Artificial Intelligence thread

[tokenanalyst]

However, is it sustainable for Chinese companies investing heavily in this only to have to give out their IPs and not make much revenue ? Afterall, they are not SOE who don’t need to make profits and are more for public goods, they are private enterprises who need to satisfy their shareholders at some point . Is there away they can monetize their products /IP? Companies like Alibaba or tencent/Bytedance wont have much of this issue since it’s not their main business and they can use their AI models to increase their monétisation of their main business which they are already benefiting hugely from, but Chinese AI startups who are at the leading edge will have a hard time sustaining this for years to come no?

As long they stay grounded in reality and not become a trillion dollar fantasies, I think they will be OK. the most important part is building the ecosystem which I think China is trying to develop. An End to End AI tech ecosystem that can sustain itself. LLMs are just part of it. With a ecosystem any company can move from rent seeking to services.

All AI startups are doomed anyway. AI is a commodity that is given away for free. You cannot make money selling AI models.

The only way to make money is to become AI hosts, essentially data center companies.

I think most AI startups will either become data center companies themselves hosting AI and also developing AI, or they will be bought by the big tech data center companies like AliBaba.

Maybe they start gating their most advanced models and open sourcing their older models, kinda like patreon vloggers. Who put their latest stuff behind a paywall for a brief period and then release it for free when the next video comes along.

OpenAI and Anthropic are not even datacenters companies, they are trying to become rent seeking businesses, in others words they don't even care about the US AI ecosystem, they want their proprietary stuff and your business hooked on it. I don't have to tell you how dangerous is to your business depend on overvalue company instead of a ecosystem of options.



The problem for them and for AI bubble in general that is becoming harder and harder to rent. Both Open and Antropic are not even making a lot of profits. So even without DeepSeek and Kimi it will reach a point that is cheaper to hire Indians than to pay anthropic tokens.
That wouldn't be a problem if OpenAI and Anthropic don't have the insane valuations they have. The expectation of investor is just to high.

 

[tokenanalyst]

This rent seeking business model will price out like 90% of the users if was not massively subsidize. Only changing the model from fixed subscriptions to tokens and keeping the users is the only way to maximize profits. And they will do it if they know that you can't move to any other option like DeepSeek Flash ( that model is really good, I hope that they keep releasing it)

 

[siegecrossbow]

Please, Log in or Register to view URLs content!

Now that’s something I can get behind.

 

[discspinner]

the elephant in the room is when/or if open source models surpass those of Anthropic/OpenAI on all metrics.

considering Kimi K3 is #1 in agentic capabilities proves that it is possible for an open source model to become the leading frontier model.

this may happen in the near future given China's power advantage, delays in data center construction in the US, and innovations in model training and inference efficiency;

this will lead to an expected thinning out of the field; AI models that are falling behind (gemini, grok, etc) are likely to adopt open source models entirely to modify in house; this means an overcapacity of hardware (eg Musk / Meta leasing out excess capacity)

meanwhile, open source breaks the duopoly of Anthropic/OpenAI and they cannot collude to keep prices high; price wars will inevitably ensue, decreasing revenue, profit margins, and valuation of the entire AI ecosystem;

 

[tokenanalyst]

this may happen in the near future given China's power advantage, delays in data center construction in the US, and innovations in model training and inference efficiency;

I don't know about the problems of power in the US. But the irony is that probably will increase demand for more processing power as more companies and providers look for hosting for these models. Is not only Anthropic and OpenAI anymore.

And will probably push OpenAI and Anthropic to release more advance model to keep themselves ahead. If wasn't for Chinese models they wouldn't have incentives to train and release models so frequently or to train larger parameter models. Anthropic could milk Opus for a year or two in a expensive per Token plan instead of releasing Fable in subscription model.

 

[tamsen_ikard]

the elephant in the room is when/or if open source models surpass those of Anthropic/OpenAI on all metrics.

considering Kimi K3 is #1 in agentic capabilities proves that it is possible for an open source model to become the leading frontier model.

this may happen in the near future given China's power advantage, delays in data center construction in the US, and innovations in model training and inference efficiency;

this will lead to an expected thinning out of the field; AI models that are falling behind (gemini, grok, etc) are likely to adopt open source models entirely to modify in house; this means an overcapacity of hardware (eg Musk / Meta leasing out excess capacity)

meanwhile, open source breaks the duopoly of Anthropic/OpenAI and they cannot collude to keep prices high; price wars will inevitably ensue, decreasing revenue, profit margins, and valuation of the entire AI ecosystem;

I dont think its necessary for open source to lead the field. What will happen is what happens in all AI fields, slowdown and saturation. If AI is already good enough to perform certain tasks, people will resort to the model that is the cheapest and simplest to run and perform that task well.

Eventually, efficiency and model simplicity will be the most dominant differentiator rather than absolute performance.

10t models will simply not fly long term. AI has to become significantly cheaper to be able run in widespread applications.

 

[tphuang]

Southeast Asia will be the next data center hotspot (because the Middle East is a bit of a mess, and nobody will choose Europe), where a large number of Chinese companies are already building servers using Nvidia chips; the infrastructure is already there. If the executive order is indeed issued, some hosting providers will definitely choose the move there

The current executive order is insufficient to prevent such behavior.

they are already a huge hotspot. A lot of the Chinese CSP data centers running Nvidia chips are there. Not exactly a secret that ByteDance has at least 500 Blackwell racks there.

the elephant in the room is when/or if open source models surpass those of Anthropic/OpenAI on all metrics.

considering Kimi K3 is #1 in agentic capabilities proves that it is possible for an open source model to become the leading frontier model.

It is currently #4 but gap with #2 & 3 & 5 are tiny. So the 2nd tier below Fable.

Your question may be more appropriate for GLM-5.5.

Please, Log in or Register to view URLs content!

I think I brought this up a while back

Z.AI, the Chinese artificial intelligence company formerly known as Zhipu and focused on developing its GLM model platform, has

Please, Log in or Register to view URLs content!

construction of a massive 1-gigawatt data center powered entirely by Chinese-made chips. The facility has started partial operations and is designed to provide the computing capacity needed to develop Z.AI's most advanced GLM systems. Its power capacity is roughly equivalent to the electricity required to supply 750,000 homes at any given moment. Z.AI has also built or operated several computing clusters containing more than 10,000 chips each, suggesting the company is expanding its infrastructure as competition among Chinese AI developers intensifies.

 

[tphuang]

super nodes are the future. Biren has told media that Chinese models are going to 5T+ parameters next year and you can see the nodes coming out to support 5 to 10T models.

Keep in mind, that Chinese models stayed smaller (Kimi had the largest ones at 1.5T until recently) due to these limitations. SuperNode, can hold all the memories.

https://twitter.com/i/web/status/2079340425085620575

 

[Michael90]

the elephant in the room is when/or if open source models surpass those of Anthropic/OpenAI on all metrics.

considering Kimi K3 is #1 in agentic capabilities proves that it is possible for an open source model to become the leading frontier model.

this may happen in the near future given China's power advantage, delays in data center construction in the US, and innovations in model training and inference efficiency;

this will lead to an expected thinning out of the field; AI models that are falling behind (gemini, grok, etc) are likely to adopt open source models entirely to modify in house; this means an overcapacity of hardware (eg Musk / Meta leasing out excess capacity)

meanwhile, open source breaks the duopoly of Anthropic/OpenAI and they cannot collude to keep prices high; price wars will inevitably ensue, decreasing revenue, profit margins, and valuation of the entire AI ecosystem;

Lmao… don’t tell me it could become like the solar industry with price wars and race to the button. lol. I actually wish it ends up like that. It would be good for us consumers globally. So no company will really benefit from that , so it becomes a public good only.

 

[Michael90]

they are already a huge hotspot. A lot of the Chinese CSP data centers running Nvidia chips are there. Not exactly a secret that ByteDance has at least 500 Blackwell racks there.


It is currently #4 but gap with #2 & 3 & 5 are tiny. So the 2nd tier below Fable.

Your question may be more appropriate for GLM-5.5.

Please, Log in or Register to view URLs content!

I think I brought this up a while back

Impressive. I believe Zhipu is the first major Chinese AI lab/company to start using only Chinese AI chips exclusively for its data center, no? They were also the first in China to start training their multimodal AI model on just Chinese chips(Huawei ascend chips) the beginning of this year if I remember correctly ?