Kimi K3 Faces Computing Power Crunch 48 Hours After Launch; Moonshot AI Suspends New Consumer Subscriptions

After Moonshot AI released the Kimi K3 model with 2.8 trillion parameters on July 16, user request volume far exceeded initial estimates within 48 hours, approaching the capacity limit of existing computing clusters. On July 19, the company announced the suspension of new consumer subscriptions.

According to the official announcement, this move aims to prioritize all available computing power for existing subscribers, avoiding task interruptions and response delays that could affect core experience. The company has also initiated expansion of its computing clusters, pledging to gradually reopen subscription slots once new resources are in place until normal access is restored.

Kimi K3 supports a context length of 1 million tokens and natively supports visual understanding. It is designed for long-context programming and end-to-end knowledge work scenarios, and is currently the world's largest open-source model by parameters. Since its release, it has rapidly gained market support far exceeding expectations, but also directly exposed the bottleneck in high-end computing supply.

The computing shortage stems from the dual amplification of model size and usage intensity. The 2.8 trillion parameters combined with a million-level context window make single inference consumption significantly higher than previous versions, while users' high-frequency calls in programming and knowledge work scenarios further increase cluster load. The explosive growth over the past 48 hours exceeded the team's earlier estimates, directly triggering the resource reallocation decision.

To improve resource allocation efficiency, Moonshot AI plans to split membership benefits when resuming subscriptions. One category is the main benefits including Kimi Web, Kimi APP, and Kimi Work, serving ordinary users' daily office needs. The other category is the Kimi Code programming benefits, sold separately to developers, to prevent high-consumption features from crowding out basic computing power.

This adjustment reflects the practical constraints of open-source large models on the path to commercialization. While free open weights can quickly attract global users, the computing costs under high-intensity usage scenarios are difficult to cover with a single subscription model. Splitting benefits is equivalent to tiered pricing by scenario, providing room for subsequent refined operations.

For developers, suspending new subscriptions temporarily limits new user access, but existing subscribers can still use Kimi Code benefits for programming tasks. Enterprise users need to assess whether their current quotas are sufficient for team size; if additional seats are needed, they may face waiting lists.

For Moonshot AI itself, the event both validates the market competitiveness of Kimi K3 and accelerates its preparation for a Hong Kong IPO. The company has sent shareholder resolution documents to investors, aiming to list in Hong Kong within six months at the earliest, while advancing a new funding round with a valuation target exceeding $30 billion.

Horizontally, the gap in parameter scale between Kimi K3 and its U.S. competitors has narrowed, but Chinese models generally adopt open-source strategies, contrasting with Silicon Valley's preference for closed systems. Open-source helps global user growth but increases the difficulty of converting free usage into sustainable profitability.

Upstream in the industry, computing power suppliers face demand fluctuations. The early phase of model release drives expectations for computing procurement, while the subsequent subscription suspension may slow short-term expansion pace, but long-term expansion plans will still maintain hardware investment.

For the downstream developer community, open-source weights lower the barrier to entry. Users can deploy locally or test long-context and visual capabilities via API calls. When selecting models, enterprises need to compare Kimi K3's million-token window with the actual throughput performance of similar models, and decide whether to prioritize applying for existing quotas based on their own programming or knowledge workloads.

The most likely development of this incident is that, as expansion resources gradually come online, Moonshot AI will restore subscriptions in batches and test the pricing effect of the split benefits. Key points to watch include the progress of new computing deployment, developer retention after the separate sale of Kimi Code benefits, and financial details disclosed in the Hong Kong listing documents.

The above assessment is based on publicly disclosed computing challenges, membership split plans, and financing rumors, and represents an analysis of public information rather than established facts.