> The hyperscalers have already outspent the most famous US megaprojects

[neilk@xoxo.zone]

@pjakobs @cwebber Inflation-adjusted. However, others point out that a percentage of GDP is a fairer comparison, see my reply.

[ghostonthehalfshell@masto.ai]

@cwebber

Reality only about 5% of 10% of all the data centers announced are actually being built. They don’t have the customers and they don’t have the energy to run the chips. If they do run, they last about 18 months.

That’s 18 months to make a return on investment.

[wronglang@bayes.club]

@cwebber yeah that should've gone into railroads, railroads and healthcare

[cassandravert@indieweb.social]

Future projections will be influenced by whether they can escape liability for the damage they cause.

[wilbr@glitch.social]

@faoluin @cwebber we won't need "taxes" when the water wars start, just copies of Paolo Bacigalupi's book "Water Knife", a weapon, and the holy guiding hand of Jesus himself

[gladtherescake@todon.nl]

@neilk @cwebber But GDP rises because of any investment, if you have a huge investment that leverages a huge part of your economy and then compare it to GDP it's going to mostly cancel out because you're putting two correlated variables at both ends of the equation. So I don't think that's a reasonable thing to do at all.

[buermann@mastodon.social]

@Spirit @cwebber

We can recycle the silicon and steel into the solar panels and wind mills we couldn't afford to build instead.

[buermann@mastodon.social]

@cwebber

What index did they use for the inflation adjustment?

[catmisgivings@stranger.social]

@cwebber now I'm looking at us railroads through the lens of a bubble investment and wondering if after all the loss of life and resettlement we could even call them a success from our standpoint in 2026

[mayintoronto@beige.party]

@cwebber What are we even doing here?

[buermann@mastodon.social]

@cwebber

Weird to adjust historic spending on infrastructure projects by a consumer price index. This is better.

X (formerly Twitter) (x.com)

[shaknais@mastodon.social]

@buermann

Percentage of GDP, for privately led projects doesn't seem a good index for me.

[criticalangle@mastodon.social]

@cwebber bubble

[buermann@mastodon.social]

@shaknais

It's a better proxy for productive capacity than a price index based on wildly different baskets of consumer goods over the centuries.

And since when were apollo or the Manhattan project or the international space station or the interstate highway system private programs?

[shaknais@mastodon.social]

@buermann

That's my point. None of those programs are private. However, the data centers _are_. It is unusual when the spending of private corporations outstrips that of their own government, on infrastructure.

Which is why GDP seems a bad fit, here. Because it uses the investments by private corps as part of its calculation, but doesn't fully cover government spending in the same period.

[iank@hostux.social]

@cwebber I suspect there is something important that is misleading about that graph, and I don't know quite what it is, but here are clues. A "data center" in practical / technical terms, is not defined by the computers inside it, but by a building (real estate), and the power & network connected to it. I walk through a floor of a data center in the Boston area every few weeks, for the past 5 years. I see maybe 200 racks, and room for 200 more. A large empty room, extremely little change. (1/2)

[iank@hostux.social]

@cwebber The company, https://en.wikipedia.org/wiki/American_Tower , has 27 data centers, 10 billion in revenue, it is not a small player, and its revenue and stock price have been flat over the past few years. So, I have to wonder how much of this kind of this stuff is at the same level of reporting as science reporting constantly telling us about the breakthrough that is about to cure cancer.

[marcel@waldvogel.family]

@darkuncle @cwebber
It's a measly 4 or 5 billion USD (4.3*10⁹ CHF)
https://home.cern/resources/faqs/facts-and-figures-about-lhc

[bms48@mastodon.social]

@cwebber OOPSIE!

[mhartle@mastodon.online]

@cwebber Contrast this with current revenue generated by AI providers running on top of these and future necessary revenue these providers would need to generate during the lifetime of these data centers for a positive ROI.