-
Cryptocurrencies
-
Exchanges
-
Media
All languages
Cryptocurrencies
Exchanges
Media
Share
Author: Zhao Ying, Wall Street Insights
Kimi, a product of Dark Side of the Moon, announced the suspension of new user subscriptions on the C-side. This decision not only confirmed the explosion in demand for its new models, but also responded to Goldman Sachs’s previous warning in an ironic way - instead of subsided, the demand for computing power has been rekindled, catalyzed by low-cost, high-performance models.
Kimi disclosed in its announcement on July 19 that since the release of Kimi K3, the number of user requests within 48 hours has greatly exceeded expectations and is approaching the capacity limit of the existing cluster. The company will suspend the subscription of new users on the C side from now on, and will make every effort to promote the expansion of computing power. Once the new computing power is in place, the subscription quota will be gradually opened.

The direct background of this scene is the wave of demand triggered by the strong performance of Kimi K3 in multiple public benchmark tests. Just before this, Goldman Sachs partner and head of equity business Rich Privorotsky had just characterized the release of Kimi K3 as a deep signal for the industry, warning that the "era of computing power expansion" may be coming to an end, and the logic of AI valuation is facing reshaping. Kimi was forced to cut off its subscription due to insufficient computing power, which is a direct refutation of this assertion.
The market impact of Kimi K3 is supported by specific data. This model reached the top of the front-end code arena in the Arena comprehensive text ranking with 1679 points, surpassing Claude Fable 5 and achieving a leapfrog improvement from 18th to 1st place compared to the previous generation product. In the front-end field, this model ranks first in 6 out of 7 subdivisions, including branding and marketing, reference design, data and analysis.
Dark Side of the Moon also claims that Kimi K3 has surpassed OpenAI and Anthropic flagship products in multiple public programming and agent benchmark tests, and its cost is about 40% lower than similar top US models. The combination of top performance and low cost directly led to an influx of users that far exceeded expectations, eventually triggering the company's active circuit breaker on subscription channels.
In the announcement, Kimi "sincerely apologizes" to users who failed to get the experience they expected, and promised that all rights and interests of subscribed users will not be affected. The company also announced that it will split the Kimi main rights and interests (including Kimi Web, Kimi APP, Kimi Work) and Kimi Code rights when the subscription is subsequently resumed in order to more accurately match computing resources.
Kimi’s embarrassment of running out of computing power is in sharp tension with Goldman Sachs’ previous predictions. Rich Privorotsky once pointed out that a Chinese laboratory that could not match the largest pre-training computing power in the West quickly narrowed the gap with the top American models through architectural innovation, synthetic data, reinforcement learning and post-training technology. He warned accordingly that "expansion" is no longer the only path to victory, and investment in large-scale computing infrastructure is expected to face repricing.
However, what Kimi’s incident revealed is exactly the other side: even if the computing power arms race is bypassed through algorithm efficiency, once the model performance is truly recognized by the market, computing power demand will still rebound in a more violent way. The low-cost model lowers the threshold for use, but may create a larger computing power gap on the demand side.
AI analyst Kim Isenberg commented on Goldman Sachs' judgment: "The entire rules of the game have changed, which will trigger a 'red alert' for some institutions." Kimi's subscription meltdown is the most intuitive way to show this red alert.
Dark Side of the Moon plans to release Kimi K3 in an open weight form on July 27, allowing enterprises and governments to customize deployment on their own systems. This path is highly similar to DeepSeek's previous open source strategy, which has once again triggered market concerns about a repeat of the "DeepSeek moment."
The combination of low cost, high performance, and open source poses a direct challenge to the existing pricing power of AI computing power. In the previous DeepSeek impact, the market has experienced a round of doubts about the necessity of large-scale computing power investment, and the valuation of related AI infrastructure stocks has been under pressure.
For investors, Kimi’s incident provides a more complex analysis framework: the expansion of computing power itself may no longer be a sufficient condition for AI competition, but the explosion of demand brought about by breakthroughs in architectural efficiency may still strain the supply of computing power.