“Why a US AI “Manhattan Project” Could Backfire: Notes from Conversations in China”, 2024-11-29 (; similar):
My recent two weeks in China suggested something surprising about its AI landscape: the biggest bottleneck isn’t compute—it’s commitment.
Despite export controls, Chinese labs can access both legal NVIDIA A800s and black-market NVIDIA H100s. Cloud computing costs are comparable to the US (maybe lower). The export restrictions just aren’t binding at current scale.
Instead, the even bigger constraint appears to be funding. Consider:
Chinese venture capital is much smaller than Western VC (according to a quick googling probably 20–40% as large).
Tech giants like Tencent and Alibaba generate ~20% of Google/Microsoft’s profits (and these have recently been declining)
While the government has deep pockets, it hasn’t yet made substantial AI allocations
This means that for Chinese companies to match Western AI investments, they’d need to bet a much larger share of their resources. But they’ve actually shown less interest in AGI, and are instead content to use trailing-edge models at a fraction of the cost. There’s no GPT-5-scale model in development.
When Chinese teams do get equal compute resources, they deliver impressive results—Alibaba’s Qwen 2.5 and Tencent’s Hunyuan rank among the best open-weight models. DeepSeek recently matched GPT-4 on many benchmarks. They have the capabilities; they just aren’t going all-in.
…So what might trigger a wake up? Most people said they didn’t know. But one suggestion was that the fastest way would be a high-profile US state-led AI project (especially if its explicit goal is US dominance…).
This means calls for a US “Manhattan Project” for AGI might easily be self-defeating. If maintaining a technological lead is your goal, better to STFU and hope the status quo persists as long as possible. (Or if you do go ahead, you need much stricter export restrictions.)
Twitter: Several people pointed out DeepSeek & ByteDance have said they’re more compute constrained than funding constrained.
I think the explanation might be that it’s been possible to buy ~10k chips, but if you try to buy ~100k, that’s much harder. Different contacts are talking about different margins.
…However, I still think this constraint could be overcome with enough funding. Maybe Chinese firms could pay a ~50% premium and get hold of leading chips (especially with govt support).
Alternatively, Epoch just released a report finding that even if your chips lag by 10 years, you can still train a leading model at 10× the cost.
That suggests if your chips lag by only two years, then the cost premium might be under 2×.
So for example if GPT-6 costs $10b to train, then China could do it for under $20b, which is still feasible for the government.