Hey everybody Rob Maori here and we should have a pretty exciting episode we've got a lot to go through including developments on Tesla's AI efforts, a possible stealth acquisition from Tesla, news on Tesla in India and more. Less exciting with the stock, Tesla down 5.5% to take closing at $259.46, the NASDAQ with a referee down 1.2% seemed to be the result of some testimony from Fed Chair Jerome Powell, but we've got too much else to talk about so we're going to skip over that macro stuff.
Anyway, before we get into the AI updates, a poignant update by Barclays today, they actually downgraded Tesla, perhaps contributing to the weakness today. Dan Levy saying in a note that the recent rally in Tesla has potentially ignored fundamentals, even though the PE is not super crazy, obviously 80s high, but for a growth company not too bad. But said quote, we believe the stock's recent rally can be best explained by the market's current AI driven thematic trade as well as excitement over recent announcements to open the Tesla supercharger network to other brands, yet while we aren't surprised that the stock has participated in the rally, we believe it is prudent to move to the sidelines.
So interesting because we've got a lot of updates on Tesla AI today, Levy said that he believes AI is a long dated opportunity for Tesla, not something that can be counted on right now, to considerably boost its valuation. So I would agree, I think Tesla's participating in the thematic rally that we're seeing here. We've talked about that also, I think, aided by the supercharger network as well as you noted there. The long dated opportunity part of this I think is debatable because Tesla's already generating profits off of FSD beta today.
Obviously, the take rate is not super high, but it does still impact Tesla's margins and earnings already. And presumably that should only improve as Tesla improves the system. So that's what we're looking at today.
So Tesla started a new account, Tesla AI on Twitter, which is nice. We can see the little organization verification there. And they tweeted this graphic as a part of a thread, which I would recommend just going reviewing this whole entire thread because they've given a lot of interesting examples of things that they're working on from an FSD perspective. But what I want to focus on today is this graphic.
A couple of really interesting things here, obviously the growth in Tesla's compute. This is what we're looking at here. They're putting this in units of A100 GPUs. And the first thing I would call attention to is that in July 2023, so next month, Tesla's indicating that as the start of dojo production. So this has been the working for years now in the works for years now. And obviously Tesla's been developing this producing, you know, sort of test or beta versions of the dojo product.
And it looks like now they're ready to actually take that into volume production as we can see with the growth and compute that they're projecting. So this is a very exciting update in terms of dojo. Elon has been pretty vocal about this being a little bit of a long shot bet for Tesla. And he wasn't really sure if it was going to work out. It seems like they are becoming more confident in it.
Tim Simon of Tesla, you know, the lead of Tesla's infrastructure team said this is real. This is happening. It's going up X speed and cannot be stopped for context. Today, our compute clusters have 0.3% idle time and 84% of jobs are high priority. We want to be in a place where we have an excess of compute.
特斯拉基础设施团队的负责人 Tim Simon 表示,这是真实存在的。正在发生的事情,它出现了 X 速度,并且无法停止。为了更好的理解,今天我们的计算集群只有 0.3% 的空闲时间,而 84% 的工作都是高优先级。我们希望拥有足够的计算能力供我们使用。
So as Tesla's fleet grows, as they are more well set up for things like auto labeling as they integrate portions of the planning system and potentially all the planning system at some point into running on a neural network. Obviously their training needs are just going to continue to grow. And the more compute that they have, the more they can train, which is clearly constrained if right now they're running at 99.7% uptime on the current compute.
I want to take a little bit closer if I look at this. Number one, very exciting about dojo. Number two, just look at, I think look at this graph. I had to bust out the pixel Photoshop on this again so we could get a sense of where things are. So if we look at this, of course, each of these segments is 69 pixels. This first starting point in January 2023 is seven pixels, which means you're talking about 10% of the 100,000 label that we see here.
So essentially you've got 10,000 equivalent A100 GPUs. If we use the 100 XFLOP scale over here, that computes to roughly three XFLOPs of compute that Tesla's running on at the beginning of this year. Now moving into June and of June where we're at right now, that is about 12 pixels. So even though it's a very minor difference on the scale of this graph, that actually represents a 70% increase in just six months in Tesla's compute capability up to about 17,000 equivalent A100 GPUs or about five and a half XFLOPs.
Tremendous growth just in the last six months in Tesla's capabilities here, but you can hardly even register it on where they expect to be going over the next 12 to 18 months.
过去六个月特斯拉的能力有了巨大的增长,但在他们未来12至18个月的发展计划中,这几乎无法引起关注。
So if we look a little bit further out in October, so towards the end of Q3, that's up to 28 pixels.
So even more than double their compute capabilities.
所以,它们的计算能力甚至超过了双倍。
简而言之:它们的计算能力是双倍以上。
I think we've already seen progress in the iteration rate from Tesla in terms of progress and FSD beta.
我认为从特斯拉的进步和FSD beta方面来看,我们已经看到了迭代速度的进展。
And as they, you know, for X compute from where they were at earlier this year, presumably that should add to the rate of progress that we are seeing so far.
So extremely exciting time right now for Tesla as this comes online.
对于特斯拉来说,现在是非常激动人心的时刻,因为这一切都正在线上进行。
Obviously, they're projecting to be top five in the entire world in terms of companies that have compute capabilities in of this, you know, order of magnitude.
显然,他们预计将成为具有计算能力的公司在全球排名前五,而且这一级别是相当高的。
And then expecting that to continue to grow very rapidly.
然后期望它会继续快速增长。意思是指对于某种事物,期待它在未来会保持高速增长的趋势。
We don't need the pixel math anymore since we actually get on the scale here.
Looks like they would be, you know, roughly three Xing again next year, even in the first three quarters.
看起来明年前三个季度,他们的大致增长率可能会再次达到三倍。
So that growth continues.
为了继续保持增长。
And what I think is, you know, potentially even exciting past this graph, which is already extremely exciting, where you're going from, you know, three to a hundred again in 18 months.
If that sort of rate of progress continues, the graph just starts to look pretty ridiculous, at least on a linear scale like this.
如果这种进展速度持续下去,图形将变得非常荒谬,至少在这样的线性比例尺下。
Obviously, a logarithmic scale would look a little bit more reasonable.
显然,对数刻度会看起来更合理一些。
But that's kind of what, you know, like a curve here fits extremely well on Tesla's projections, and they could be looking at, you know, getting up to 300 extra flops or so sometime towards the middle of 2025.
And he said that, you know, right now we've got one of the biggest neural network training systems in the world.
他说,现在我们拥有世界上最大的神经网络训练系统之一。
And we expect to increase that capability by an order of magnitude by the end of this year.
我们预计到今年年底将把这一能力提升一个数量级。这意味着我们将大幅提高我们的能力水平。
And remember, this was earlier this year, so end of 2023, he's talking and then says, and probably another order of order of magnitude by the end of next year with some combination of Nvidia and dojo.
And then if we go to the end of 2024, this graph would not even be, you know, technically a full order of magnitude, depending on how you're looking at that, defining that by the end of 2024.
So Elon's comment there would actually be a little bit ahead of this pace, which no surprise.
埃隆在那里的评论实际上会比这个进度快一点,这不出所料。
But at least if you're talking another, you know, 10 times increase from October 2024, that's where you get up to this sort of million equivalent a 100 GPUs, which maybe could happen sometime in 2025.
But maybe they find ways to even leverage this better, and that could make it even a faster curve.
也许他们会找到更好的方法来充分利用这一点,这可能会使曲线变得更快。
So these are the kind of things that Elon is seeing and thinking about and probably why he's consistently been a little bit wrong on the timelines for FSD beta because, you know, he tries to think in this exponential way that a lot of people are mostly thinking when you're if that so I'm pumped about it.
Tesla's pumped about another Tesla team member here saying Tesla already build some of the largest GPU clusters in the world. Now we're going to do greater than 10 X that next year, dojo's compute production rate will be on the same order as Nvidia's. So again, going back to Tim's omen, this is real. This is happening. It sounds like they've reached a point with dojo where they feel like, you know, it's it's ready to roll. So very exciting.
Elon said I should clarify that dojo V one is highly optimized for vast amounts of video training versus general purpose AI dojo V two will address these limitations. So it's important context when we look at things like this, you know, not all flops are necessarily created equal in terms of what they're meant to do, but still for what Tesla's meant to do. Obviously, these are going to be very valuable, you know, units of compute.
埃隆说我应该澄清道:Dojo V one非常优化大量视频培训,而不是通用AI,Dojo V two将解决这些限制。因此,当我们看待这些事情时,重要的是要有这个背景,你知道,并不是所有Flops在其意图上都是平等的,但对特斯拉要做的事情来说,显然,这些将是非常有价值的计算单元。
And then in terms of whether this would be competitive with Nvidia, he said not exactly a competitor, but we aren't as we aren't selling dojo computers, although it could be sold as a service, which Elon has talked about before potentially being somewhat comparable to AWS. And continue to say we just need a crazy amount of training compute to get full self driving 10 X safe within a human, but can't get enough GPUs plus prices are extremely high. So in addition to these comments about orders of magnitude increases in training capability, you want us also talked about an order of magnitude decrease in cost for that training. So obviously dojo playing a big role on the cost side as well.
So I think just thinking about all of this, you know, when we think about FSD and autonomy and the competitive set, and we see these updates from crews and Waymo about slightly increasing their service areas and things like that, not to knock them. But when you look at the kind of things that Tesla is doing, it's just hard to imagine, especially for a company like Cruz, how they would necessarily be competitive with sort of the back end that Tesla is running with their training models. So pretty interesting.
I guess probably to wrap up maybe not to throw a ton of cold water on all of this, but just as a reminder, this was Elon on the Q1 call. So back in April, only a couple of months ago, he said, I look at dojo as kind of a long term bet or sort of as kind of a long shot bet. But if it's a long shot bet that pays off, it will pay off in a very, very big way. But potentially, yeah, potentially in a very, very big way, like in the hundred in the multi hundred billion dollar level, but I think it's still put in the long shot category. So again, two months ago, it does seem like from these comments that this is maybe a little bit of outdated in terms of the comments and maybe they had some sort of breakthrough. But probably wise to keep that in mind. And especially when we look at these longer term projections, you got to remember like Tesla's still right here, right?
Zoom in a little bit there. That's what has actually been accomplished thus far and the rest would be, you know, a forecast. So we'll see. Hopefully things progress in this way. And Tesla seems to feel confident in it, which is exciting to see.
All right, so moving on from that topic, we do have an update on Tesla and India as we had talked about Elon met with the prime minister of India yesterday. And he had quite a few comments after that meeting to the media. Let me just find my notes here.
Probably the most impactful one regarding Tesla was that quote talking about the prime minister quote, he really cares about India because he's pushing us to make significant investment in India, which is something that we intend to do, just trying to figure out the right timing and quote. So probably not too surprising given all of the, you know, rumors that we have had about Tesla and their conversations with India.
He did say that the prime minister did extend an invitation to Elon to visit India and Elon is tentatively planning to visit again next year. So I don't know if that gives us any sort of hint on, you know, the right timing that you want to first do there, but it would be another touch point. And then he did also mention that he thinks there's tremendous possibility for sort of all three pillars of sustainable energy in India, solar energy storage and EVs. And then a little bit separately from Tesla also mentioned that hopefully looking forward to bringing Starlink to India as well.
So in general, he said the conversations went really great. He spoke very highly of the prime minister. And I think, you know, it gives us a lot more confidence that that relationship is developing positively, you know, over the last few months.
Next we've got kind of an interesting possible stealth acquisition. This is being reported, I believe this is coming out of Germany. Check to translate here. Yeah, out of Germany. So this is reported from business insider in Germany. They say that Tesla is said to be interested in buying a wifurian, which I think is probably how you pronounce it, which is an EV slash, maybe not EV, but like a vehicle and I guess EVs, depending on how you classify an EV, EV and sort of production, you know, robots and things like that, wireless charging.
So apparently, the shareholders for wifurian have indicated their intent to sell their shares in the company to Tesla International BV by means of a purchase agreement, which is Tesla's entity in Europe. And let's see. Yeah. So developing an inductive charging system for EVs and robots. And it looks like according to previous rounds raised by this company, Tesla probably has paid somewhere in mid to high double digits for this startup if this deal is eventually confirmed. So we may not hear any confirmation for a while. Sometimes these things are a little bit hard to get information on outside of these sort of reports, but an interesting acquisition here potentially for Tesla.
Again, this is the kind of, you know, machine that you can see that they would be charging outside of potentially be, you know, cars, which Tesla, many other automakers have machines like this. You know, if you ever do a tour of Giga, Texas, you're going to see plenty of machines like this out there, you know, transporting things from, from one place to another. So in addition to that, we've also seen Tesla tease a little bit at investor day as well. A little bit of, you know, wife wireless charging there, potentially on this vehicle and the garage.
So, you know, they can't forget to cool, can't forget to do cool stuff and teasing potentially wireless charging, making this acquisition of a wireless charging company. This definitely seems to be on Tesla's roadmap in some way, shape or form. Again, probably with this acquisition in a major way in factories, but then also here potentially for vehicles, which would probably be maybe not as focused on like a supercharging, although that could be very beneficial for autonomous driving. But Tesla may be intentionally here putting the vehicle in a garage, maybe a little bit more focused on destination type of charging or home charging situations that from a little bit of that I've read on my theory, and it does seem to be not necessarily like the super highest power like a supercharger would be, but relatively high enough power to, you know, charge in the factory or charge at home. So it will be interesting to see if we get any more information on that potentially not, but you know, Tesla continues to do things, you know, in this wireless charging space.
All right, on supercharging, we've got an update on scroll here, update on the NACS standard and Texas has commented and said that for any federal funding going to charging stations that are installed in Texas, they'll be requiring both CCS and NACS to be a part of those charging systems to be eligible to receive funding from the federal funding.
So pretty big win, I think for Tesla in Texas and I think a recognition of the progress that NACS is making, they did comment to Reuters that quote, the decision by Ford GM and now Rivian to adopt NACS change requirements for phase one of the rollout. I guess that's the end of the quote, but saying that they would require fast chargers to have one CCS and one NACS connector. So it'll be interesting to see again, if Tesla continues to actually gain traction on NACS, if there really ends up being much of a point to CCS, I think, I guess, until you really have everyone on board, you probably still have to offer support for CCS.
But I mean, Tesla's already what 60, 70% of the EV market in the US. And if they can get a significant chunk of that other 30%, as those other brands, as we've talked about my grow market share a little bit faster since they're coming from a smaller base, if you get that to an 80 or 90% number, all of a sudden it starts to look pretty silly to incentivize having CCS systems on these on this infrastructure. So we'll see, but definitely a big win for Tesla there and for the NACS in Texas.
Then a couple of other updates out of Texas. First on the Gigafactory Gigataxis, Joe Tegmeyer has spotted that Tesla has now started to tear down some of the walls in preparation for their expansion, which we had talked about the plans that they had submitted for that previously.
So nice to see that work begin, although I'm sure it's already begun internally, but nice to be able to see it externally. And well, I'm sure Joe will keep us updated on that.
And then the other thing going on in Texas is there's been apparently a heat wave, nothing new there, but apparently Tesla energy, Tesla electric plans that are integrated with power walls have been able to sell power back to the grid during periods of high demand at up to about $5 per kilowatt hour after potentially generating them or storing them from a solar system or storing from the grid at a previous time at something maybe around 10 cents a kilowatt hour.
So although obviously this is for peak periods and it's not necessarily going to sustain for a super long time where you can generate a ton of revenue, still pretty interesting to see. And also the price I think reflects how important that sort of an infrastructure is and how that can help the grid in areas like this where they might experience challenges over time.
So kind of cool to see, you know, a few examples posted on Twitter of that.
真的很酷,你知道,能在 Twitter 上看到一些例子。
And then the last couple of things for today, we have an updated American made index for 2023 from cars.com. We talked last year about how Tesla took the number one and number two slots with the model Y and the model three for most American made vehicles. This year they've taken the top four slots.
So the model Y, the model three, the model X and model S from all cars that are eligible in this index, which is not all cars, but all cars that are eligible. Tesla has taken the top four slots with their four vehicles that they sell.
So the S and the X didn't quite make the list last year because they were still my understanding of what cars.com said was that because they were still sort of getting production up and running from the refresh, that there were some things that made them not eligible. So at any rate this year, they say that they have become eligible again. And now they pop back up to the top.
So they say that a substantial boost in your, your employment at, at Gigatexas has aided at all Teslas in their workforce calculations, like going to their rankings or the index and lifted the S and the X, which I guess lagged last year.
So nice to see this. Again, something that Tesla does not get enough credit for or probably enough awareness of. And you know, depending on what Tesla's thinking here for advertising, could be something to play up a little bit more, a little bit more frequently.
So really nice to see that consistent ranking there and the next rejoining.
真的很高兴看到那持续的排名以及接下来的重回。
And then finally, quick update on Ribian. So they had a meeting or investor day with Deutsche Bank last week. And as part of that, their CFO commented on the R2, which would be their next generation vehicle after the R1s and the R1T, of course, the delivery van.
最后,对于 Ribian 的最新情况简要更新一下。他们上周与 Deutsche Bank 进行了会面或投资者日。在其中,他们的首席财务官评论了 R2,这将是 Ribian 推出的下一代车型,继 R1s 和 R1T(当然还有送货车)之后。
Apparently, that's going to be launched in 2026. I think that timeline is what they had previously stated. Originally, it was targeted a little bit earlier, but they had previously postponed that. So nothing new on that.
So it'll be very interesting to see that vehicle. It's obviously a long ways away still. Here's a little bit of a render from car buzz. So not an official vehicle. It sounds like they're going to be unveiling it next year or rebealing it next year.
So we'll keep an eye out for that. But if it does look like something like this, I think they can find a lot of success there, again, focusing a little bit more on that adventure type of category.
All right. I think that is it for today, then. Again, very exciting episode here. And I can't wait to hear more from Tesla in a few weeks on their progress. I'm sure they'll talk more about it on the Q2 call. But that'll wrap it up for today.
So as always, thank you for listening. Make sure you're subscribed and signed up for notifications. You can also find me on Twitter at Tesla podcast. And we'll see you tomorrow for the Thursday, June 22 episode of Tesla Daily. Thank you.