Chinese AI startup Moonshot AI has temporarily halted new consumer subscriptions for its flagship Kimi K3 model after demand overwhelmed its computing infrastructure. The move underscores the intense pressure on AI developers to scale hardware resources amid surging user interest.
Compute Constraints Force Subscription Pause
Moonshot announced on Sunday that Kimi K3, launched just days earlier, attracted far more users than anticipated, creating what the company described as “unprecedented compute challenges.” User requests in the 48 hours following launch spiked sharply, nearing the limits of existing GPU clusters. As a result, new subscriptions are suspended immediately, while existing paid subscribers retain full access to available capacity.
The company plans to restructure future memberships into two tiers, including a dedicated coding plan, to better align compute resources with user needs. Moonshot stated that new subscription slots will reopen gradually as additional capacity comes online.
Funding and IPO Preparations Advance
Moonshot is simultaneously pursuing fresh capital and laying groundwork for a potential Hong Kong initial public offering. According to a fundraising document seen by Reuters, the startup raised over $2 billion in May from investors including Meituan, China Mobile, and CPE, bringing total historical funding to more than $5.5 billion. The company is now seeking up to $2 billion in additional capital and reached a $30 billion valuation in June.
Two sources familiar with the matter told Reuters that Moonshot is unwinding its offshore corporate structure as part of IPO preparations. The company has engaged Goldman Sachs and China International Capital Corp (CICC) as financial advisers, though the IPO timeline remains subject to change.
Kimi K3: A Massive Model with Heavy Compute Demands
Kimi K3, described by Moonshot as a 2.8 trillion-parameter system and the world’s largest open-weight AI model, is designed for coding and agent-based tasks. These workloads require repeated model calls and significant inference capacity, making them costly to operate at scale. While open-weight models allow users to download and customize the system, analysts note that few will host a model of this size due to hardware expenses.
Moonshot said Kimi K3 has delivered competitive performance against leading US models in certain technical benchmarks, with independent evaluations confirming strong results. The company added on X: “Kimi K3 has received far more love than we expected, and our GPUs are feeling it.”
Broader AI Infrastructure Race Intensifies
The capacity crunch at Moonshot reflects a wider trend across China’s AI sector, where startups are racing to secure computing power amid US export controls on advanced Nvidia chips. Competitors like DeepSeek have also sought external funding to expand compute capacity. Meanwhile, Alibaba, an investor in Moonshot, launched its own 2.4 trillion-parameter Qwen3.8-Max-Preview model on Sunday, further heating the competitive landscape.
For more on the AI monetization challenge, see Big Tech Earnings: Wall Street Demands Proof of AI Monetization on $725B Capex. To understand the broader implications of Kimi K3’s launch, read Moonshot's Kimi K3: 2.8 Trillion Parameters Challenge US AI Dominance. For context on compute leasing deals, see Meta in Talks to Lease AI Compute to Anthropic in Potential $10B Deal.
This article is for informational purposes only and does not constitute financial advice.
