Most places de facto depend on a single provider that becomes a cost center when using cloud providers too. While you can shop around, in effect it tends to get entrenched and get more and more dysfunctional as people try to exploit characteristics of your approvals process for using resources from your chosen provider. For a big company, you need a team to manage these resources - I've lost count of the number of recruiters who has contacted me just this year for roles involving management or architecture of cloud strategy etc. because they're building whole supporting organisations around a specific paradigm for hosting rather than building an organisation responsible for understanding and providing the best possible substrate for their applications.
I'm all for hybrid solutions, but mostly because it means you usually end up being able to cut the cost of on prem, colocation or managed servers even more relative to cloud setups because you can increase the utilisation rate (e.g. run your base load on cheap Hetzner servers but being able to spin up EC2 instances on short notice to take spices or handle failures up to and including all of Hetzner falling off the face of the earth). Most larger managed providers today also offers cloud services and many also offers colo services, so you can often easily mix and match as long as you're conscious off egress pricing which is often the biggest barrier to such setups today (especially if the big cloud providers are in the mix, though it's better than it was).
That said, I prefer not putting my eggs in one basket, and instead going the direction of putting in place layers to treat a multi-provider setup as one. In the past I've e.g. done zero downtime migrations between AWS->GCP->Hetzner that way, and also had systems split between actual on prem (as in a data room at our office), multiple colo's, dedicated servers at Hetzner, and a couple of VM providers where everything was transparent to our engineers (they didn't need to know on what continent a service ran unless doing performance optimisation or reliability engineering). We prepped to tie in AWS resources too, but it never become cost effective for that company to do so vs. the prices of the other providers used - it would have been if we had sudden extreme load spikes, but it was a business where traffic was linked closely to physical capacity at restaurants and nightclubs, and so traffic was very predictable.
I'm all for hybrid solutions, but mostly because it means you usually end up being able to cut the cost of on prem, colocation or managed servers even more relative to cloud setups because you can increase the utilisation rate (e.g. run your base load on cheap Hetzner servers but being able to spin up EC2 instances on short notice to take spices or handle failures up to and including all of Hetzner falling off the face of the earth). Most larger managed providers today also offers cloud services and many also offers colo services, so you can often easily mix and match as long as you're conscious off egress pricing which is often the biggest barrier to such setups today (especially if the big cloud providers are in the mix, though it's better than it was).
That said, I prefer not putting my eggs in one basket, and instead going the direction of putting in place layers to treat a multi-provider setup as one. In the past I've e.g. done zero downtime migrations between AWS->GCP->Hetzner that way, and also had systems split between actual on prem (as in a data room at our office), multiple colo's, dedicated servers at Hetzner, and a couple of VM providers where everything was transparent to our engineers (they didn't need to know on what continent a service ran unless doing performance optimisation or reliability engineering). We prepped to tie in AWS resources too, but it never become cost effective for that company to do so vs. the prices of the other providers used - it would have been if we had sudden extreme load spikes, but it was a business where traffic was linked closely to physical capacity at restaurants and nightclubs, and so traffic was very predictable.