Fellow AI nerds, beware: Google Cloud glitch leaves Nvidia T4 GPUs off estimated bills for some virtual machines

If something seems too good to be true, it probably is. There appears to be a bug in the Google Cloud Platform online user interface that may lead engineers into thinking they're renting GPU-accelerated virtual machines for free, when, really, they're not. Anyone hoodwinked by the glitch will realize the Compute Engine …

[YetAnotherJoeBlow]

Thanks for the heads up, but alas I already believed in something to good to be true, and sadly I too will be drinking tap water tonight as well. I was doing some testing on some SHA3 algorithm tweaks that I made for modern hardware. I used 10 T4s unbelievable throughput on many GBs of data.

[tfewster]

It may be worth trying to dispute the charges. If you can keep a straight face when claiming you didn't know you had to pay for high-end processors.

[John Brown (no body)]

Considering the American business model of "contract is king" and their absolute dedication to weaselling around contract wording and the law looking for any possible way to screw over the customer, I think it's very much worth a try. The US has some quite strong bait and switch laws.

This isn't like the online ordering system where a TV accidentally goes on sale for $5 instead of $500 and they cancel all the chancers orders before they are dispatched. This is something that the purchaser actually sees as not increasing the cost of the ordered service and then actually getting the ordered service, ie the order as requested has been fulfilled with a price displayed. You can't then go on and say "oops, sorry, you owe us $000's now.

[Anonymous Coward]

It's not so far fetched that they might have been running a promotion that week.

[Raphael]

"If you can keep a straight face when claiming you didn't know you had to pay for high-end processors." The article does point out that there are some scenarios where you legitimately don't have to pay for them. (Just not the scenario they setup)

[DCFusor]

random?

While one would predict a certain number of errors are going to happen in anything so complex as this and many other systems, why is it that I hear about the kind that wind up blasting the consumer FAR more often than those where the company makes good on the error they made, or make one in the opposite direction. Pure randomness would say that there are a roughly equal number of errors in both directions - you know, there should be $1000/h instances listed on price sheets now and then - stuff like that. But no.

[Richocet]

Re: random?

This point is also true in Australian wage theft (there has been a surge in cases over the last 2 years).

The employers who own up to it use the excuse that their payroll system made errors. Funnily enough it was always to the detriment of the employee. What are the odds?

[ITBloke]

Can some of you please do me a favour and prove me wrong? Asking the Google assistant if it will cause world war 3 or if it will start world war 3 seems to never get recognised correctly. It always chooses world war 2 no matter how I say it. Now maybe it is because I'm a Yorkshireman with the associated accent and not because Google is plotting against us but can all you non Yorkshiremen put there give it a go and see if it recognises the same question, will Google cause wwiii? Or will Google start wwiii. Please post your results to this paranoid Brit. Thanks