Anthropic's IPO Math Just Got Harder: A $567 Million Loss After a $974 Million Profit
Anthropic's amended IPO filing shows a $567 million operating loss in Q3, reversing a $974 million operating profit in Q2. Sales grew to roughly $16.96 billion but at a slower clip, and that gap is what Nasdaq investors will price on debut day.
Anthropic swung to a $567 million operating loss in Q3. Three months earlier it posted $974 million in operating profit. That's a $1.54 billion reversal, disclosed in an amended IPO filing ahead of the company's Nasdaq debut, and it's the line item that will shape how the deal gets priced.
Revenue didn't collapse. Sales came in around $16.96 billion, but growth slowed. The top line is still climbing, just not fast enough to cover what's happening underneath it. Compute costs aren't falling. Neither is the price of the researchers building Claude.
Here's what matters: Anthropic's pitch has always been two things at once, frontier safety and enterprise-grade margins. A quarter like this tests the second half of that story, and it tests it in public, with a prospectus attached.
So which quarter is the real Anthropic? The $974 million profit or the $567 million loss?
My read: the Q3 loss is closer to the truth, and the Q2 profit was the outlier. One strong quarter of operating income doesn't establish a trend, especially at a company still building out training capacity at a pace that eats cash. The numbers tell the story, and they say costs are scaling faster than revenue right now.
But context cuts both ways. A loss driven by a deliberate ramp, new data centers, a hiring wave, a major model launch, is an investment. A loss driven by unit economics is a problem. Investors won't get a clean answer from a headline figure.
From a risk perspective, the number I'd hunt for is gross margin split by product line. API usage and enterprise contracts carry very different cost profiles, and blending them hides where the pressure actually sits. If the filing gives you that breakdown, read it before you read the revenue chart.
Watch the Q4 figures and whatever management says about capital spending on the roadshow. That's where this thesis lives or dies.
