[rsynnott]

At present, no, of course not. No company maintains surplus capacity sufficient to absorb its competitors' business if they withdraw from or are excluded from the market. You'd assume in practice there'd have to be a transition period, during which the likes of Hetzner would be... busy.

(More likely, there's another round of negotiation, and some new bandaid solution is produced; not like it's the first time. No-one, or almost no-one, really _wants_ this to break down entirely; the fallout would be widespread.)

It does seem reasonable to expect that the rate of companies moving stuff out of US-based infrastructure providers will increase, though; the whole thing is very fragile.

[pjc50]

I don't think you can rely on "no one will smash US institution X because that would be bad for everyone" any more.

[rsynnott]

You've got to imagine there are limits, tho. I note that Trump backed off most tariffs, at least for now, when the markets got unhappy (I mean, you could believe it was due to symbolic troop movements if you wanted, I suppose?) And cloud services to Europe are _big_ business; it would not be a small market shock.

If there were to be a major migration from AWS and Azure to the likes of Hetzner, OVH and friends, also, that would likely be _permanently_ lost business for US megacorps; no-one does that sort of migration unless they really have to, so it's improbable that anyone would move back if and when the situation was resolved.

[pjc50]

I thought he only backed off the Canadian tariffs, and only because there were retaliatory tariffs plus some symbolic concession? The China ones and the de minimis change are still in place.

Bezos turning up at the inauguration and directing the WaPo to not endorse Harris are strong hints that Amazon is probably going to be fine, but I would say that nothing is certain when dealing with someone who's deliberately unpredictable and willing to threaten allies.

[rsynnott]

> and only because there were retaliatory tariffs

Those were completely inevitable, though; the game theory behind all this stuff essentially requires them.

> plus some symbolic concession

A really utterly meaningless one, though. I'm fairly convinced that pissed-off markets were the major factor.

> I thought he only backed off the Canadian tariffs, and only because there were retaliatory tariffs plus some symbolic concession? The China ones and the de minimis change are still in place.

Also Mexico. I'd suspect most of the Chinese ones aren't long for this world, either.

[yibg]

But this isn't just cloud infra / platform like AWS etc right, but also various SaaS products. e.g. Office 365, gsuite, github etc. Are there even equivalent (enough) versions of all those that are not only hosted but owned by EU companies?

[rsynnott]

Those are also less _critical_/replaceable, though. Up until recently, most companies didn't use cloud-y office things, they used, typically, Microsoft Office 2xxx (ie the non-cloud version). Microsoft's cloud-y Office solution is only 7 years old; while Google's is older, it wasn't taken particularly seriously for a long time. Many companies (actually I would suspect _most_ companies) _still_ use on-prem Office/Sharepoint Server/Exchange Server setups, and Microsoft still sells this stuff (Office 2024 LTSC is the latest one for enterprise, Office 2024 for consumers).

As for Github, self-hosted or vendor-hosted GitLab would be the obvious solution (self-hosted Github _is_ a thing, but only for large enterprises IIRC); other GitHub-like things are available.

I also suspect that Github in particular, and maybe MS, could, if desired, rework their services such that they didn't actually touch personal data in a form that they could disclose to the US government (which is the core issue here). This could be managed via using a third-party auth service (which typically these sort of services already support for enterprise integrations) and, for the Office-y apps, end-to-end encryption.

Replacing AWS and Azure and friends would in many ways be the big problem, especially if all this were to happen quickly (in practice, there'd almost inevitably be a significant grace period if things broke down). There's a big capacity problem there; all of these sorts of services operate basically at capacity, because economically it makes no sense to do anything else. That said, in the doomsday scenario, Amazon et al would presumably end up selling off a lot of data centres in Europe (restricted to only non-personal-data applications, they'd need fewer).

[yibg]

I agree each thing is not difficult to replace, but replace all SaaS products that are used might be more challenging. Some data out there from a quick google is saying companies use on average 300+ SaaS products. Not all of them are going to be from US companies, but probably a large amount is, so we're looking at probably over a hundred or 2 of products that need to be replaced with local or EU ones across all companies in Europe. That seems like a lot of work and disruption.

[rsynnott]

Oh, yeah, it’d be a complete nightmare for everyone involved. But it’s likely doable; in particular providers would be _strongly_ incentivised to provide compliant solutions (again, many SaaS providers could manage this by avoiding directly touching PII).