The Valuation Event

0:00 Edmund Hale — This is The Fathom Report, extended edition, with Edmund Hale.

0:04 Edmund Hale — A word before we start. This edition draws on The Leadsman, the analytical newspaper. Where something is its judgement rather than established fact, we will say so.

0:14 Edmund Hale — The headlines. Anthropic has filed to go public, and its finances are visible for the first time. Israel voted with neither bloc near a majority. And China's ruling party met to discuss discipline.

0:25 Edmund Hale — After three short reports, our editor, Rose, joins me to take the first story apart: a stock market filing with one very large number on the front and a much larger one underneath.

0:35 Edmund Hale — First, Israel. In a report published on election day, the paper described the sixth national election since twenty nineteen, and the first in thirty-eight years to run its full parliamentary term.

0:46 Edmund Hale — Its final polls put Gadi Eisenkot's bloc at about fifty-four seats and Benjamin Netanyahu's at about forty-eight, with sixty-one needed. It forecast a hung parliament. That was a forecast, so check the count itself.

0:59 Edmund Hale — Next, China. According to the reporting, the Communist Party's fifth plenum opened on the twenty-sixth of October, a four-day meeting on party discipline, and the last before the congress expected in autumn twenty twenty-seven.

1:12 Edmund Hale — Ahead of the closing communique, it predicted language on high-quality development and technological self-sufficiency, and no major personnel announcements.

1:21 Edmund Hale — And in Europe, the analysis looked at France, where the October flash survey put activity below the fifty mark separating growth from contraction, for a third quarter running. Germany stayed in growth.

1:33 Edmund Hale — Ahead of the European Central Bank's meeting on the twenty-ninth, the reporting had markets pricing a hold at two and a half percent. Its point: the bank must explain a hold with inflation at three point eight percent.

1:45 Edmund Hale — Now to the story at the top of the bulletin. With me is our editor, Rose. Rose, welcome.

1:51 Rose Calloway — Thank you, Edmund. Happy to be here, and happy to be discussing something with this many zeros.

1:55 Edmund Hale — The facts first, as the reporting gives them. Anthropic filed its registration statement on the twenty-sixth of October, planning to list on the Nasdaq under the ticker A N T. No share price had been disclosed.

2:07 Edmund Hale — What is the number everyone is quoting?

2:09 Rose Calloway — Revenue for the first half of this year: sixteen point two nine billion dollars. A year earlier: one point one six billion. The article gives that as an increase of one thousand three hundred and four percent.

2:19 Edmund Hale — Which sounds like a typing error.

2:21 Rose Calloway — It does, so I checked, and this is the station's own arithmetic. Sixteen point two nine over one point one six is just over fourteen. That is an increase of just over thirteen hundred percent. The sum holds.

2:31 Edmund Hale — And the quarterly picture?

2:33 Rose Calloway — That is where I would look first. The article's sources give eleven point six billion dollars for the second quarter of this year. Subtract that from the half-year total and the first quarter was about four point seven billion.

2:44 Rose Calloway — So the second quarter was up a hundred and forty-seven percent in a single quarter. The article offers three explanations: company contracts, government contracts, and software teams. It says the filing cannot tell them apart.

2:55 Edmund Hale — Why does it matter which?

2:57 Rose Calloway — Because, as the piece notes, government contracts often arrive in large, lumpy tranches. If a good part of that quarter was a few big payments, it is not a run rate. It is a very good quarter.

3:06 Edmund Hale — Now the figure that grabbed headlines in the other direction. A net loss of eighty point seven four eight billion dollars. Eighty billion, lost in six months.

3:15 Rose Calloway — Which is the reading the headline invites, and the article says it is wrong. The operating loss was six hundred and ninety-three million dollars. Almost all the rest is a charge that involves no cash.

3:25 Edmund Hale — Explain that charge to someone in a car.

3:27 Rose Calloway — The company has issued bonds and preferred shares that convert into equity. When it is valued higher, those rights are worth more, and the accounting rules book the increase as a loss. The better it does, the bigger the number.

3:38 Edmund Hale — Then I would go further than the paper. A charge that grows whenever the valuation grows is a distraction, and anyone who stops at eighty billion has been misled.

3:48 Rose Calloway — I would push back. Not cash does not mean not real. Those rights are claims on future shares, which is dilution, and the piece calls it theoretical dilutive value.

3:56 Rose Calloway — It is not money leaving the bank, but it matters to anyone who owns the stock.

4:00 Edmund Hale — But dilution is a cost a buyer prices in. It does not mean the business is failing.

4:04 Rose Calloway — I concede that. The analysis puts the operating loss at about four point three percent of revenue and says that is not a distressed profile. On the evidence it gives, that is fair.

4:14 Rose Calloway — My remaining quarrel is narrower. The article treats the operating loss as close to cash burn, and it cites no cash flow figure. I would want one before treating the two as the same thing.

4:23 Edmund Hale — Margins, then. Cost of revenue was two point three eight billion dollars against sixteen point two nine in revenue. A gross margin of about eighty-five percent.

4:33 Rose Calloway — Yes. But compute and infrastructure spending in the half-year was eight point nine five billion, nearly four times the cost of revenue. That comparison is mine, not the article's.

4:42 Rose Calloway — It suggests much of the computing bill sits outside the cost of serving customers. The article does not break that down, and an eighty-five percent margin depends on where that line is drawn.

4:52 Edmund Hale — Then the largest figure in the piece. Four hundred and seventeen billion dollars of contracted computing capacity.

4:58 Rose Calloway — Contracts beginning in twenty twenty-six or twenty twenty-seven and running from under a year to as much as twenty years. The Leadsman assesses, with high confidence, that the counterparties are Google, Amazon and Microsoft.

5:10 Rose Calloway — And it says the figure is about twenty-five times first-half revenue, annualised. I think that is a slip.

5:16 Edmund Hale — You are checking the paper's arithmetic.

5:18 Rose Calloway — Yes, and this is the station's own sum. Its valuation section annualises first-half revenue at thirty-two point six billion. Four hundred and seventeen over that is nearer thirteen, not twenty-five.

5:28 Edmund Hale — Perhaps it uses a different base.

5:30 Rose Calloway — Perhaps, but it does not say, and the base it uses elsewhere gives thirteen. Thirteen years of revenue is still an enormous promise. But it is an error, and I would flag it.

5:38 Edmund Hale — Let me press that. Thirteen years of revenue, against contracts running up to twenty years, does not sound like a cliff.

5:44 Rose Calloway — Fair challenge. Spread evenly over twenty years, it is a little under twenty-one billion a year, well below annualised revenue. But that is the best case. Many contracts are shorter, and the article gives no schedule.

5:56 Rose Calloway — And the paper's own arithmetic points the other way. Compute spending is tracking to about eighteen billion a year, which would take over two decades.

6:03 Rose Calloway — So spending has to rise. I concede the total alone is not the problem. The schedule is.

6:08 Edmund Hale — The article's bottom line goes further. It says the industry has crossed the revenue threshold, and the question now is servicing the infrastructure.

6:16 Rose Calloway — Crossed it for one company, for six months, and this is the first filing of its kind, so there is nothing to compare it with.

6:22 Edmund Hale — And where does the article get its numbers?

6:25 Rose Calloway — From press coverage of the filing, including CNBC, Yahoo Finance, Fortune and BetaNews, rather than the filing itself. A slip anywhere in that chain travels with the number. I would read the filing first.

6:35 Edmund Hale — Now the price. The piece cites implied valuations of one point four three to one point eight eight trillion dollars, from outside estimates.

6:43 Edmund Hale — The last private round, back in May, put the figure at nine hundred and sixty five billion dollars.

6:49 Rose Calloway — Those are estimates, not an offer price. The analysis then applies its own multiples of annualised revenue.

6:54 Rose Calloway — Forty times thirty-two point six is one point three trillion. Fifty times, one point six three. Sixty times, one point nine six.

7:01 Rose Calloway — It sets those against Microsoft at about twelve times, Google at seven, Nvidia at twenty-five.

7:05 Rose Calloway — It adds that OpenAI trades privately at fifty to eighty times, a figure no public filing verifies. So the top of the range leans on an unverified comparison.

7:13 Rose Calloway — Then the base. Those multiples double the first half. The second quarter alone, times four, is about forty-six billion, and that is my arithmetic, not the paper's.

7:22 Rose Calloway — At one point six three trillion, that is about thirty-five times instead of fifty.

7:26 Edmund Hale — Which makes the price look cheaper.

7:28 Rose Calloway — Only if the second quarter repeats, and we agreed it may not. The valuation question is really a question about which quarter is the typical one.

7:34 Edmund Hale — The article points to a Forbes analysis headlined as the biggest risk to the artificial intelligence boom. What is the risk?

7:42 Rose Calloway — Not that revenue stops growing. It is whether four hundred and seventeen billion in commitments gets absorbed by growing revenue, or turns out to be overcapacity that no revenue could justify.

7:52 Rose Calloway — The paper says the filing poses that question and does not answer it.

7:55 Edmund Hale — The paper makes two dated predictions. First, a first-day close between one point six and two point one trillion dollars, at moderate confidence, with a nervous stock market its stated way to be wrong.

8:07 Edmund Hale — Second, that OpenAI will disclose an intention to list, or a concrete process towards one, within ninety days of Anthropic's listing. The reasoning is competitive pressure, since a rival now has public financials.

8:19 Edmund Hale — If you had to put the piece in a sentence?

8:21 Rose Calloway — The revenue looks real, the headline loss is mostly accounting, and the commitment is the one to watch. Those are the article's conclusions and, with the corrections I have made, mine.

8:31 Edmund Hale — Rose, thank you.

8:32 Rose Calloway — Thank you, Edmund. Mind the zeros.

8:34 Edmund Hale — That was The Fathom Report, extended edition. Our thanks to Rose, and to The Leadsman, whose analysis we drew on throughout. This is Fathom FM.