Governments are making a dangerous bet on the AI boom
economist.com
economist.com
What governments do (which is the actual content of the article) is betting on economic growth to exceed the growing debt.
1) economic reality of an aging population and the gdp = population * efficiency equation, which now in every last EU country requires constant reductions in government spending until, at minimum 2050
2) find some excuse to predict economic growth and keep (or even increase) spending, note: this is not something these governments necessarily actually believe, this is something to tell investors
This is an extremely self-serving money NOW effort where someone used a secondhand crayon to write "strategy" on it. But it's not like governments actually have a real choice in this, a message the people for example gave to French government, a big unmissable message set in flame across the streets of Paris.
The truth is that stock market margin debt, what usually sets of crises, is not at cycle highs, surprisingly, it's at maybe 60-70%. Government debt is high but is at about 70% of the hillbilly point, the point where new borrowing only gives you interest payments and no longer enables you to spend more. Which means company spending will rise and government spending will rise for ~2 years from this point forward.
It's the private debt which is the problem since individuals and enterprises don't have the capability the governments have to roll it over forever.
Now in 2020s, it seems only China is tailing US and in turn, US is wary of China. What gives?
The reality is LLMs are just tools, and not even close to as revolutionary of tools as computers themselves. Even Chinese companies understand that. It is the US alone that is in machine god mania, gambling their economy on this idea that the next model will let them rule the world if only we give it three more months.
Why Japan and Europe don't develop a model that is 20% worse at 0.001x the cost?
1. Price of electricity is low because of China's excellent energy policy. Admittedly this should absolutely be a priority of Japan/EU, but it isn't, so first you'd have to start there...
2. Because of #1, China already had massive infrastructure in place for eg. crypto mining that was readily repurposed. Japan/EU never engaged in this fad, so they have to start effectively from scratch in building out the GPU infrastructure.
3. Chinese talent is relatively isolated from the US economy. US-aligned economies must compete with the US for talent. It's hard to develop technology at a reasonable cost when all of your top talent is leaving to get paid 10x as much by speculative gamblers. If you don't have a very good reason to believe the gamblers are making a good bet, it's better to wait for them to go bust.
Added on to all of that... where is the upside? Chinese models are open anyways. There is nothing that merits this being a matter of urgent government intervention at all, it's basically pure downside.
US has really cheap energy and far more compute by a very large margin.
The real reason is chinese gov wants to crush US ai industries and own the market. Same thing they have successfully done multiple times in other industries now the difference is this time the US isn't seeming to back down.
I have yet to see any credible evidence that the Chinese government was involved with either the creation of DeepSeek or the open publication of it, in any way. Seems like usual yellow scare discourse, attributing everything to the bogeyman.
There is effectively no divide between private companies and state.
Jack ma being a perfect example.
So unless they think deepseek useless don't kid yourself
Personally I also always just use Gemini-3.5-Flash since for me it's actually a good search tool access that makes a model great.
But I'm not very typical European we have 2000x B300 and I'm in USA and China all the time, and work 16hrs a day without weekends and holidays.
“The Fifth Generation Computer Systems…was a 10-year initiative launched in 1982 by Japan's Ministry of International Trade and Industry (MITI) to develop computers based on massively parallel computing and logic programming” [1].
That sounds an awful lot like GPUs and HBM.
[1] https://en.wikipedia.org/wiki/Fifth_Generation_Computer_Syst...
A lot of what's happening now is the (very) belated delivery of what everyone was talking about when they talked about "expert systems."
Generally, they are a trailing indicator, like someone appearing on the Forbes millionaire under 30 list (Theranos, FTX). However, they are not always wrong.
Think of Economist writers as intelligent interns - people who couldn't make it in the world of finance, and with a particular ideological bent imposed on them by their editors.
What a load of crap. Most of the AI investment by governments is defense spending and it's been happening for years.
This whole article's premise is trying to rearrange the reasons for the large amount of spending and muddy waters. This isn't a bet on economic growth. This is purely a capabilities arms race and the cold war never ended.
The stock market is still business as usual. This spending is all coming from taxes and private equity mostly detached from those investments. The scaremongering about employment is a red herring. We all know it doesn't work that well. It doesn't matter. The only real losers are those trying to run a business too cheaply with AI. They tend to be the same startups always expected to lose anyway.
The counterfactual was never "cutting pensions and medicare". That wasn't on the table and no government would ever commit suicide and pass it. It was either, spend until the economy completely collapses under the weight of debt and hyperinflation, or do exactly the same thing but pray for a miracle while doing so.
So, we're doing the second one.