Anthropic Profitability Reports Rely on Adjusted Operating Income

Reports that artificial-intelligence lab Anthropic achieved profitability for a second consecutive quarter rely on an adjusted operating income metric that excludes significant expenses, including model training costs and partner revenue shares.

While tech companies routinely adjust operating income to exclude stock-based compensation, Anthropic's reported figures omit costs that typically form a substantial portion of an AI developer's outgoing expenditures. According to a report by Tiger Brokers, the company's stated gross margins above 80% do not account for revenue shared with distribution partners or the capital required to train frontier models.

"Reports citing multiple people familiar with the matter stated that Anthropic told investors it will profit on a measure called adjusted operating income," noted the Tiger Brokers report detailing the financial structure.

How Frontier AI Costs Contrast With Tech Norms

The exclusion of training costs highlights the unique financial architecture of generative AI companies, where capital expenditure for compute power and data acquisition rivals traditional research and development. Training advanced large language models requires massive outlays for GPUs, electricity, and engineering talent before a single commercial API call generates revenue.

For comparison, data cited in the report regarding SpaceX's AI segment illustrates the heavy cash burn associated with model development. SpaceX's artificial intelligence research and development expenses reached $2.03 billion in the second quarter, outpacing its $1.47 billion in revenue during the same period, driven largely by training costs for models like Grok and associated datacenter infrastructure.

Valuation Metrics and Private Market Speculation

Despite the narrow accounting definitions underlying its reported profitability, market enthusiasm for generative AI developers remains elevated in private and speculative trading venues. According to market data from the Tiger Brokers report, a futures contract tied to Anthropic's valuation on the Hyperliquid exchange recently implied a $2.15 trillion valuation upon a potential initial public offering.

This valuation disconnect underscores how investors evaluate high-growth artificial intelligence firms, frequently weighing forward-looking revenue projections against the near-term realities of high infrastructure costs and non-standard accounting adjustments.