Are AI Companies Entering the Era of Profitability?

This article is part of the PRO Members Communication, and follows "PRO Members of Machine Heart" at the end of the article to read more on the topic.

Over the past few years, leading-edge model companies have generally suffered high losses, with capital markets continuing to bleed, computing, infrastructure, and the triple-pressure of commercialization, and model companies relying solely on API calls or C-end subscriptions, often unable to cover high operating costs. Recently, Anthropic disclosed its first single-season earnings in its public offering, prompting a heated market debate focusing on "can Anthropic's profitability continue" and "will AI enter the second half of profitability?"

Directory

01. What does Anthropic's profitability mean?
What is Anthropic doing to make money and why is it so common for AI companies to lose money?...

What does Anthropic's profitability mean?

1 Anthropic's recently disclosed IPO shows that the company achieved a first quarter GAAP operating profit of $974 million in the second quarter of 2026, becoming the world's first leading-edge model company to disclose a quarterly operating profit in its official financial statements, sparking industry heated debate.[1-1]

1 Anthropic reported quarter-on-quarter revenue of $115.55 billion, operating profit of $974 million, and adjusted operating profit of approximately 19% for the second quarter of 2026. Net account loss for the same period was $80.748 million, of which more than $7.2 billion came from changes in the fair value of convertible debt and derivatives.

2 Following the elimination of non-cash accounting losses resulting from an increase in valuation, Anthropic posted net profit after adjustment for the first half of the year of $26.71 billion and operating cash flow of $3.9 billion.

2. Over the past few years, model companies have been widely seen as "black holes" that swallow capital, and concerns about the bursting of the "AI bubble" in capital markets have continued to spread. Anthropic's profitability has been seen as a landmark milestone in the industry, proving that Scaling Laws have been established not only in improving the performance of model technology but also in the economies of scale business.[1-2]

1 As the model size and technology matured, the unit reasoning cost decreased significantly, while the extremely strong model performance generated an explosion of Agent (smart body), which led to an exponential increase in the total number of Token calls, which allowed stronger models to gain premium, super-linear growth in the consumption of enterprise customers, and the reasoning margin rose with scale.

3. In recent years, LLM providers have generally been in a long-term state of loss. The industry attributes the previous industry's loss to the high cost of computing, infrastructure investments and fragile business models, and believes that it will be difficult for model companies to get rid of their reliance on continuous outflows of external capital in the short term.

1 In terms of computational costs, the overlap between the cost of model training and the cost of reasoning makes computational costs account for more than 50% of the operating expenses of most manufacturers.[1-3]

2 The rigid capital commitments of long-term arithmetic and the pressure to support the continuous rise and rise of infrastructure costs are widespread in infrastructure, and it is difficult for model companies to quickly cover long-term compliance expenses even with increasing revenue. [1-4]

3 In addition to computing power and rigid infrastructure spending, the low-price competition and capital subsidy strategies adopted by model companies in the early years of commercialization to compete for market share created a contradiction between revenue growth and gross overhead, the higher the call volume of models, and the greater the cumulative consumption of reasoning costs that exceed single token revenue.[1-5]

4 With the IPO, the global AI community has begun to focus on reviewing the business model of Anthropic. However, there is still no consensus in the industry as to whether Anthropic has continued profitability across cycles and whether the case could be a precursor to profitability for large model manufacturers collectively.

Are LLM providers entering the second half of profitability?

1. With the industry's gradual analysis of Anthropic's initial public offering, the industry believes that Anthropic's earnings of $1.62.86 billion in the first half of the year benefited from three aspects of "consumption-based pricing", "focused on high-net-worth enterprise customers with high retention rates", and "the deep channel distribution strategy of the debt cloud giant".

1 Anthropic's charging structure is based on mass-charging, and the book of common stock shows that Anthropic earns about $1.31 billion in mass-charging revenue, accounting for about 80% of total revenue.[1-1]

2 Anthropic enterprise customers have a net revenue retention rate of about 900%, which is about nine times the current quarter's revenue contributed by the same group of qualified enterprise customers compared to the same period a year ago; among enterprise customers using Claude Code, the highest monthly spending of 5% is about 20 times that of non-Code enterprise users.[1-1]

3 In the first half of 2026, sales generated by Anthropic through AWS, Google Cloud and Azure, the three largest cloud markets, accounted for 48% of total revenues (approximately $7.8 billion), up significantly from 11% in 2023.[1-1]

2 Although Anthropic has advantages in revenue structure, channel distribution and reasoning efficiency, there is considerable disagreement about whether the market will be able to generate sustained cross-cycle profitability in the future and whether its profitability path will be replicated by other model companies.

3 Optimistic analysis suggests that Anthropic's billing model and the burst of Agent demand form the "Javens Paradox" of positive feedback that even if the model pricing drops, the total use and revenue of the model will increase dramatically.