Moonshot AI pauses new subscriptions, saying that over the past two days Kimi K3 demand nearly passed the limit of its capacity, and also reworks its plan tiers

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. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and

@kimi_moonshot 2026-07-20

Context & Ripple Effects

Moonshot’s capacity constraint arrives days after its Kimi K3 launch, which the company positioned as a 2.8T-parameter model competing at the frontier. The abrupt move from launch to admission control makes serving capacity—not just model claims—the immediate test of the release.

Moonshot had also been emphasizing efficiency through lower reasoning-token use in K2.7-Code. K3’s demand surge shows that efficiency improvements do not eliminate the operational pressure created when a new model attracts heavy interactive use.

First-order effects

Prospective Kimi K3 customers cannot newly subscribe for now, while existing subscribers receive protected access as Moonshot holds demand within available GPU capacity.
Reworked plan tiers become Moonshot’s near-term tool for rationing access and aligning subscriber usage with a constrained serving pool.

Second-order effects

The pause gives rivals an opening to target users unable to obtain Kimi access, while Moonshot must balance growth against service quality for its existing base.
Capacity management becomes part of K3’s commercial positioning: tier design can steer high-intensity demand toward plans that better reflect the cost of serving it.

Third-order effects

If similar launch surges persist, frontier-model competition will be shaped increasingly by inference capacity and allocation policies rather than benchmark performance alone.
Subscription tiers and temporary admission controls could become more common mechanisms for AI providers to manage scarce compute, especially around major model releases.

The trend: Frontier AI providers are moving toward capacity-aware product design as demand for newly released models tests the limits of available inference infrastructure.

Discussion

@kimmonismus @kimmonismus on x
Demand for Kimi k3 is so high that no new subscriptions are currently paused. k3 is a resounding success; you couldn't ask for better publicity.
@daveshapi David Shapiro on x
When the signs all say SOLD OUT there is no bubble. Sorry. We are dramatically, sorely, epochally underbuilt. The correct amount of compute for humanity is probably at least one trillion times what we have today, and that's just entry level for the next phase of
@ehrmantrautcap_ @ehrmantrautcap_ on x
Bullish for stocks related to compute.
@jukan05 Jukan on x
Who said Kimi K3 is bearish for compute? Lol.
@ckcapitalxx @ckcapitalxx on x
The hottest model in the world just started turning away paying customers because there isn't enough compute to serve them. This is the neocloud thesis in one post.
@bigeagle_xd @bigeagle_xd on x
we're working extremely hard with our partners to open the weights as soon as possible
@0xdevshah Dev Shah on x
2 types of companies when they run out of compute to serve new customers - > Moonshot - stops new revenue. no lag for existing customers. > Anthropic - cuts 50% usage of existing customers to serve new ones that tells you everything you need to know why anthropic hates open
@davidondrej1 David Ondrej on x
this is called Product Market Fit
@andrewcurran_ Andrew Curran on x
Kimi is pausing new subscriptions due to extremely high demand, and will be splitting their plans into Kimi Membership and Kimi Code Membership.
@highyieldharry @highyieldharry on x
R.I.P. the bear case July 16-July 19
@nicbstme Nicolas Bustamante on x
Guys, the world is so compute constrained. I honestly don't understand how anyone can think otherwise. Almost nobody is seriously using AI today, and we're already hitting capacity. Everything is a bottleneck: land for data centers, permits, electricity, grid connections, chips,
@anni_sen Anni on x
Kimi K3 is short compute. Demand is so high , they can't take new subs. Wasn't Kimi K3 the Deepseek moment that was going to kill the AI infrastructure trade? Long live compute, interconnects, memory and DC energy. July 22nd $GOOGL will reiterate capex and if I were to wager
@boringbiz_ @boringbiz_ on x
Jensen Huang right now [image]
@khushiirl Khushi on x
imagine shipping a model so good you have to pause new subscriptions [video]
@patrickmoorhead Patrick Moorhead on x
Wait, you mean that alternative models with a few superior metrics need more infrastructure to support the inference demand? No way!!!! 😉
@kevinsxu Kevin S. Xu on x
Alibaba is a major investor of Moonshot. Kimi, in turn, uses a big chunk of AliCloud GPUs (oftentimes competing away Qwen team's resources) 10 days ago, Alibaba previewed quarterly earnings saying AliCloud is growing 40+%. This was ofc before K3 Could be a monster quarter for
@hypertechinvest Daniel Romero on x
> Moonshot doesn't have enough compute > OpenAI doesn't have enough compute > Anthropic doesn't have enough compute The market: Let's liquidate semis, neoclouds, and data center stocks There are so many attractive valuations right now This drawdown is retarded [image]
@natolambert Nathan Lambert on x
Huawei about the get a blank check from the government to make inference chips
@healthranger @healthranger on x
So much demand for Kimi-K3 that the company is forced to halt new subscriptions for the moment. That's because everybody is ditching Anthropic and OpenAI and switching to Kimi-K3. It's just as good in nearly every way, yet a fraction of the cost. I wonder how long it will take
@anshuc Anshu on x
extremely unprofessional. if kimi wants to make it as a frontier lab, they need to act like one: perhaps silently route people to worse models, and maybe write a blog post about the collapse of humanity
@ultralinx @ultralinx on x
Wtf this is actually crazy. Until I see RAM and SSD prices come down, I'm continuing to invest in AI related stocks. Your tech devices aren't going to get cheaper any time soon.
@alexpilar Alex on x
open source rules
@jason @jason on x
.@Kimi_Moonshot could have very well been an American company if it wasn't for our idiotic immigration policy The 46th and 47th administrations has both failed to solve the insanely easy immigration problem/opportunity Wake up @potus @vp we're in a war for the most elite

Chronicles