For years, Elon Musk has talked about Dojo — the AI supercomputer that would be the cornerstone of Tesla’s AI ambitions. It’s vital sufficient to Musk that in July 2024, he mentioned the corporate’s AI staff would “double down” on Dojo within the lead-up to Tesla’s robotaxi reveal, which occurred in October.
However what precisely is Dojo? And why is it so essential to Tesla’s long-term technique?
In brief: Dojo is Tesla’s custom-built supercomputer that’s designed to coach its “Full Self-Driving” neural networks. Beefing up Dojo goes hand-in-hand with Tesla’s objective to succeed in full self-driving and produce a robotaxi to market. FSD, which is on tons of of 1000’s of Tesla autos in the present day, can carry out some automated driving duties however nonetheless requires a human to be attentive behind the wheel.
Tesla’s Cybercab reveal has come and gone, and now the corporate is gearing as much as launch an autonomous ride-hail service utilizing its personal fleet of autos in Austin this June. Tesla additionally mentioned throughout its 2024 fourth-quarter and full-year earnings name on the finish of January that it plans to launch unsupervised FSD for U.S. prospects in 2025.
Musk’s earlier rhetoric has been that Dojo can be the important thing to reaching Tesla’s objective of full self-driving. Now that Tesla seems near nearing that objective, Musk has been mum on Dojo.
As a substitute, ever since August 2024, speak has been round Cortex, Tesla’s “large new AI coaching supercluster being constructed at Tesla HQ in Austin to unravel real-world AI.” Musk has additionally mentioned it is going to have “large storage for video coaching of FSD & Optimus.”
In Tesla’s This fall shareholder deck, the corporate shared updates on Cortex, however nothing on Dojo.
Tesla has positioned itself to spend massive on AI and Dojo — and now Cortex — to succeed in its objective of autonomy for each vehicles and humanoid robots. And Tesla’s future success actually hinges on its means to nail this down, given the elevated competitors within the EV market. So it’s value taking a better take a look at Dojo, Cortex, and the place all of it stands in the present day.
Tesla’s Dojo backstory
Musk doesn’t need Tesla to be simply an automaker, or perhaps a purveyor of photo voltaic panels and vitality storage programs. As a substitute, he needs Tesla to be an AI firm, one which has cracked the code to self-driving vehicles by mimicking human notion.
Most different firms constructing autonomous car expertise depend on a mix of sensors to understand the world — like lidar, radar and cameras — in addition to high-definition maps to localize the car. Tesla believes it could actually obtain totally autonomous driving by counting on cameras alone to seize visible knowledge after which use superior neural networks to course of that knowledge and make fast selections about how the automotive ought to behave.
As Tesla’s former head of AI, Andrej Karpathy, mentioned on the automaker’s first AI Day in 2021, the corporate is principally attempting to construct “an artificial animal from the bottom up.” (Musk had been teasing Dojo since 2019, however Tesla formally introduced it at AI Day.)
Corporations like Alphabet’s Waymo have commercialized Degree 4 autonomous autos — which the SAE defines as a system that may drive itself with out the necessity for human intervention underneath sure situations — by means of a extra conventional sensor and machine studying method. Tesla has nonetheless but to supply an autonomous system that doesn’t require a human behind the wheel.
About 1.8 million individuals have paid the hefty subscription worth for Tesla’s FSD, which at present prices $8,000 and has been priced as excessive as $15,000. The pitch is that Dojo-trained AI software program will ultimately be pushed out to Tesla prospects through over-the-air updates. The size of FSD additionally means Tesla has been in a position to rake in tens of millions of miles value of video footage that it makes use of to coach FSD. The concept there’s that the extra knowledge Tesla can gather, the nearer the automaker can get to truly reaching full self-driving.
Nevertheless, some business specialists say there is likely to be a restrict to the brute power method of throwing extra knowledge at a mannequin and anticipating it to get smarter.
“Initially, there’s an financial constraint, and shortly it is going to simply get too costly to try this,” Anand Raghunathan, Purdue College’s Silicon Valley professor {of electrical} and pc engineering, informed TechCrunch. Additional, he mentioned, “Some individuals declare that we would truly run out of significant knowledge to coach the fashions on. Extra knowledge doesn’t essentially imply extra data, so it relies on whether or not that knowledge has data that’s helpful to create a greater mannequin, and if the coaching course of is ready to truly distill that data into a greater mannequin.”
Raghunathan mentioned regardless of these doubts, the development of extra knowledge seems to be right here for the short-term a minimum of. And extra knowledge means extra compute energy wanted to retailer and course of all of it to coach Tesla’s AI fashions. That’s the place Dojo, the supercomputer, is available in.
What’s a supercomputer?
Dojo is Tesla’s supercomputer system that’s designed to operate as a coaching floor for AI, particularly FSD. The title is a nod to the area the place martial arts are practiced.
A supercomputer is made up of 1000’s of smaller computer systems known as nodes. Every of these nodes has its personal CPU (central processing unit) and GPU (graphics processing unit). The previous handles general administration of the node, and the latter does the advanced stuff, like splitting duties into a number of components and dealing on them concurrently. GPUs are important for machine studying operations like those who energy FSD coaching in simulation. Additionally they energy giant language fashions, which is why the rise of generative AI has made Nvidia probably the most worthwhile firm on the planet.
Even Tesla buys Nvidia GPUs to coach its AI (extra on that later).
Why does Tesla want a supercomputer?
Tesla’s vision-only method is the primary purpose Tesla wants a supercomputer. The neural networks behind FSD are skilled on huge quantities of driving knowledge to acknowledge and classify objects across the car after which make driving selections. That signifies that when FSD is engaged, the neural nets have to gather and course of visible knowledge constantly at speeds that match the depth and velocity recognition capabilities of a human.
In different phrases, Tesla means to create a digital duplicate of the human visible cortex and mind operate.
To get there, Tesla must retailer and course of all of the video knowledge collected from its vehicles around the globe and run tens of millions of simulations to coach its mannequin on the info.
Tesla seems to depend on Nvidia to energy its present Dojo coaching pc, however it doesn’t wish to have all its eggs in a single basket — not least as a result of Nvidia chips are costly. Tesla additionally hopes to make one thing higher that will increase bandwidth and reduces latencies. That’s why the automaker’s AI division determined to provide you with its personal {custom} {hardware} program that goals to coach AI fashions extra effectively than conventional programs.
At that program’s core is Tesla’s proprietary D1 chips, which the corporate says are optimized for AI workloads.
Inform me extra about these chips
Tesla is of an analogous opinion to Apple in that it believes {hardware} and software program needs to be designed to work collectively. That’s why Tesla is working to maneuver away from the usual GPU {hardware} and design its personal chips to energy Dojo.
Tesla unveiled its D1 chip, a silicon sq. the dimensions of a palm, on AI Day in 2021. The D1 chip entered into manufacturing as of a minimum of Might this yr. The Taiwan Semiconductor Manufacturing Firm (TSMC) is manufacturing the chips utilizing 7 nanometer semiconductor nodes. The D1 has 50 billion transistors and a big die dimension of 645 millimeters squared, in keeping with Tesla. That is all to say that the D1 guarantees to be extraordinarily highly effective and environment friendly and to deal with advanced duties shortly.
“We will do compute and knowledge transfers concurrently, and our {custom} ISA, which is the instruction set structure, is totally optimized for machine studying workloads,” mentioned Ganesh Venkataramanan, former senior director of Autopilot {hardware}, at Tesla’s 2021 AI Day. “This can be a pure machine studying.”
The D1 continues to be not as highly effective as Nvidia’s A100 chip, although, which can be manufactured by TSMC utilizing a 7 nanometer course of. The A100 comprises 54 billion transistors and has a die dimension of 826 sq. millimeters, so it performs barely higher than Tesla’s D1.
To get a better bandwidth and better compute energy, Tesla’s AI staff fused 25 D1 chips collectively into one tile to operate as a unified pc system. Every tile has a compute energy of 9 petaflops and 36 terabytes per second of bandwidth, and comprises all of the {hardware} obligatory for energy, cooling and knowledge switch. You’ll be able to consider the tile as a self-sufficient pc made up of 25 smaller computer systems. Six of these tiles make up one rack, and two racks make up a cupboard. Ten cupboards make up an ExaPOD. At AI Day 2022, Tesla mentioned Dojo would scale by deploying a number of ExaPODs. All of this collectively makes up the supercomputer.
Tesla can be engaged on a next-gen D2 chip that goals to unravel data stream bottlenecks. As a substitute of connecting the person chips, the D2 would put the whole Dojo tile onto a single wafer of silicon.
Tesla hasn’t confirmed what number of D1 chips it has ordered or expects to obtain. The corporate additionally hasn’t supplied a timeline for the way lengthy it is going to take to get Dojo supercomputers working on D1 chips.
In response to a June publish on X that mentioned: “Elon is constructing an enormous GPU cooler in Texas,” Musk replied that Tesla was aiming for “half Tesla AI {hardware}, half Nvidia/different” over the following 18 months or so. The “different” could possibly be AMD chips, per Musk’s comment in January.
What does Dojo imply for Tesla?
Taking management of its personal chip manufacturing signifies that Tesla would possibly sooner or later be capable of shortly add giant quantities of compute energy to AI coaching packages at a low price, significantly as Tesla and TSMC scale up chip manufacturing.
It additionally signifies that Tesla could not need to depend on Nvidia’s chips sooner or later, that are more and more costly and arduous to safe.
Throughout Tesla’s second-quarter earnings name, Musk mentioned that demand for Nvidia {hardware} is “so excessive that it’s usually troublesome to get the GPUs.” He mentioned he was “fairly involved about truly having the ability to get regular GPUs once we need them, and I believe this due to this fact requires that we put much more effort on Dojo as a way to make sure that we’ve acquired the coaching functionality that we’d like.”
That mentioned, Tesla continues to be shopping for Nvidia chips in the present day to coach its AI. In June, Musk posted on X:
Of the roughly $10B in AI-related expenditures I mentioned Tesla would make this yr, about half is inside, primarily the Tesla-designed AI inference pc and sensors current in all of our vehicles, plus Dojo. For constructing the AI coaching superclusters, Nvidia {hardware} is about 2/3 of the price. My present greatest guess for Nvidia purchases by Tesla are $3B to $4B this yr.
“Inference compute” refers back to the AI computations carried out by Tesla vehicles in actual time and is separate from the coaching compute that Dojo is liable for.
Dojo is a dangerous wager, one which Musk has hedged a number of instances by saying that Tesla won’t succeed.
In the long term, Tesla might theoretically create a brand new enterprise mannequin based mostly on its AI division. Musk has mentioned that the primary model of Dojo can be tailor-made for Tesla pc imaginative and prescient labeling and coaching, which is nice for FSD and for coaching Optimus, Tesla’s humanoid robotic. Nevertheless it wouldn’t be helpful for a lot else.
Musk has mentioned that future variations of Dojo can be extra tailor-made to general-purpose AI coaching. One potential drawback with that’s nearly all AI software program out there was written to work with GPUs. Utilizing Dojo to coach general-purpose AI fashions would require rewriting the software program.
That’s, until Tesla rents out its compute, much like how AWS and Azure lease out cloud computing capabilities. Musk additionally famous throughout Q2 earnings that he sees “a path to being aggressive with Nvidia with Dojo.”
A September 2023 report from Morgan Stanley predicted that Dojo might add $500 billion to Tesla’s market worth by unlocking new income streams within the type of robotaxis and software program providers.
In brief, Dojo’s chips are an insurance coverage coverage for the automaker, however one that might pay dividends.
How far alongside is Dojo?
Reuters reported final yr that Tesla started manufacturing on Dojo in July 2023, however a June 2023 post from Musk instructed that Dojo had been “on-line and working helpful duties for a couple of months.”
Across the identical time, Tesla mentioned it anticipated Dojo to be one of many prime 5 strongest supercomputers by February 2024 — a feat that has but to be publicly disclosed, leaving us uncertain that it has occurred.
The corporate additionally mentioned it expects Dojo’s complete compute to succeed in 100 exaflops in October 2024. (One exaflops is the same as 1 quintillion pc operations per second. To succeed in 100 exaflops, and assuming that one D1 can obtain 362 teraflops, Tesla would wish greater than 276,000 D1s, or round 320,500 Nvidia A100 GPUs.)
Tesla additionally pledged in January 2024 to spend $500 million to construct a Dojo supercomputer at its gigafactory in Buffalo, New York.
In Might 2024, Musk noted that the rear portion of Tesla’s Austin gigafactory can be reserved for a “tremendous dense, water-cooled supercomputer cluster.” Now we all know that it’s truly Cortex, not Dojo, that’s taking on that area in Austin.
Simply after Tesla’s second-quarter earnings name, Musk posted on X that the automaker’s AI staff is utilizing Tesla HW4 AI pc (renamed AI4), which is the {hardware} that lives on Tesla autos, within the coaching loop with Nvidia GPUs. He famous that the breakdown is roughly 90,000 Nvidia H100s plus 40,000 AI4 computer systems.
“And Dojo 1 can have roughly 8k H100-equivalent of coaching on-line by finish of yr,” he continued. “Not large, however not trivial both.”
Tesla hasn’t supplied updates as as to whether it has gotten these chips on-line and working Dojo. Through the firm’s fourth-quarter 2024 earnings name, nobody talked about Dojo. Nevertheless, Tesla mentioned it accomplished the deployment of Cortex in This fall and that it was Cortex that helped allow V13 of supervised FSD.
This story initially printed August 3, 2024, and we are going to replace it as new data develops.
