'Italia
- Kind of source
- Preprint, not peer reviewed
- About events on
- Published
Probably right?
One AI thinks this is probably right, and the other says it’s too early to tell. The evidence isn’t conclusive, so the verdict above stays unresolved. It counts only when you pick Speculative under “How sure?”.
- Anthropic’s AI: Anthropic's S-1 shows net losses every year, including 2025 and H1 2026, plus an H1 operating loss. Reporting on OpenAI's audited accounts shows a 2025 operating loss of about $21B. One profitable Anthropic quarter and its adjusted profit don't change that losses dominate, or that profitability isn't convincingly established.
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Where we had to interpret
- The factual claim concerns the companies' financial performance and their industry's demonstrated profitability. The proposed acquisition price is a recommendation, not a financial valuation being tested.
- The ordinary reading of 'mostly money-losing ventures' is that losses characterize both businesses overall. Company-wide net results are an appropriate measure. It does not mean every product, transaction or quarter loses money.
- To assess this ongoing description without making one quarter decisive, examine the latest completed calendar year and the completed first half preceding the post: January 2025 through June 2026. These dates operationalize the present-tense claim; Marcus supplied no reporting period.
- In context, 'industry' means businesses developing and selling generative AI models and services. Profits from unrelated activities of diversified technology companies do not establish that these businesses are profitable.
- 'Convincingly establish profitability' means demonstrated, repeatable profits after the costs of developing and delivering the business's products. Revenue growth, financing, positive gross margins and forecasts of future profits are insufficient by themselves.
- Marcus supplies no accounting convention or attribution to financial estimates. Record that omission as unknown without treating it as evidence against his ordinary-language claim. Examine reported net results and explain material adjustments separately.
Right if: Evidence shows that losses characterize both companies over the recent reporting history, allowing occasional profitable periods, and sufficiently broad evidence shows that sustained profitability has not been demonstrated across the relevant generative AI business sector.
Wrong if: Evidence shows that sustained company-wide profitability, rather than losses, characterizes either named company during the target period, or that sustained profitability was already convincingly demonstrated across the relevant industry. One profitable quarter or one profitable product is insufficient to overturn 'mostly'.
We wrote this test after the outcome was known.
What would settle it
Comparable company-wide financial results across the target period, plus broader evidence of sustained profitability or persistent losses among generative AI businesses, would resolve the substantive gaps. Formal supported or contradicted scoring would also require a permitted accounting-basis mapping for this ordinary-language claim.
We found 17 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, China has all but caught up. The US is not going to “win” the AI war. Here’s what we should do instead., paragraph 37.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2026-07-20.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, China has all but caught up. The US is not going to “win” the AI war. Here’s what we should do instead.
Buy out the big AI labs, OpenAI and Anthropic, altogether, for two cents on the dollar — which is generous, since they are both mostly money-losing ventures in an industry that has yet to convincingly establish profitability.
- Where
- paragraph 37
- Published
- 2026-07-20
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.