Tesla wants to monetize its cars to process AI workloads
EV carmaker Tesla is considering a wonderful money-making wheeze – use all of that compute power in its vehicles to process workloads for cash, like a kind of AWS on wheels.
The Elon Musk-led outfit said in its recent earnings conference call for calendar Q1 that it had noticed its vehicles spend a considerable amount of their time just sitting there not moving. Many pack in a decent amount of processing power, so why not get them to do something useful and earn some cash for the company as well?
Speaking on the conference call, Musk said that he thought most Teslas were probably used for about a third of the hours in a week.
"But then there's certainly going to be some hours left for charging and cleaning and maintenance in that world, you can do a lot of other workloads," added Tesla's Director of Autopilot Software, Ashok Elluswamy.
He claimed that AI companies have batch workloads where they send "a bunch of documents" that are run through large neural networks and "take a lot of compute to chunk through."
"And now that we have already paid for this compute in these cars, it might be wise to use them and not let them be, like, buying a lot of expensive machinery and leaving to them idle. We don't want that. We want to use the computer as much as possible and close to like basically 100 percent of the time to make full use of it," Elluswamy said.
"It takes a lot of intelligence to drive the car anyway. And when it's not driving the car, you just put this intelligence to other uses, solving scientific problems like a human or answering dumb questions for someone else," he added.
Musk compared it to Amazon Web Services, saying that people didn't foresee that AWS would become the most valuable part of Amazon.
"They [Amazon] found that they had excess compute because the compute needs would spike to extreme levels for brief periods of the year, and then they had idle compute for the rest of the year. So what should they do with that excess compute for the rest of the year? That's kind of, yeah, monetize it," Musk explained.
"If you get, like, to the 100 million vehicle level, which I think we will at some point get to, and you've got a kilowatt of usable compute – I think you could have on the order of 100 gigawatts of useful compute, which might be more than anyone, more than any company, probably more than any company," he mused.
Of course, all this compute capacity isn't sitting conveniently clustered together in a datacenter. It is distributed here and there, reached via a cellular connection in each Tesla, or possibly via Wi-Fi if the car is on the owner's driveway.
- Musk schmoozes Chinese Premier as Tesla Full Self-Driving remains parked
- Musk moves Tesla's goalposts, investors happily move shares higher
- Tesla misses the mark on all fronts in quarter of chaos
- Tesla slashes vehicle and self-driving-ish software prices as shares plummet
So the model Tesla would be looking at is perhaps more akin to edge computing, such as Heata in the UK, which uses heat from servers in homes to provide domestic hot water and rents out the compute capacity via cloud company Civo.
Among the issues we can see is that Tesla would be effectively using electricity that the car owner has paid for to run any workloads while it is idle, so would they get a cut of the money generated?
Yes, it seems. CFO Vaibhav Taneja, said "the capex is shared by the entire world. Sort of everyone owns a small chunk, and they get a small profit out of it maybe."
Would Tesla owners be able to opt out if they didn't want their vehicle used this way? That wasn't answered but it seems likely, that is, if the plan ever gets off the drawing board. There's no guarantee that it will actually come to fruition, for various reasons.
IDC Senior Research Director for Digital Infrastructure Andrew Buss said the idea sounds technically feasible, but the potential downsides are perhaps too big to justify it being actually implemented.
"They'd not even be edge processing nodes as the code and data would have to be centrally managed and stored and then packaged and sent for processing before being returned once complete," he told The Register.
Other downsides include third-party code and data running on a private asset, Buss said, and if taking power from the battery, this would accelerate the degradation of these, which are the single most expensive and crucial part of a Tesla and need to be kept in as optimal a shape as possible for longevity and consistency of range.
In other words, Tesla might well find that implementing this idea may prove more trouble than it is actually worth for the returns it generates.
And as The Register noted after the earnings conference, Elon has a habit of throwing out wild ideas when things aren't going well to distract the punters and energize investors. This could well be one of them. ®