Leading American automaker Tesla has long been striving to advance its self-driving technology. However, it needs high enough computing capability to achieve its target of full autonomy.
In order to do that, Tesla announced plans to invest billion dollars in the project’s development.
About the investment
Tesla will allocate a whopping $1 billion to its Dojo supercomputer. Specifically, the automaker will use the funds for the project’s capital expenditures and R&D.
As EV-a2z previously reported, the technology is a powerful AI training platform that will aid the automaker in achieving true FSD capabilities.
In that sense, CEO Elon Musk stated that Tesla would require high computer power to achieve level 5 autonomy at the company’s Q2 2023 Earnings Call. This level of computing power will support accelerated neural network training, which has been a major impediment to the technology’s advancement.
“The fundamental rate limiter on the progress of full self-driving is training. If we had more training compute, we could get it done faster.”
Tesla CEO Elon Musk
CEO Musk further noted during the company’s recent earnings call that Tesla will soon start producing its training computer.
As per the plan, production will begin sometime this month. However, it still must be noted that Tesla may encounter some delays.
As of now, Tesla already has an Nvidia GPU-powered supercomputer. Meanwhile, the new Dojo supercomputer will leverage Tesla-developed chips.
See Also:
- Tesla boss allusively confirms Dojo supercomputer now operational
- Tesla Chief updates his outlook for the coast-to-coast FSD demonstration
- Tesla announces plans to begin Dojo supercomputer production next month
- Tesla CEO comments on Cruise Robotaxi’s mishap
- Tesla discreetly tests FSD software in Australia, per report
Tesla CEO Musk claimed that the new Dojo could support an exaflop, or one quintillion floating-point operations, every second.
In addition, the automaker anticipates the technology to generate 100 exaflops by October 2024.
That said, Dojo can accomplish certain tasks in just a second, while a regular desktop computer will take billions of years to perform.
