> You will actually experience a lot less downtime by investing in your own wafer saw (or your own servers) rather than outsourcing (to a foundry or a cloud pro

> You will actually experience a lot less downtime by investing in your own wafer saw (or your own servers) rather than outsourcing (to a foundry or a cloud provider). Analogy doesn't work; "in-sourcing" of previously-IaaS resources as a cost-cutting measure, is used almost exclusively for "base load" (vs. elastic load, which is the comparative advantage of IaaSes); and so your in-house servers are usually running pinned at 100%. So the hardware components will wear out just as fast, if not faster, than the IaaS's servers will; and when they do, your ops team won't have as many spare parts on hand as the IaaS does; nor the ability to live-migrate the enforced VM to another exactly-equivalent substrate host in the fleet (which was already warm) to avoid downtime altogether. Also, apart from that, there are economies of scale in reliability. Your in-house backups are probably just a ZFS pool or a RAID5+1 array or something. They're certainly not hosted within a 17+-copies Dynamo-ish system, the way most IaaS object storage (incl. archival storage) is, because the costs of doing that in-house are ridiculous.