Whilst manufacturing has more practical value than much of what's in datacentres[1] it doesn't necessarily follow that it's the more valuable strategic play or route to long term economic growth (bubble or no bubble)

Food is more critical to us than Alibaba crap or even top notch smartphones and would be missed rather more if we had a nuclear exchange, but Chinese entrepreneurs would rather own factories exporting manufactured goods than rice paddies, and for good reason. The Chinese government would like to be self sufficient with food production, but unlike many less technologically developed countries it isn't. Most of those countries stay poor though. Food doesn't change or scale as much as manufacturing, which possibly doesn't change as fast or scale as big as compute. Plus in theory at least, some of the compute is being devoted to automating away dependency on offshore labour for manufacturing, although I suspect China is generally ahead in the manufacturing automation areas of AI anyway...

[1]I'd rather have kitchen appliances than novelty image generation, spamblogs or ad retargeting, but then again I'd also rather have access to knowledge and communications than plastic toys...

> but then again I'd also rather have access to knowledge and communications than plastic toys

My point is that you can get that with ~1/1000th the compute we spend. The truly useful parts of the internet are a tiny fraction of the tech footprint. There's more net value for us in Wikipedia than there is in ~a trillion dollars of the more vapid tech firms. And it runs on 3m/year in hosting costs.

You are right that China has a food security problem, though. I am, however, assuming that its government isn't blind to it - the country is self-sufficient for high-calorie staples, and wouldn't starve even if all food imports stopped.