Sep 30, 2026 · 32 min
Listen from 29:15
Listen at 29:15
Anthropic’s losses test frontier AI valuations
Anthropic’s Financials Revealed — The Losses Are Stunning
Anthropic’s filing raises a central market question: can frontier AI companies sustain enormous training costs while presenting a credible path to profitability?
- 1Anthropic’s reported losses look materially different when training costs and revenue-sharing obligations remain in view.
- 2Frontier AI valuations depend on continued spending to preserve a competitive moat, not simply on current revenue growth.
- 3Oura’s IPO delay shows how investor feedback, valuation expectations, and limited cash needs can outweigh strong operating fundamentals.
Don't miss
Paul Kedrosky argues that frontier AI companies cannot simply stop training without risking their competitive position against cheaper Chinese rivals.
The brief
Paul Kedrosky and Ed Elson examine Anthropic’s filing, arguing that adjusted profitability can obscure the training costs and revenue-sharing obligations behind frontier AI.
The valuation debate turns on a difficult tradeoff: continued training may protect Anthropic’s moat, but stopping could leave it vulnerable to cheaper Chinese competitors.
Jay Ritter explains why Oura paused an IPO despite strong fundamentals and an oversubscribed order book, emphasizing valuation expectations, feedback, and the absence of urgent cash needs.
The episode closes with Manchester City’s alleged financial misrepresentations and sham contracts, shifting the discussion from startup finance to sports governance.
What was said on this episode
23 statements · 2 positive · 17 negative · 3 mixed · 1 neutral
Two Anthropic customers account for approximately 25% of revenue.
“2 customers are on the order of 25% of revenues”
Listen at 4:10
Approximately six Anthropic customers account for 60% of revenue.
“something like 6 customers are 60% of revenues”
Listen at 4:16
Anthropic’s training costs are ordinary operating costs, not exceptional expenses.
“The training costs are just the day-to-day running of the business”
Listen at 6:49
Training costs should be included when valuing Anthropic’s earnings.
“it should be reflected in the earnings that we look at to value the business”
Listen at 7:00
Anthropic’s inference-only operations could generate several billion dollars of positive cash flow.
“it could be as high as a couple of billion dollars in positive cash flow just from inference alone”
Listen at 8:38
The first frontier-model company to stop training and focus on inference may win.
“the first frontier model company to stop training models and just do inference is probably going to win”
Listen at 9:10
Wall Street will penalize AI companies for excluding training costs.
“Wall Street's going to give them a wake-up call on that”
Listen at 9:31
A viable AI cash-flow business requires substantially lower training spending.
“I think there is a cash flow business here, but it requires far less money spent on training”
Listen at 10:24
Wall Street should punish Anthropic for excluding training costs from earnings.
“Should Wall Street punish them for it, yes.”
Listen at 10:35
Anthropic’s approximately $2 trillion valuation is excessive.
“Well, it's a ridiculous price.”
Listen at 10:44
Stopping frontier-model training would be catastrophic for Anthropic and OpenAI.
“it's actually catastrophic for them if they do that”
Listen at 12:00
China would overwhelm frontier AI firms in industrial token production.
“China crushes them with cheaper power and vastly larger industrial token production”
Listen at 12:09
Frontier AI companies lose their competitive moat if they stop training models.
“if you don't keep training, you have no moat”
Listen at 12:47
Anthropic and OpenAI will become premium performance providers rather than mass-market manufacturers.
“they become like Ferraris”
Listen at 13:13
Anthropic’s existential-risk disclosures will trigger numerous lawsuits.
“Expect a litany of lawsuits over it for exactly this reason.”
Listen at 15:14
The Anthropic IPO may not occur.
“I actually have a standing bet that it doesn't happen”
Listen at 15:42
If Anthropic’s IPO occurs, it will initially succeed before immediately declining.
“I do expect it comes out when it comes out and it's successful for sort of in a SpaceX sense of successful. And then we have the immediate slide lower.”
Listen at 16:30
A high valuation can make a strong company a poor investment.
“there's a price at which a great company is not a great investment”
Listen at 21:23
Investors are concerned that Oura faces competing personal-health measurement products.
“some investors have the concern here with the Oura Ring. Well, they've got a great product, but it's not as if there aren't any competing products”
Listen at 23:40
Most companies that pause IPOs historically never go public.
“Historically, the majority of companies that have paused their IPOs have never gone public.”
Listen at 26:08
Oura is more likely than not never to go public.
“that's the most likely outcome here”
Listen at 27:06
Manchester City’s owners illegally bought the club’s sporting success.
“if the company's owners illegally bought their way to that success”
Listen at 28:59
Rewarding fraudsters increases public distrust in institutions.
“the more that we see how our system rewards fraudsters, the more we will distrust the system itself”
Listen at 29:49
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.