Deep dive: How a Convertible Note Creates a $72B Paper Loss

Why today

Reuters reviewed Anthropic's draft prospectus for its planned stock market listing and reported on October 21 that the company lost US$80.7 billion in the first half of 2026. That number travelled fast. Most of it is not cash. The loss includes a $72 billion accounting charge. Reuters describes it as reflecting an increase in the estimated value of bonds that could eventually convert into Anthropic shares. The timing matters because the listing is expected in November 2026. These are the first detailed half-year figures to come from a filing rather than from investor briefings. A week earlier, reports based on what Anthropic told investors put third-quarter revenue at $16.8 billion. So in one week we got a huge loss and a fast-rising top line. Today we stay narrow. How does a convertible note produce a loss this big, and which numbers deserve the attention instead?

Plainly

Start with the instrument. A convertible note is a loan that can turn into shares. The lender hands over cash today. Instead of expecting to be repaid, the lender usually expects to convert into stock on terms set when the note was written. Now the accounting. US rules give the company that issues the note two broad routes. It can split out the conversion feature and treat it as a derivative. Or it can choose the fair value option. In the words of the advisory firm Stout, the issuer then "measures the entire hybrid debt instrument at fair value each reporting period." Either way, changes in that value run through earnings. Here is the counterintuitive part. The note is a liability, something the company owes. If the note converts into a fixed slice of the company, that slice grows in value as the company does. A bigger liability means a loss on the income statement. Good news for the share price shows up as bad news in net income. A worked example. No cash moved. The product, the customers and the servers are unchanged. The loss records that the noteholder's claim on the company got more valuable. There is a public mirror image. On the holder's side that is a gain. On the issuer's side the same rise in value is a loss. We don't know from the reports how much of the $72 billion relates to Amazon's notes. Reuters reports that the estimated fair value of the notes tripled year over year as Anthropic's valuation climbed. That rise is what produced the $72 billion.

What the evidence shows

First, what is established, and how. The core figures come from a draft prospectus that Reuters reviewed. That is a primary document seen through a news report, not a public filing we can read ourselves. Draft numbers can change before the final version. Per that report, the net loss was $80.7 billion, including the $72 billion charge. A year earlier, the first-half net loss was $13 billion. For all of 2025, the net loss was $42 billion, and Reuters says most of it came from a similar charge. Strip that out and the 2025 operating loss was about $8.1 billion, on revenue of $4.6 billion. Spending on computing capacity and infrastructure was nearly $9 billion in the first half, up from $2.3 billion a year earlier. The company has committed to spend at least $518 billion over a decade on cloud, compute and infrastructure. Now the gaps. It can include interest, taxes and other items that sit below the operating line. The reports we found do not give a first-half 2026 operating loss. They do not give an inference gross margin, meaning revenue minus the cost of serving the models. Next, what is claimed but not shown. The quarterly revenue figures come from what Anthropic told investors, as reported by the Financial Times. They are self-reported. On those reports, revenue went from $4.7 billion in the first quarter to $11.6 billion in the second. The third quarter reached $16.8 billion. Anthropic also said it was profitable on an adjusted basis in the third quarter. The Next Web put the problem plainly: None of the reports say which costs the adjusted measure leaves out, and no profit figure was given. So the evidence is lopsided. The accounting mechanism is clear and well documented. The size of the mark is reported from the draft filing. The underlying cost structure is still mostly missing.

How people read it

The same facts support several honest readings. The AGI-forward reading says the charge is a side effect of success. The notes became expensive because the market's view of the company tripled. Revenue is compounding fast, and the cash loss looks modest next to quarterly revenue in the tens of billions. On this view, the right response to the $80.7 billion is a shrug. The AI-realist reading agrees the mark is noise but looks at what's left. Compute and infrastructure spending nearly quadrupled year over year. Then there are commitments of at least $518 billion stretching over a decade. Revenue is rising, but so are the obligations. Adjusted profitability without a definition proves little. The market reading adds a point both readings can miss. The mark is non-cash, but it is not free. It measures value moving to noteholders. When the notes convert, those holders own a larger share of the company. Existing shareholders and new buyers at the listing own correspondingly less. Strip out the charge for operating analysis, then add the dilution back when you value the shares. The policy angle is narrower. A draft prospectus goes through securities-regulator review. Adjusted profit measures in public filings must be defined and reconciled to the standard accounting figures. That process should turn several of today's open questions into disclosed numbers.

What it means for builders and investors

Our read: the $80.7 billion headline tells you about the capital structure, not the business. We would set it aside and anchor on three other lines. First, the operating loss, once it is disclosed for the half. Second, inference gross margin, because that tells you whether each new dollar of revenue is profitable to serve. Third, compute commitments set against revenue, because $518 billion of promises shapes the company's flexibility for a decade. This matters for late-stage pricing, and it is relevant at seed and Series A as well. The same mechanism applies to any company carrying convertible instruments at fair value. An up-round can make reported net loss jump while nothing operational changes. The reverse also holds. A falling valuation can flatter net income. Founders and investors reading early-stage statements may see the same distortion at smaller scale. One question worth exploring in diligence might be this. When a company reports adjusted profitability, which costs sit outside the adjustment? How much of its compute is committed rather than used? We would hold this view provisionally. A disclosed gross margin that holds up as usage grows would make the bullish case much stronger.

What to watch

Three things would settle most of this. First, the public prospectus. We'd look for the operating loss line, cost of revenue and the footnote on the notes. That footnote should show the conversion terms and whether the notes convert at the listing. Second, the listing price itself. If the notes are marked again before conversion, a price above the last mark could mean another charge. That is a possibility, not a reported fact. Third, the first quarterly report as a public company. That report should define the adjusted measure and give a gross margin we can track. The listing is expected in November 2026, so the first two could arrive within weeks.