Moonshot’s Kimi uses 20k Nvidia chip cluster from Alibaba
bloomberg.com
bloomberg.com
I guess at least partly a reflection of all the optimizations in the Kimi 3 architecture.
In the recent leaked DeepSeek investor meeting, they also mentioned only having a 20K GPU cluster (unclear if NVIDIA, or Huawei).
Source?
And yes, I understand the stolen data was expensive to make, so I understand the owners of it are also frustrated, but that's partly a problem with current law. Would the authors of the world be rich if OpenAI bought a single copy of their book to legally scan? For best sellers, that's somewhere around pennies, so no. Should the authors get a share in OpenAI? Current laws says, unambiguously, "no".
Frustration all around is reasonable.
The way Anthropic are using "distillation" is just in a very broad vague sense to claim that some data generated by their model was used to help train another one. They are not talking about something like internal logits, expensive to derive, that would be useful to train a smaller model, but rather about any output from their model, even outputs with redacted reasoning (i.e. incomplete outputs that do NOT reflect the underlying knowledge of the source model).
Given the way Anthropic are using the word, IMO it's better just to think of this as cheap training data, and indeed very similar to the way Anthropic themselves got cheap training data just by taking it (even in cases where that was illegal - copyright). The alternative for Moonshot would be to pay for human generated reasoning data, just as the alternative for Anthropic would have been to pay human developers for coding data etc, not just take it from wherever they could lay their hands on it.
So, I guess Moonshot may have violated Anthropic's TOS, in using Anthropic output to compete against Anthropic, but unlike Anthropic they at least didn't break copyright law since model output is not copyright, and technically they may not have even violated Anthropic's Terms of Service unless they owned the accounts used to access Anthropic's models (perhaps not - they may have used one of the anonymizing Chinese token resellers).
So yeah - pot calls kettle black.
> but unlike Anthropic they at least didn't break copyright law
Reference? All the recent lawsuits have sides with Anthropic, that I've seen. What exact case do you have in mind?
I am, clearly, not talking about the act of not paying to obtain the works. That is wrong, and they've been sued successfully, because the law is clear about that. I'm talking about the integration of works into the model, without permission, which everything I've seen says none is required in a "sufficiently transformative" framing.
Moreover, it’s very plausible (and expected) to use multiple clusters and GPU types for RL rollouts which could very well not be included in this count.
No part of this pipeline is fixed in stone.
[1] https://www.chinatalk.media/p/how-to-buy-cheap-claude-tokens... [2] https://xcancel.com/synthwavedd/status/2078514339552628880
I think the word "distillation" needs to be used a bit more selectively here. If their pre-training run was complete before Fable was released that implies that ZERO Fable data went into the base model. Perhaps the timeline allows for a few weeks at best of incremental post-training on some limited amount of Fable data, but calling this "distillation" seems a bit dramatic especially given the redacted outputs that would have been available. A more factual speculation would just be that they may have had time to post-train using a limited amount of Fable output in some fashion (LLM as judge? SFT? Who knows ...).
In this sense, any advance in intelligence is a performance improvement and vice versa
When OAI released gpt-oss it was released as an mxfp4 checkpoint.
OAI, Ant, et al are also obviously employing QAT.
Are you sure they are using all of their compute on training? Didn't they rent out a ton to other AI companies?
the only way is down for the massive valuations and 'a.i' revenue projections.
Its all a fucking capitalistic farce to display to other rich elite that "Look at how much clout I have! I can make these peons dance around and do my bidding! Im a slave-owner!"
https://infosec.exchange/@david_chisnall/116991627711001827
He noted that that Silicon Valley doesnt really want to SOLVE problems. They want to find already-solved problems with problem matching. And of course, we just throw more people and more compute instead of optimization and understanding.
The Chinese are being actively constrained with bullshit politics around a second Red Scare moment. And, well, they're winning. A lot.
I just bought 2 switches, 48 port 10GbE with 4x QSFP+ at 40Gb fiber. $110 each.
You can even get 24 port QSFP+ @100Gb networking devices for $350.
Yeah while ram and gfx is $$$$$, networking is rock bottom prices..
(new in terms of the most recent models, not just "something with X number of ports")
Can’t possibly be intentional media strategy by a geopolitical target, right? If its going to negatively impact valuations and revenue of major rivals, seems like a desirable strategy?
We just saw OpenAI take the steps to significantly lower the cost of one of their models, which confirms that at least one western lab has a large margin on inference, not a large inference cost.
Meanwhile, we don’t know how many experiments these east/west labs are performing relative to each other. We also know that many western labs have a whole portfolio of models too, which is product breadth not necessarily waste.
> A person familiar with Moonshot’s procurement strategy confirmed that the company does indeed have a channel for accessing Blackwell processors via Southeast Asia. They didn’t specify whether this was a rental channel — which is legal, in most cases — or direct purchases, which are a breach of US regulations. The Information reported this week that Moonshot is seeking additional Blackwell processors to train its next model.
> The 20,000 chips Moonshot accesses via Alibaba, meanwhile, are from Nvidia’s earlier generation of Hopper products, the people familiar with the agreement said.
So the 20k GPUs from Alibaba is only a lower bound on how many you need to train a model like Kimi K3.
The advantage of having more GPUs in any case is not so much that you can train bigger models, but that the turnaround time is faster, so you can run more experiments to dial in training choices. It's entirely possible that Musk has more than enough compute, but can't hire the talent to run all those experiments. (That would also explain why he has excess capacity he can rent to Google.)
A few quotes from the transcript:
> Our current computing capacity is approximately 20,000 H-equivalent units, most of which have just arrived within the past month or two
> Regarding the Huawei 950, Huawei currently provides us with 16,000 SIM cards
> A Huawei 950 [cluster] with 16,000 cards is equivalent to only a B-series card [cluster] with 4,000 cards.
I’m not American nor Chinese, and I’m from a much more US-aligned country. But, Christ, you lot really are asking for it.
The truth is probably Anthropic/OpenAI/Google are pretty efficient but less efficient than the Chinese labs, the Chinese labs probably have more compute than they say to undermine US spending and distillation is quite efficient at bridging the gap in compute.
High quality data is expensive. Synthetic data will get you so far, but after that you need to start paying experts to create data for you, which has been going on for a long time. The latest thing is paying for human written LLM-as-judge AI-output evaluation "rubrics", trying to extend RLVR into areas where "looks like it checks the boxes" is the best you can do.
When anyone, Chinese or not (Elon Musk cheerfully admits to distilling OpenAI models) uses the output of someone else's model to train their own, then what they are primarily getting is cheap training data, but you still need to train your model on this data! You may have reduced the cost/speed of training data acquisition, but if you are training a 3T param model (Kimi 3) then you still need the compute to do that - that did not change.
There was an interesting mention of the cost of training data in the recently leaked DeepSeek investor meeting, where their CEO referred to the cost of human-generated training data in China (i.e. using Chinese labor) as being the same as that in the US, which seems surprising. He also mentioned the time such data takes to be created. No doubt the Chinese will catch up in this area - this is just time and money, not Dutch technology (ASML) that the US is blocking them from buying.
And btw, yes it's an objective fact from every technical angle that the Chinese only have competitive models because of US tech. Why do you think they try so hard to smuggle NVIDIA GPUs, and now exposed infrastructure to create cheap imitations of American tech?
The Trump administration, having first blocked China from buying NVIDIA H100's, has since done a U-turn and is now allowing them to buy the more powerful NVIDIA H200, on a case-by-case basis.
Now, the CHINESE government is blocking Chinese companies from buying these H200s, at least in part because it turns out that being denied US tech has been a great accelerator for Chinese tech, with Huawei now producing the entirely domestic Ascend 950 chip, which according to NVIDIA's Jensen Huang performs about the same as NVIDIA's own H100 (which while not NVIDIA's most powerful is still plenty capable, and is what Elon Musk's Colossus-1 data center mostly uses, currently being rented out to Anthropic).
The CCP tried to promote their own chips, but later regressed and started allowing ByteDance, Alibaba, and Tencent to purchase more than 400,000 H200s. And they are trying to acquire millions.
Thus, export restrictions accelerated Chinese substitution, but also made that substitution slower, costlier, less scalable, and technically inferior to unrestricted Nvidia access.
> According to Jensen Huang, Ascend 950 performs about the same as H100
Source?
(We've changed the title to what the article says now.)
This feels like a very American way of designing things - just throw more horse power at it, bigger is better! The rest of the world is usually a bit more resource constrained and efficient at using those resources.
See also Mustangs vs German sports cars, giant American fridges, giant American suburban McMansions vs livable cities etc etc.
I don't think so. Bruteforcing problems is a well established strategy in any field that involves computation of any form. Once you get something working, you can get results right now if you throw resources at it. In the meantime, any improvement in efficiency can easily be back ported to the same computational resources you're using.
The goal will be to develop smaller models with more efficient architectures, that have similar or even better performance than larger models.
A VAX 11/780 was good, but an 80386 was a lot better, since the latter could run on 3 AA batteries and the former needed 6,000 watts of 3 phase.
Continual learning tends to imply individualized models, else there is no data privacy (the secrets learned on the job at your company now being available to your competition), which really turns the current AI business model of a single centralized model served to everyone on it's head. If every customer has a different model that essentially means the end of batch processing with the same weights loaded into the GPU.
The direction this suggests is a move away from centrally served common models to locally served individual ones, which generally requires them to be smaller, even if some larger companies may be willing to invest in beefier hardware.
I think this is at least in part why the AI companies are trying NOT to implement true continual learning and see if they can instead finesse it by implementing continual compacted(?) memorization instead, since then it's "just" additional context that needs to be recalled and fed into every request, not weights that need feeding into the GPU. I don't think memorization is any substitute for learning, especially learning of practiced skills, but since it's far easier to implement, and non-disruptive to the cloud-based API business model, this is what we will see first.
The recent news of NVIDIA' investment in Sutskever's SSI has a tiny hint of this also, talking about SSI advising on NVIDIA's future architectural direction - apparently pushing it in a different direction than current (cloud-based, pre-trained) models. NVIDIA may be quite happy to see a move towards local models.
The frontier labs would be well served in carving out 20k sub-clusters and giving research teams carte blanche in building things with radically different architectures - with full permission to distill whatever they want from the flagship models. We'd expect to see more product lines that feel "different" from the flagship models if this were already being done.
I am adding multi-modality to https://github.com/guilt/TinyToT, and I see that dis-aggregating capabilities, very similar to how our own sensory organs work, seems to be paying off quite well.
If they got a lot cheaper and only a little dumber, and we got a little smarter about how we use them, they could appear to the bystander as much more useful than they are.
isn't that, in the end, the case with all/most technologies?
i can get a petaflop of compute capacity in a dgx spark for relatively cheap nowadays. that used to be a whole supercomputer like 20 years ago.
Someone will find a way to make cheaper compute, and since nothing fundamental changed there (LLM didn't change how silicon were made), that's bound to happen.
Either they have Blackwell with native 4-bit floating math, or they use have Chinese domestic NPU that support mxfp4 natively.
The article’s statement does not make sense.
Probably either the author or the people they interviewed got the specific GPU details mixed up or wrong. Maybe it was B200s.
(Kimi K3 tech report section 4.1.1 https://arxiv.org/pdf/2607.24653)
Personally, I think 20K nvidias is a stop gap solution because they really don't have the capacity to serve their models to earn any money right now.
From the article:
A person familiar with Moonshot’s procurement strategy confirmed that the company does indeed have a channel for accessing Blackwell processors via Southeast Asia. They didn’t specify whether this was a rental channel...
Both W4A8 and W4A16 schemes are supported by Hopper GPUs and commonly used to serve mxfp4 Kimi models on Hopper.
I wonder how much spending is motivated by the phenomenon of sudden emergent performance in LLMs. Clearly some people who are smarter than me expect something like emergent AGI, or at least they think the odds justify spending whatever it takes to see if that would happen.
That leaves a lot of room for efficient aggressive followers.
If Alibaba can't bring the chips into China, but can buy a whole bunch in Thailand or Singapore or whereever (or secure exclusive rights via JV partners where relevant) and then just provide them as a service to its customers in China - what is the point? I'm sure many customers would actually prefer such an arrangement.
[1] https://www.tomshardware.com/tech-industry/artificial-intell...