#Anthropic盈利增速放緩 The competitive dynamic has begun to change. In the next phase, AI companies will shift from “burning money to scale first” to focusing on whether the money they burn can be recovered faster.

Anthropic generated $16.8 billion in third-quarter revenue, up approximately 46% quarter-on-quarter, while achieving adjusted profitability and positive cash flow. The problem is that growth has clearly slowed: monthly growth has fallen from double digits to single digits, and the target of reaching $100 billion in annualized revenue by year-end also appears challenging[3]. The market is starting to use a different yardstick. Previously, it focused more on users, token usage, and annualized revenue. Now, a more realistic set of questions has emerged:
How much revenue can each dollar of compute investment generate?
After API prices are cut, can increased usage make up for the lost revenue?
Will enterprise customers continue renewing?
Can operating profit and free cash flow improve?
So, the real signal from Anthropic this time is that adjusted profitability in a single quarter is merely the entry ticket; the quality of growth is the key to valuation.

The pressure differs across companies
Independent model companies are under the most strain. Training, inference, R&D, and customer acquisition all cost money, while open-source models are pushing API prices lower. More usage does not necessarily mean more profit.
Large platforms have more room to maneuver. They can integrate AI into cloud services, advertising, office software, and e-commerce, so their models do not necessarily have to generate profits on their own.
Alibaba’s AI cloud and computing services generated 48.44B yuan in revenue this quarter, with adjusted EBITA of 5.63B yuan, indicating that some cloud and computing businesses have already begun contributing profits.

How things will change next
Expansion will continue, but project selection will become more stringent. High-cost, low-paying general-purpose applications will come under pressure, while enterprise coding, workflow automation, industry models, and cloud computing will be more favored by capital. Financing will also become more difficult, with investors paying greater attention to the speed of cash recovery rather than the scale of the story alone.
In one sentence: The AI industry has not stopped burning money; it is shifting from “who burns the most” to “who burns most effectively.”