I talked with Herbert Ong about Tesla compute and how I believed Tesla had 100,000 H100 ordered. I only wondered if they were installed and operating.
I also told Herbert that I expected large Nvidia B200 orders by Tesla and XAI so that they each add another 800 Exaflops of compute by the end of this year for a total of 1200 Exaflops or more.
It was from about 4:30-5:30 when I specifically discussed the number of H100s that Tesla has. I also gave a talk to over 20 people on April 13, 2024 where I used the slide above describing the 100,000+ Nvidia H100s that I believed Tesla already had.
This compute will drive the accelerated development of full self driving (Tesla FSD) and the development of robotaxi.
Chris Zheng confirms that Tesla has bought over 100,000 Nvidia H100s. This purchase by Tesla is confirmed by a source at Nvidia. Nvidia H100s have 4 Petaflops of compute. The 100,000 Nvidia H100s is 400 Exaflops of compute.
Tesla bought a ton of compute and now has over 100,000 H100s, which Elon will probably talk about on the earnings call.
— Chris Zheng (@ChrisZheng001) April 19, 2024
by someone from NVIDIA.
— Chris Zheng (@ChrisZheng001) April 19, 2024
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
“Tesla Has 400 Exaflops of AI Training Compute”
This is massive news. It seems like the mainstream press and the majority (read all?) of the popular enthusiast web-news sites (e.g. CleanTechnica) have no insight into what is happening around AI, particularly the rate of change. OK, it’s not directly comparable (the super computer is probably a 64 bit FP), but the biggest/fastest super computer in the world from 2020 had only 400 ExaFLOPS – now 4 companies are pretty much in the same place – at least with respect to ML.
I am really excited to see the results are we both learn more about tuning and training alongside applying these mammoth compute arrays. I’m expecting FSD, for example to exceed human capabilities (i.e. make fewer errors) in one year from now. Also, on the ChatGPT side, logic and reasoning is something else I will be watching closely for progress.
This aI race is brutal. The numbers for meta and Ms were probably for 2023. They have much more of h100 now, way more than Tesla. I doubt Tesla can compete. They will focus on self driving and vision based approach for robots. It is possible lets wait for Grook 2. Meta is probably the worst here, I dont trustcthem a bit, than Ms a bit better, Google and I would trusk Musk with AI the most of the bunch.
I’m wondering what apple is doing
Tim Cook has his thumb up his nose. Tim missed the EV party. Tim is a no show for the AI party. Tim has a dumb watch to sell you. Any color.
Selling overpriced phones and making good bucks of it. Their market is not in danger from meta, Ms,.. so they dont need to go into AI race.
They already have an AI in their phones, called Siri, and by today’s standards it’s junk. They could really stand to catch up. Google will certainly get some good AI in Android phones, probably before all that long.