@freagle@lemmygrad.ml pertinent to a discussion we had earlier
I think this analysis misses a critical point, pretty early.
He states that US manufacturing for the components needed, like steel and copper (not to mention other critical components like fiber, coolant, etc), is 30% more expensive, at least. Which is true.
Then he says that on one side big tech is sucking up all the world’s capital and on the other side manufacturing is collapsing.
But clearly these things are deeply interrelated. Big tech, with all of that capital, is going to put out orders for manufacturing. Those orders will go everywhere, including China. But clearly some of those orders will end up with US factories, European factories, LatAm factories, etc.
So the manufacturing sectors of the West clearly have an opportunity to take on new orders, and that will drive secondary investments as people try to position themselves to receive the money big tech is spending. Sure, you can’t build every type of factory needed in 1 year, but it’s entirely possible for people to put down 5-year bets right now to build new factories in LatAm, where it’s cheaper, and in the meantime max out what manufacturing is still present in the West.
The combination of these two things can defer the economic collapse being referred to. Even this essay talks about the timeline when referring to $1.5T in bonds by 2028 - 2 years.
Yes. There are critical gaps in the supply chain - magnets and chips in particular - but the shortages aren’t here yet, only the risks of shortages. The shortage that I’m most aware of right now is actually gas generators, where there’s a 1 - 2 year wait-list right now.
So I do think there’s serious problems, I do think that the system is hurtling towards collapse, but I don’t think it’s going to happen in 2026, I don’t think it’s going to happen in 2027. And until we know what happens in 2026, we won’t know what changes will attenuate the crises and extend the timeline.
Edit:
Part of the issue is that people are using historical data center examples for their analysis. The reality on the water situation is that evaporative cooling is being phased out, electricity usage overhead are coming down, new form factors are already being rolled out to increase power density while decreasing costs.
But the real problem is going to be in the real estate world. The hyperscalers do not own every data center. They have commercial leases for many of them. Those leases are obviously securitized by Wall St. And when those default, there will be a 2008 problem.
This is where we disagree. You’re assuming that big tech will make meaningful investments and actually produce things. I think they’re just pumping up their stock and nothing of value will actually be produced. The fact that the proposed data centres will supposedly be built in places like Texas and Arizona where there’s water scarcity kinds of highlights that none of these plans are serious. Efficiency and water usage may be going down, but that’s more than made up for by the sheer scale of proposed rollouts.
Meanwhile, the shortages are absolutely here already as evidenced by the current crisis over yttrium supply. China’s export restrictions have effectively halted shipments to the US, and caused European prices to skyrocket by over 4,000%. The lack of supply is creating severe shortages for aerospace and semiconductor manufacturing today.
Market analysts forecast significant supply deficits by 2030, while China is projected to retain 78% of global refining capacity even as far as 2040 under the most aggressive diversification scenarios.
And given that these GPUs have a duty-cycle that’s just two to three years, and as low as 54 days when they are run hard, that gives us a clue to the scale of the shortages we can expect. The current crop of chips will be nearing the end of their life cycle just as the rare earth shortage is about to hit the hardest.
Scale of data center doesn’t matter for water consumption when the cooling system is a closed loop, which is what data centers are moving too. Evaporative cooling doesn’t even work for the current crop of GPUs, it doesn’t transfer heat fast enough. So there are multiple pressures against evaporative cooling and towards closed loop liquid cooling. That’s an example of investments going to something real - the cooling solutions is one of those areas that orders are going to go up for and there will be non-tech investments in it in order to capture profit from the irrational exuberance of the a.i. gold rush.
I have seen the technologies changing in the data center world due to my work. I am aware that there are actual changes happening and that literally the CEOs of some of the inglorious 7 are having private meetings about these specific topics with people who are allocating capital to solving these specific problems. It’s not pure theater.
It could be too little too late, bit it’s not merely hand waving. There’s mobilization actually happening on the ground.
You’re right, if the proposed data centers end up using closed loop cooling then water might not be an issue there. It could be a serious effort happening in the US, but I still don’t see how the rare earth problem is going to be addressed. Exports from China are going to be drying up because China is increasingly using them for domestic production, and domestic demand will only accelerate as China expands production of advanced chips, solar panels, EVs, robots, and so on. Given that current chips will burn out right around when we expect to see a rare earth supply bottleneck in 2030s, where are the materials for new chip production going to come from?
Venezuela, that’s where they are gonna come from. Oil also a big player, but rare earths in the Bolivarian nation are they key target of US imperialism.
I don’t see how Venezuela helps here. It’s not the mining of rare earths that’s the really hard part, but the refining tech that China has an effective monopoly on. Replicating that will be a decades long project.
I can only see from the current US administration motivations to wrap up existing conflicts and tribulations (Israel and Gaza, Ukraine) to have the resources necessary to open up new fronts for imperial aggression and secure raw materials to keep propping up their economy to avoid total collapse from the current AI bubble.
The current US military presence off the coast of Venezuela, legitimized by the farce of the “War against Drugs,” aims at control of critical raw materials and the geopolitical containment of BRICS partners, namely China and Russia. In the context of these resources, oil plays only a subordinate role: Venezuela possesses with bauxite, nickel, gold, and coltan raw materials that are of central importance for the global race for technological supremacy and especially for the development of Artificial Intelligence.
Venezuela: Burnt Dragon (kritikpunk)
And it kinda shows i mean, the “Donroe” doctrine is real and is here to stay
This is the only way a dying empire can cope with the fact they cannot compete with China and that they lost.
Addendum: Ecuador is going to shit, Perú is in political turmoil and soon Chile will be up for grabs in the “soft” dominance front by way of the incumbent second round of presidential elections this december (if José Antonio Kast is elected)
The rare earth problem can’t be addressed. China will decide how it plays out. In one sense, it seems like they will string the US along. In another it seems like they will just let the US starve. In another it seems like they’ll just let it flow as long it serves their development goals. I have hunches, but really it’s up to their strategies.
One way it could work is that China becomes an exporter to the US for these things in a sort of mutual agreement for the US bubble to maintain. It could be that a.i. in the US is running off Huawei in 3 years. I don’t really know.
I think it’s all doomed for the US project. But it’s not next year unless something big happens. And it’s unlikely to be in 2027, but things are operating on a 3-month cycle so 2027 could be the year given what happens next year. I don’t really know.
I just think the US has sufficient positioning, sufficient reserves, sufficient capabilities and capacities, and sufficient drice to innovate that it can keep this going for a while.
I expect that China will prioritize domestic use, because why wouldn’t they. And if that’s going to be the case then it’s no longer a question of politics, the economics are going to be the deciding factor here. Chinese businesses are going to need rare earths as inputs, and there won’t be enough left to export. It’s really that simple in my opinion. Even if CPC wanted to help the US out for some reason, it would be politically impossible for them to say we’re going to starve our own industries to feed the US.
I agree it’s unlikely that this particular problem will surface in the next couple of years. I expect the big problems will start closer to 2030 when we start seeing serious supply shortages and the current crop of chips starts burning out. But a lot of other things can happen before then as well. The US economy is clearly in a very fragile state right now, and there could be any number of black swan events before this particular bottleneck is hit.
Yes. This matches my thinking as well.
I found a YouTube link in your post. Here are links to the same video on alternative frontends that protect your privacy:
