Anthropic’s $80.7 Billion Loss Is Mostly Non-Cash. Alphabet’s $9 Billion Cash Burn Is Not. Here’s the Difference Your Business Needs to Know.

In the same week, two of the biggest names in AI published numbers that look backwards. Anthropic filed to go public and showed an $80.7 billion loss for the first half of the year. Alphabet reported $26.2 billion in net income for the quarter and its stock fell about 10 percent. The loss that scared nobody was the big one, and the profit that scared investors was the small one. The reason is the oldest lesson in running a business, and it applies to you at any size: profit on paper and cash in the bank are different things.

Two Headlines That Look Backwards

Anthropic filed its S-1 on October 26 to list on Nasdaq under the ticker ANT. It reported about $16.3 billion in first-half revenue, up from roughly $1.2 billion a year earlier, and a net loss of $80.7 billion. About $72.4 billion of that loss is a non-cash fair-value charge on convertible notes and similar instruments. No money left the company for that line. It is an accounting mark that moves with the company’s valuation.

Alphabet reported on Wednesday that its third-quarter net income was about $26.2 billion. But it spent $51.4 billion on capital projects, mostly AI infrastructure, against $42.3 billion of operating cash flow. That left free cash flow of negative $9.07 billion, its second negative quarter in a row, against positive $24.5 billion a year earlier. Revenue and earnings also came in slightly under forecasts. Shares fell about 10 percent after hours. Microsoft reported positive free cash flow of $18.6 billion on the same night, down 27 percent, and its shares rose.

What Is Paper and What Is Not at Anthropic

Strip out the non-cash charge and the picture changes. The filing shows an operating loss of about $693 million for the half, down from about $2.1 billion a year earlier, and an operating profit of about $974 million in the second quarter, the first in the six quarters it shows. According to analyses of the filing, operating cash flow was positive by about $3.9 billion. One analysis also finds that compute spending fell from about 201 percent of revenue a year ago to about 55 percent. Those are the unit economics we wrote about on September 8.

That does not make the risks small. The filing describes long-term compute commitments in the hundreds of billions of dollars, with outlets citing figures from roughly $450 billion to more than $500 billion, most of it non-cancelable. Those costs arrive on a schedule whether or not revenue keeps growing. Reports also differ on customer concentration, with one saying about 20 percent of first-half revenue came from two customers and CNBC quoting the prospectus saying no single customer exceeded 10 percent over nine months. We use Claude every week, and that does not change how we report any of this.

On August 18, we wrote that Anthropic had just turned a profit. The filing shows what that meant. It was an operating profit in the second quarter, sitting beside a huge net loss made mostly of non-cash charges. Both statements are true, which is exactly why you have to ask which number someone is quoting.

What Went Wrong at Alphabet

Alphabet is hugely profitable. The problem is that its spending is outrunning its cash. Investors treat free cash flow as the line between a company funding its own ambitions and one that needs outside money. Alphabet kept its 2026 spending guidance at $195 billion to $205 billion and said 2027 spending will increase significantly. Markets did not punish the earnings number. They punished the cash.

Read Your Own Business the Same Way

Four numbers do most of the work. The first is your cash balance and how many weeks of expenses it covers. A profitable month with unpaid invoices can leave you unable to make payroll, and a loss caused by a one-time or non-cash item can look scary without threatening you at all.

The second is the gap between profit and cash. Large purchases, unpaid invoices, and annual prepayments all move cash without moving profit. Ask your bookkeeper where the two numbers differ and why.

The third is your fixed commitments. List the contracts, leases, retainers, and annual software subscriptions that do not shrink if sales slow, and note when each one ends. This is the small-business version of take-or-pay compute.

The fourth is concentration. If two or three clients make up a large share of your revenue, you have the same risk the filing flags, and it matters whether they are on contracts.

Add a fifth if you can. Know what it costs you to deliver one unit of what you sell, whether that is an hour, a project, or a customer served.

What to Do This Week

Put the numbers on one page. Write down your cash balance, what you expect to come in and go out over the next 90 days, your fixed commitments with end dates, your top three clients as a share of revenue, and your cost per unit. If you cannot fill in a line, that is the line to fix first. This is not accounting advice, so go over it with your bookkeeper or accountant.

The Honest Takeaway

Nothing here is investment advice, and the numbers in the coverage differ by outlet, so check the filings yourself before relying on any single figure. The lesson does not depend on who wins the AI race. The headline number is rarely the number that decides whether a business survives. Cash, commitments, and concentration usually are.

---

If you want the unit-economics side of this, this connects directly: Anthropic’s IPO Investors Want Revenue Per Token. Here’s the Founder-Sized Version of That Question.