Performance wise that P100 is close to GeForce 1080. That GPU currently retails for $550.
For the price of 1 month of preemptive google’s IaaS you can buy the GPU and use it for as long as you want only paying very small amount (depending on where you are $0.02-$0.1/hour) for electricity.
I think its more about the flexibility of having access to more resources than you can ever need. If you buy a 1080 then you are limited by its performance and say you wanted to train to models at the same time, you are even more limited. Whereas with something like this you could train 10 models right when you need them trained, and you aren't stuck with 10 gpus that you bought so you could do that.
I understand the IaaS is way more flexible, e.g. in cases when you need lots of GPUs but not for long the cloud is a clear winner.
However, IMO the price is just not OK.
Compare with traditional servers. For $0.7/hour you can use e.g. c4.4xlarge amazon instance. The hardware is much more expensive than $550, just the RAM is already around $400. You won’t be able to purchase an equally-performing server for 1 month of that IaaS rent, I think it’ll be like 3-6 months (which is IMO reasonable).
They didn't add the K80s until February of 2017, did you really think they'd literally throw out hardware after less than 12 months? "Accelerated" means 18-36 month shelf life, not 10.
You’ll need at least some more components to run your GPU like a CPU, motherboard, etc... not to mention a reliable network, power and other ancillary services.
The real power of cloud is not a replacement of one instance but the ability to turn on tens, hundreds or even thousands of nodes.
Performance wise that P100 is close to GeForce 1080. That GPU currently retails for $550.
For the price of 1 month of preemptive google’s IaaS you can buy the GPU and use it for as long as you want only paying very small amount (depending on where you are $0.02-$0.1/hour) for electricity.