Alibaba is having its smart hardware exhibition and unveiling new multimodal interaction dev kit Pre-loaded w/ > 10 Agent & MCP tools
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
Alibaba is having its smart hardware exhibition and unveiling new multimodal interaction dev kit Pre-loaded w/ > 10 Agent & MCP tools
It´s not difficult to understand why the Chinese govt is not thrilled about the prospect of Chinese AI companies importing more US GPUs from Nvidia. Why help cement US hegemony in chip sales?
The US compute advantage is real and massive. Yet, while compute is important, it is also not everything. Meta had a huge budget but didn't deliver much. DeepSeek and others show that high talent density punches way above its weight. I would not bet against Chinese AI companies, but we have to acknowledge that they will have compute disadvantages compared to their US peers for years to come. It's just silly to pretend otherwise.
And here's broken down by country. Note: this includes Chinese companies buying NV GPUs.
The US compute advantage is real and massive. Yet, while compute is important, it is also not everything. Meta had a huge budget but didn't deliver much. DeepSeek and others show that high talent density punches way above its weight. I would not bet against Chinese AI companies, but we have to acknowledge that they will have compute disadvantages compared to their US peers for years to come. It's just silly to pretend otherwise.
And here's broken down by country. Note: this includes Chinese companies buying NV GPUs.
this chart is probably crap. Anyone that thinks US currently has 10 to 15x China's compute access is not worth following.
China‘s total computing capacity is around 788 EFLOPS, while US get around 10 million H100 which is roughly equivalent to 10,000 EFLOPS, so 10x is reasonable.
But the total revenue of AI model is around 1/10 of the cost of buying these chips, so 10x also means huge deficit.
But the total revenue of AI model is around 1/10 of the cost of buying these chips, so 10x also means huge deficit.
Last edited:
actually, we don't know exactly how much compute China has or has access to in surrounding asian data centers.
But a cursory look at how much revenue Nvidia gets from its Asia region will show just how much compute is bought by Chinese players who either put the data center in Malaysia, Japan, Indonesia or smuggle them in.
Qwen's open-source multimodal model is quite impressive, but its Qwen3 MAX is already significantly behind, and thinking still hasn't been fixed. I haven't used it for a long time.
The Information reports that Deepseek will release V4 around the Spring Festival. Source is kind of "Trust me, bro". They're probably just guessing based on the previous mHC paper.
Such hype expectations are somewhat of a psychological preparation and hedging strategy. If Deepseek releases as they say, then it's "expected." If not (Deepseek hasn't promised any timeline), these media can spout old dog whistles like "computing bottlenecks" or "training failures."
Such hype expectations are somewhat of a psychological preparation and hedging strategy. If Deepseek releases as they say, then it's "expected." If not (Deepseek hasn't promised any timeline), these media can spout old dog whistles like "computing bottlenecks" or "training failures."
Last edited:
This was a long-standing rumor before Jukan posted about it, and honestly... Jukan just reads public information from what I can tell.