Latest / Elon Musk Podcast / Tesla Will Double Its Dojo D1 Supercomputer Chip Orders

Transcript

Tesla Will Double Its Dojo D1 Supercomputer Chip Orders · Elon Musk Podcast

• 0:01
  Hi everybody and welcome back to another episode of the Elon Musk
• 0:04
  Podcast where we discuss all things technology
• 0:07
  entrepreneurship in the future of humanity.
• 0:10
  In today's episode, we'll be focusing on the recent news
• 0:13
  regarding Tesla's plan to ramp up the orders for its Dojo D1
• 0:19
  supercomputer chips. This move is part of a larger
• 0:22
  strategy to advance Tesla's capabilities and artificial
• 0:25
  intelligence, particularly for their autonomous driving
• 0:29
  technologies. Now these chips are manufactured
• 0:31
  by Taiwan Semiconductor Manufacturing Company, or TSMC,
• 0:35
  and according to insiders, Tesla has plans to double the D1 chip
• 0:40
  orders for the coming year. Firstly, let's talk about the
• 0:43
  Dojo D1 chips and what they are and why they're so important.
• 0:47
  Tesla's D1 chips are application specific integrated Circuits or
• 0:51
  ASICS, specifically designed to enhance the capabilities of
• 0:55
  Tesla's Dojo Supercomputer. Now this machine is expected to
• 0:59
  play a critical role in the training of Tesla's driver
• 1:02
  assisted systems and their self driving AI models.
• 1:06
  And the main advantage here is that Tesla can more efficiently
• 1:09
  analyze extensive video data from its vast vehicle fleet,
• 1:13
  negating the need for manual human input.
• 1:17
  This also represents a significant stride toward
• 1:19
  Tesla's aspirations for full self driving technology as well
• 1:23
  as other initiatives like dedicated robo taxis and
• 1:27
  optimist robots. As Tesla rolls out more a I
• 1:30
  dependent projects, the Dojo D1 chip becomes an increasingly
• 1:34
  central component to the company's technological
• 1:37
  infrastructure and the Dojo supercomputer has the potential
• 1:40
  to advance not only Tesla's automotive applications, but
• 1:44
  could also be integrated into other autonomous systems in the
• 1:47
  future. Now moving on to the business of
• 1:50
  this report suggests that Tesla will be ordering 10,000 units of
• 1:55
  the Dojo D1 chip from TSMC for the upcoming year.
• 1:59
  This is not just good news for Tesla, but also positive
• 2:02
  development for TSMC, which has seen an increase in its high
• 2:06
  performance computing or HPC related orders, partially fueled
• 2:11
  by Tesla's orders. And the dynamics between Tesla
• 2:14
  and TSMC are becoming increasingly symbiotic,
• 2:18
  benefiting both companies and pushing the envelope in
• 2:21
  supercomputing and AI. Now firstly, we need to
• 2:25
  understand the implications of Tesla shift away from NVIDIA.
• 2:29
  For years, Nvidia's powerful GPUs have been the workforce for
• 2:32
  many companies in the tech industry, Tesla included.
• 2:36
  Nvidia's technology has been particularly critical in Tesla's
• 2:39
  development of machine learning algorithms for autonomous
• 2:42
  vehicles, and the strategic decision to ramp up the
• 2:45
  production of Dojo D1 chips marks a turning point for Tesla,
• 2:49
  allowing the company to have more control over its own
• 2:53
  destiny. And the Dojo D1 chips are not
• 2:56
  just a piece of hardware, they represent Tesla's determination
• 2:59
  to control its own intellectual property and their own
• 3:03
  resources. Now let's talk about the supply
• 3:05
  chain. Here in the global semiconductor
• 3:08
  market, supply constraints have become a significant concern for
• 3:11
  tech companies and the recent chip shortages have demonstrated
• 3:14
  the vulnerabilities and relying too heavily on 3rd party
• 3:17
  vendors. By investing in its own chips,
• 3:20
  Tesla is essentially laying the foundation for a more resilient
• 3:24
  and adaptable supply chain. This in house production
• 3:27
  capability gives Tesla an edge, enabling faster integrations of
• 3:32
  its technology and potentially leading to a quicker rollout of
• 3:35
  full self driving capabilities. Now on the future projection
• 3:40
  side, Tesla has ambitious plans for the Dojo supercomputer which
• 3:44
  serves as the brain Center for their Dojo D1 chips.
• 3:47
  According to various reports, the next Gen.
• 3:49
  Dojo system is expected to be operational by 2024 with a
• 3:54
  target achieving 100 exaflops of compute power.
• 3:58
  This would not only make Dojo one of the most powerful
• 4:00
  supercomputers on the planet, but also give Tesla the
• 4:03
  computational capabilities required for sophisticated
• 4:07
  machine learning algorithms and data analysis.
• 4:10
  Now with a more powerful Dojo, Tesla's rate of innovation could
• 4:13
  accelerate, impacting everything from autonomous vehicles to
• 4:17
  smarter energy solutions. Now the financial implications
• 4:21
  should not be overlooked either. This price tag associated with
• 4:24
  NVIDIA H100GP U's was reported to be around 300,000,000 for the
• 4:30
  setup that Tesla was considering.
• 4:32
  Now producing the Dojo D1 chips in house, despite the initial
• 4:35
  research and development costs, could end up being a more cost
• 4:38
  effective solution for Tesla in the long run.
• 4:41
  In addition, having a proprietary chipset could become
• 4:45
  a new revenue stream for Tesla as they could eventually license
• 4:48
  this technology to other companies and sectors like
• 4:51
  robotics, healthcare, or even in aerospace.
• 4:55
  Now let's not forget that the Dojo supercomputer isn't
• 4:58
  exclusively for autonomous driving applications.
• 5:01
  Test the signals that the Dojo capabilities could be leveraged
• 5:05
  in other areas like robotics and different kinds of autonomous
• 5:07
  systems. Tesla's hardware 4.0 self
• 5:10
  driving computer, another collaborative between test and
• 5:14
  TSMC, has been a significant component in the development of
• 5:18
  autonomous technology. And with such diverse
• 5:21
  applications, Dojo's future could very well make it a
• 5:24
  standard tool in the computational universe, applied
• 5:28
  in various sectors, requiring immense computational power and
• 5:31
  machine learning capabilities. Tesla CEO Elon Musk has even
• 5:35
  disclosed plans to invest over $1 billion in the Dojo project
• 5:39
  by the end of 2024, underlining its critical role in Tesla's
• 5:44
  future and potentially the future of a I and robotics.
• 5:49
  Now, we can't have an Elon Musk podcast episode without touching
• 5:52
  on a little bit of controversy. Tesla is currently under federal
• 5:55
  investigation for allegedly misappropriating company funds
• 5:59
  to build a new mansion for Elon Musk.
• 6:01
  And while this is indeed a critical issue that could affect
• 6:03
  Tesla's public image and even stock prices, it's essential to
• 6:07
  remember that this is still under investigation and nothing
• 6:10
  has been proven yet. Regardless of these challenges,
• 6:13
  the momentum behind Tesla's technological advancements like
• 6:16
  Dojo and the One Chips doesn't seem to be slowing down.
• 6:20
  If anything, the accelerated investments indicate an
• 6:23
  unshakable belief in the company's vision into the future
• 6:26
  of autonomous systems and AI. Now, thank you so much for
• 6:30
  tuning into the Elon Musk podcast today.
• 6:33
  We found this episode informative and want to keep up
• 6:35
  with all things Elon Musk and all of his companies do.
• 6:38
  Please hit the subscribe button or the follow, but on your
• 6:41
  preferred podcast platform. Remember, the subscribing's
• 6:44
  Free, helps us out and takes just a second.
• 6:46
  And until next time, this has been the Elon Musk podcast.
• 6:49
  My name is Will Weldon. Take care of yourselves and each
• 6:52
  other, and I'll see you in the next one.