Kimi K3 перегрузил кластеры за 48 часов — Galaxy Securities ждёт бума спроса на суперноды

After Moonshot AI officially open-sourced the Kimi K3 model, requests to the system during the first 48 hours completely exhausted the capacity of existing clusters — the company was forced to suspend registration of new paid subscribers in the consumer segment.

Kimi K3: What Happened in the First 48 Hours

The cluster overload within the very first day confirmed two key theses: the high performance of Kimi K3 and the acute shortage of computing resources in the industry. At the moment of release, the model received Day 0 compatibility with several major Chinese cloud platforms — that is, literally from the first day the code was published. This is rare for the market: usually adapting a new open checkpoint to specific hardware takes days or weeks.

Key facts from the Galaxy Securities analytical report:

• 48 hours — the time from code publication to the overload of all Moonshot AI clusters
• Moonshot suspended registration of new subscriptions for consumer (C-end) users
• Day 0 compatibility: Kimi K3 was adapted to several national computing platforms from day one
• Model performance closely approaches the world's leading closed systems

Why

Does Open-Source Kimi K3 Accelerate Rather Than Reduce GPU Demand?

According to Galaxy Securities analysts, top-tier open models do not take market share away from infrastructure manufacturers — they expand it. Kimi K3 puts pressure on the leaders of the closed systems market from two directions simultaneously. First, when an open model reaches the level of the best closed counterparts, leading labs are forced to scale up training and accelerate iterations to maintain differentiation — this requires more compute for training.

Second, the wide availability of an open model generates enormous demand for inference capacity: the broader the audience, the more requests per second. According to Galaxy Securities data, Kimi K3 in its first days already activated demand for Chinese supernodes — high-performance computing nodes based on national GPU platforms.

Which Sectors Will Receive Orders?

Galaxy Securities explicitly names the segments that will benefit from growth: servers, network switches, optical modules, liquid cooling systems, and power supplies. Demand for supernodes pulls the entire national AI infrastructure manufacturing stack along with it — from silicon chips to rack equipment.

"Kimi K3 does not restrain demand for computing resources — it continuously explodes it," — this is the key conclusion of the analytical report by

Galaxy Securities bank, published by Jiemian.

Analysts recommend paying attention to Chinese supernode manufacturers and companies across the entire AI cluster supply chain. The shortage of computing capacity exposed by the Kimi K3 launch is not a one-time episode, but a persistent structural deficit.

What This Means

The open-sourcing of Kimi K3 is a catalyst, not a brake, for the AI hardware market. The closer open models get to closed leaders, the higher the stakes for all participants in the race: more compute for training, more servers for inference, more orders for national stack manufacturers. For the Chinese AI industry, this is a concrete market signal.

Frequently Asked Questions

Why

Did Moonshot Suspend Subscriptions After Open-Sourcing Kimi K3?

Moonshot AI suspended registration of new paid subscribers because within 48 hours of open-sourcing Kimi K3, requests completely exhausted the capacity of all the company's existing clusters. The infrastructure was insufficient to handle the explosive traffic growth.

What Are Chinese Supernodes and Why Are They Needed?

Supernodes are high-performance computing nodes that combine many national GPU accelerators for training and inference of large language models. Kimi K3 received Day 0 compatibility with several such platforms, which sharply increases demand for the entire corresponding manufacturing stack.