Kimi K3 Demand Surge Forces Moonshot AI to Halt New Sign-Ups

Chinese AI startup Moonshot AI has temporarily suspended new user registrations for its Kimi K3 AI assistant, just two days after launch, after overwhelming demand pushed its GPU computing infrastructure to capacity. The move, announced on July 19, 2026, underscores the explosive growth in China’s consumer AI market and has sent shockwaves through global technology stocks.

The GPU Crunch

In a statement on X (formerly Twitter), the company acknowledged the strain: “Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity.” As reported by Caixin Global, the company is prioritizing compute resources for existing subscribers while it works to scale its infrastructure.

Existing subscribers remain unaffected, and Moonshot plans to reopen new subscription slots in controlled batches rather than all at once. The company has also split its membership into two tiers — a general “Kimi Membership” for web, app, and office use, and a “Kimi Code Membership” for programming workflows — to allocate computing resources more efficiently. This distinction matters: coding and agentic tasks involve long chains of reasoning and large context windows, making each user far more compute-intensive than a casual chat session.

A Model That Demands Attention

Kimi K3, unveiled at the World AI Conference (WAIC) in Shanghai on July 16-17, is a 2.8 trillion-parameter mixture-of-experts model — the largest open-weight model globally at launch, surpassing DeepSeek V4’s 1.6 trillion parameters. According to The Next Web, it scored 57 on Artificial Analysis’s Intelligence Index, trailing only Anthropic’s Claude Fable 5 (60) and OpenAI’s GPT-5.6 Sol (59), with particular strength in front-end coding and long-context tasks.

The model features a one-million-token context window with native vision capabilities, and its API pricing is set at $3 per million input tokens and $15 per million output tokens — competitive with Anthropic’s Claude but higher than some Chinese rivals, as Dataconomy detailed.

Market Tremors and the DeepSeek Echo

The K3 launch triggered immediate market reactions. AI and semiconductor stocks dropped sharply on July 17 as investors drew comparisons to the “DeepSeek moment” of early 2025, when a Chinese lab released a model matching US rivals at dramatically lower cost. CryptoBriefing noted that analysts characterized the selloff not as a shock but as an expected continuation of a trend.

Chinese AI stocks Zhipu AI and MiniMax fell sharply on July 20, with K3’s performance and pricing driving valuation resets. The BBC reported that the selloff was so severe that China mounted one of its broadest efforts in years to steady the stock market on July 21, with regulators and state-backed investors moving to shore up confidence.

The Open-Weight Paradox

Kimi K3 is billed as an open-weight model, but the full weights won’t be released until July 27, 2026. Until then, all demand lands on Moonshot’s own servers — a tension TechNode described as the “open-weight paradox.” Once the weights are public, large customers and cloud providers can host K3 themselves, potentially easing the crunch for casual users who will still rely on Moonshot’s apps.

A Broader Industry Squeeze

Moonshot is not alone in its capacity struggles. In the same week, Anthropic cut usage limits on its Claude Fable 5 model for subscribers, citing demand that was “hard to manage.” Alibaba also rushed out an open-weight Qwen model at a steep discount, aimed squarely at competing with Kimi K3. This pattern points to a broader industry challenge: a wave of capable AI models is arriving at prices that undercut US labs, but the barrier to using them is increasingly whether anyone has enough chips to keep them running.

Analysis: A Turning Point for AI Economics?

The Kimi K3 episode carries implications far beyond one company’s capacity constraints. Global investors view K3’s aggressive pricing as a negative signal for model-layer margins, potentially marking an industry turning point where Chinese competition compresses profits for US AI companies. Atreides Management founder Gavin Baker commented that K3 could mark a turning point for the AI industry, as noted by Caixin Global.

The model’s performance also challenges the assumption that AI leadership requires massive, sustained infrastructure spending — the very thesis that has driven Nvidia’s meteoric rise. If Chinese labs can produce competitive models at lower cost, the investment case for US data center buildouts may weaken.

What’s Next

Moonshot AI is racing to add capacity ahead of the July 27 open-weight release. Once the weights are public, large customers and cloud providers can host K3 themselves, which should ease the strain on Moonshot’s servers. Casual users, however, will still rely on the company’s apps, so the crunch may ease rather than disappear entirely.

The company’s financial trajectory adds urgency. Annualized recurring revenue reached $300 million in June 2026, up from $200 million in April, and Moonshot is reportedly preparing for a Hong Kong IPO that could value it at over $30 billion. The demand surge is a powerful signal of market appetite, but the GPU capacity constraints raise real questions about scalability. For investors and industry watchers, the key question is whether Moonshot can build infrastructure fast enough to match its soaring ambition — and whether the global GPU shortage will let it.