Tesla Q2: What's Being Overlooked / FSD 12.5 Reviews / Elon's Mistake ⚡️

Welcome to Electrified, it's your host Dylan Loomis, quick shout out to my newest patron David H. Thank you for choosing to support the channel. Rather than doing my typical post earnings video, I think this one will be more valuable to hop around a little bit and touch on some things that I think are going overlooked or not talked about at all and then clarifying some other things. But before we even get into it, I think it's important to reiterate Elon's sentiment given the valuation where it is. If you don't think Tesla is going to solve autonomy, you should not be in Tesla stock.

欢迎来到 Electrified，我是你的主持人 Dylan Loomis。首先向我最新的赞助者 David H 致以诚挚的感谢，感谢你选择支持这个频道。这次我不会像往常那样做一个收益报告视频，我认为今天的视频会更有价值，因为我会谈到一些被忽视或根本没有被谈论的事情，并澄清一些其他问题。但在我们开始之前，我觉得有必要重申一下埃隆的观点，考虑到特斯拉目前的估值。如果你不认为特斯拉能够解决自动驾驶问题，你就不应该持有特斯拉的股票。

One thing that I think is not being celebrated nearly enough is that Tesla said by quarter four of this year they're on track for dry cathode production for the 4680s, which means if that goes well, Tesla's 4680s will then be significantly cheaper than any alternative in the market. They already have a validation cyber truck with the dry cathode that was made on Tesla's mass 4680 manufacturing equipment that most of which Tesla has designed in house and some of which was actually built in house Tesla acquired Maxwell's DBE tech back in May 2019.

有一件事我认为没有被足够庆祝，那就是特斯拉表示他们预计在今年第四季度实现4680电池的干电极生产。如果一切顺利，特斯拉的4680电池将比市场上的任何替代品都便宜得多。他们已经使用干电极生产了一个验证性的Cybertruck电池，这个电池是在特斯拉的大规模4680制造设备上生产的，这些设备大部分由特斯拉自行设计，部分甚至在特斯拉内部制造。特斯拉在2019年5月收购了Maxwell的干电极技术。

So Tesla has been working on solving this problem for the last five years. And it's this solution that if they can scale it will actually remove all of that wet process, remove all of the drying and the huge machines and the time that goes with it. And that should dramatically accelerate both 4680 production and the cost of goods sold for 4680s coming down aggressively.

所以，特斯拉在过去五年里一直致力于解决这个问题。如果他们能够将这个解决方案大规模应用，它将彻底取消所有湿法工艺，消除所有干燥过程、大型机器以及相关的时间。这将极大地加速4680电池的生产，并且使4680电池的销售成本大幅下降。

Now we likely won't see those benefits hit Tesla's margins until mid 2025 and into 2026. But people were talking about Tesla abandoning the 4680 program all together by the end of this year. And the way I see it this confirmation from Tesla should totally take that off the table, even though personally I never thought it was on the table to begin with. This means Tesla's original goal with battery day to make batteries in house with new technologies that will be cheaper than any alternative from suppliers and with better energy density is now on track and likely to be solved by the end of this year and into next year. It's not a done deal yet, but when it comes to Cybertruck and other vehicle profitability, whatever vehicles Tesla will ultimately use 4680s in, this is a major win for the company for the next three to five years and beyond.

我们可能要到2025年中期甚至2026年才能看到这些好处反映在特斯拉的利润率上。但人们曾预测特斯拉会在今年年底前彻底放弃4680电池项目。而在我看来，特斯拉的确认应该完全排除了这种可能性，尽管我个人从一开始就不认为这会成真。这意味着特斯拉在电池日提出的通过新技术自行制造电池、成本低于供应商并具有更高能量密度的初衷，现在正在按计划推进，并有望在今年年底到明年实现。这虽然还没有完全落实，但对于Cybertruck以及其他需要使用4680电池的车辆来说，未来三到五年甚至更久，这都是特斯拉的一大胜利。

Let's shift focus to autonomy and robotaxies because that really is the main driver for Tesla stock at least in the near term. People out there are mistakenly saying, see, I told you that hardware three will never be fully autonomous. I don't believe that's true. Elon actually said that hardware three would run the same parameter count, meaning the five X increase that we're seeing in 12.5, but it requires extra work to optimize the code for hardware three, not that it's never going to work on hardware three, just that Tesla needs a bit more time for that. When it comes to 12.5 and the early reviews overwhelmingly, they are very positive and more positive than it was with the 12.4 releases. That's not to say this version is perfect and that FSD is all of the sudden solved. However, it does seem to be a solid upgrade from 12.3.6 and 12.4.

让我们把重点转向自主驾驶和自动驾驶出租车，因为这才是特斯拉股票在近期的主要推动力。有些人误以为硬件三永远不能实现完全自主驾驶，但我不认同这个观点。埃隆实际上说过，硬件三将运行相同的参数数量，也就是我们在12.5版本中看到的五倍增加，但这需要对硬件三的代码进行额外优化工作，并不是说硬件三永远无法实现，只是说特斯拉需要更多时间来完成这项工作。关于12.5版本的早期评测，整体反馈非常积极，比12.4版本的反馈更为正面。但这并不意味着这个版本是完美的，也不代表全自动驾驶已经完全解决。然而，相比12.3.6和12.4版本，12.5版本确实是一个坚实的升级。

Some lesser known FSD users who were just saying that Tesla's FSD is effectively terrible are now saying 12.5 is by far the best it has ever been. AI driver has said 12.5 is a step change in improvement from both 12.3 and 12.4, saying the quickness and decision making on this version is on another level. A shock said there was a huge focus on improving both safety and smoothness in this release. One of my personal tests is to not spill an open cup of coffee while on FSD. 12.5 was the first version where I was able to do that for 30 minute drives.

一些不太为人知的FSD用户曾表示特斯拉的FSD（全自动驾驶软件）非常糟糕，但现在他们纷纷表示12.5版本是迄今为止最好的版本。AI驾驶员认为12.5版本相对于12.3和12.4有了质的飞跃，决策速度和准确性都达到了一个新的高度。一位用户提到，这次的更新特别注重提高安全性和平顺性。我个人的一个测试方法是看在FSD模式下汽车能否在不洒出咖啡的情况下行驶。12.5是第一个让我在30分钟的驾驶中成功完成这个测试的版本。

It's still not available on the Cybertruck, looking like that will come with a future dot release. Dirty Tesla said I'll say this is the safest, the best version of FSD ever and should go to everyone with FSD worldwide. And on Chris's 45 minute ride to work, he said it was by far the most impressive it's ever done that route. As I said though, from Chris, it's definitely not perfect as this clip shows it driving right into a do not enter. Chris also mentioned that it was still struggling with blinking red lights where it would just come to a complete stop and then stay there. And AI driver had a clip of 12.5 not slamming on the brakes at a yellow light, which I'll be honest personally, I was hoping to see.

这功能在Cybertruck上还是不可用，看起来会在未来的小版本更新中推出。Dirty Tesla表示，这是有史以来最安全、最好的全自动驾驶版本，应该推广给全球所有拥有全自动驾驶功能的用户。而Chris在45分钟的上班途中提到，这是这条路线以来表现最为出色的一次。然而，正如Chris所说，它绝对不是完美的，因为这段视频显示它驶入了一个禁止入内的区域。Chris还提到它在应对红灯闪烁时仍然有问题，会完全停下来然后停在那里。AI司机有一段12.5版本在黄灯时没有急刹的短片，老实说，我本质上是希望看到这一点的。

It's early, but at this point barring anything unexpected, it feels like 12.5 will be the release one of these dot releases that does make it to a wide release. One of these years, Elon's prediction about robot taxi timing will be right, but we really can't put much stock in his guesses at this stage. What we can do however, is look at what Tesla is doing with compute. This is where we are right now. Let's just call it sitting around 38,000 H100 equivalents. Tesla's plans for the end of this year, they're expected to be up to around 89,000 H100 equivalents. The math on that is a 2.3 X in compute capacity over just the next five months. I'm certainly still expecting improvements this year as Tesla really learns how to actually train with all of this data and customize it in the way that they want. But once the South Extension at Gigatexus is filled up with compute and it's turned on and begins training 2025, maybe the year where the real magic happens. By that time, Tesla will have multiples of the compute power it has now, plus roughly a year under its belt of learning how to train this end to end network.

这还早，但目前来看，如果没有什么意外，12.5将成为广泛发布的一个版本。虽然埃隆关于机器人出租车的预测有一年会准确，但在这个阶段，我们不能太相信他的猜测。我们能做的，是关注特斯拉在计算能力方面的进展。目前，特斯拉大约拥有相当于38,000个H100的计算能力。根据特斯拉的计划，到今年年底，这个数字预计将增加到大约89,000个H100。这意味着在未来五个月内，计算能力将增加2.3倍。我当然仍然预计今年会有改进，因为特斯拉正在学习如何利用所有这些数据进行训练，并按照自己的需要进行定制。但一旦Gigatexas的南扩展部分装满计算设备并开始训练，2025年可能会是奇迹发生的一年。到那时，特斯拉将拥有现在多倍的计算能力，并且在整个网络训练方面已有大约一年的经验。

And to those of you in the comments that have been saying Tesla doesn't really train the system, it's just the data in and then controls out. Well, it's not that simple. Tony a Tesla engineer just said a lot of hard work in many late nights went into making FSD 12.5. Many ideas were simplified and reworked from first principles. Plus Tesla is in control of the actual video footage that makes it into the system for training. Another example is when Elon was talking about 12.4, he said part of the issue was too much training on interventions and not enough on normal driving. So this is the type of thing I'm referring to when I say that Tesla is learning how to train this new end to end neural net on Tesla's compute build out Elon said it's important to note we also use the Tesla AI for computer in the training loop within video GPUs currently at roughly a one to two ratio. By the way, hardware four is now being called AI for just like hardware five is now AI five. Elon said that means around 90,000 H 100 plus around 40,000 AI for computers.

在评论区里有些人说，特斯拉并没有真正训练系统，只是把数据输入然后输出控制。其实事情并没有那么简单。特斯拉的一名工程师托尼刚刚说到，FSD 12.5的开发过程中花费了大量深夜的辛勤工作。许多想法被简化并从基本原理重新设计。此外，特斯拉还掌控着用于训练系统的实际视频数据。 另一个例子是，当埃隆·马斯克谈论12.4版本时，他提到问题的一部分是对干预操作的训练过多，而对正常驾驶的训练不足。这就是我所指的，特斯拉正在学习如何在其计算架构上训练这种新的端到端神经网络。埃隆还说，很重要的一点是，我们在训练循环中使用了特斯拉AI计算机，结合视频GPU，目前大约是1比2的比例。 顺便说一下，现在硬件版本四被称为AI4，就像硬件版本五现在被称为AI5一样。埃隆提到，这意味着大约有90000个H100芯片加上约40000个AI4计算机。

Thus, it seems like this chart we just had on the screen would actually not include those 40,000 AI for computers. Elon said Dojo one will have roughly 8,000 H 100 equivalent of training online by end of year, not massive but not trivial either. So it sounds like looking at that chart any H 100 equivalents would actually be Dojo and then the AI for may not be included in that chart at all. But those details are not as important as knowing that heading into 2025 Tesla's compute will be on an entirely different level and that's both for FSD and Optimus. That leads us to the Tesla and X AI conversation. Elon's poll about if Tesla should invest $5 billion into X AI currently sits at 68% for 32% against this poll has no bearing on anything Tesla would still need board approval and a shareholder vote. This is just testing the waters. Now this leads to a potential mistake that I think Elon made on the call. We've been told Tesla is not compute constrained right now but Elon said that he was quite concerned about getting state of the art in video GPUs when they need them.

因此，看起来我们刚刚屏幕上显示的图表实际上并没有包括那4万台用于计算机的AI。Elon说Dojo One到今年年底大概会有约8,000个相当于H100的在线训练能力，不算巨大但也不容小觑。所以，从那个图表来看，任何相当于H100的设备其实都是指Dojo，而那些AI可能根本没有被包括进图表里。但这些细节并不如知道进入2025年特斯拉的计算能力将达到一个全新水平重要，这不仅适用于全自动驾驶(FSD)，也适用于Optimus。这就引出了特斯拉和X AI的讨论。Elon关于特斯拉是否应该投资50亿美元到X AI的投票目前是68%的人支持，32%的人反对。这次投票对任何事情都没有直接影响，特斯拉仍然需要董事会批准和股东投票。这只是试水行为。现在，这就导致了我认为Elon在电话会议中可能犯的一个错误。我得知目前特斯拉并不受计算能力的限制，但Elon说他对在需要的时候获得最先进的NVIDIA GPU非常担忧。

Sadly I can almost guarantee you there's a nefarious Tesla investor out there somewhere thinking wait a second X AI just bought 100,000 H 100's and that's 100,000 that Tesla now doesn't get. Now please do not misunderstand me I am not trying to argue that that complaint is somehow justified. But if Tesla gets to a place in the next two years where they are again compute constrained I can guarantee you these conflict of interest arguments between Tesla and X AI get even louder. Thanks to that Nvidia demand Elon did say they're doubling down on Dojo to not be so reliant on that but he also said Tesla really needs to make Dojo work which implies that it's not working at least at the scale they had wanted it to. He said they'll figure it out and they still see a path to be competitive with Nvidia but at least right now we would have been hoping for an update to hear that Dojo was actually accounting for a larger percentage of Tesla's training compute. But that's not what we heard after the last Dojo update which was that it was actually online and doing some training.

很遗憾，我几乎可以肯定，某个心怀不轨的特斯拉投资者正想着：“等一下，X AI刚买了10万个H100，这意味着特斯拉现在少卖了10万个。”请不要误解我的意思，我并不是在为这种抱怨辩护。但如果特斯拉在未来两年内再次面临计算能力受限的问题，我可以保证特斯拉和X AI之间的利益冲突讨论会变得更加激烈。由于英伟达需求高涨，埃隆确实表示他们会加大对Dojo的投资，以减少对英伟达的依赖，但他也说特斯拉真的需要让Dojo发挥作用，这暗示着Dojo至少在他们期望的规模上还没有达到预期。他说他们会解决这个问题，并且仍然看到与英伟达竞争的路径，但至少目前，我们希望听到的是Dojo在特斯拉的训练计算中所占的比例更大一些。但在最后一次Dojo更新后，我们听到的却是它已经上线并进行了一些训练。

When it comes to Tesla investing in X AI personally I would be on board I wouldn't be overly excited about it if you just run some simple numbers let's just say Tesla would end up with between 15 and 20% of X AI which after the last round based on a 5 billion dollar investment is where Tesla would land even if X AI grew to be a 300 billion company that's still only 60 billion dollars in valuation for Tesla. Now from an ROI standpoint I think that's just fine no one would be complaining about that but to think that it's going to transform Tesla into a trillion dollar company in my opinion with a 5 billion dollar investment that's really not in the range of outcomes for the next few years. But I also disagree with people like Gene Munster who said that he thinks Tesla should invest 20 billion dollars to make it more worthy Tesla investor while but that seems crazy to me because Tesla has plenty of AI endeavors and planned CapEx in-house that they should be spending money on before they invest that type of cash with X AI. You guys know I'm in the camp that X AI and Tesla are not really direct competitors they may be competing for certain resources like H100s but in LLM like Grock is definitely not the same thing as real-world AI like Tesla is working on and as I've said all along I think long term Tesla and X AI will be very mutually beneficial to each other.

当谈到特斯拉投资于X AI时，我个人是支持的，但我不会过度兴奋。如果你只做一些简单的计算，假设特斯拉最终持有X AI的15%到20%的股份，在最近一轮融资中，这相当于50亿美元的投资，即使X AI成长为一家市值3000亿美元的公司，这对于特斯拉来说也只有600亿美元的估值。从投资回报率的角度来看，我认为这没问题，没有人会对此抱怨。但要认为这会让特斯拉变成一家市值万亿美元的公司，在我看来，5亿美元的投资在未来几年内并不会产生这种效果。 不过，我也不同意像吉恩·芒斯特这样的人所说的特斯拉应该投资200亿美元来让它更有价值的观点，因为这对我来说很疯狂。特斯拉有许多内部的AI项目和计划中的资本支出，他们应该在这些领域投入资金，而不是把这么多现金用在X AI上。你们知道，我一直认为X AI和特斯拉并不是直接竞争对手，他们可能会在某些资源（如H100）上竞争，但像Grock这样的自然语言模型与特斯拉正在研究的现实世界AI并不一样。正如我一直所说的，我认为从长远来看，特斯拉和X AI会对彼此非常有利。

Elon confirmed that much on the call saying that Tesla has already learned a lot from X AI again though I do think this would be a Tesla investment that investors should be going into with eyes wide open because I do think it at least opens the door a bit more for new attack vectors for Tesla Elon and X AI from a legal standpoint and not that I'm saying they're justified I'm just saying I'm not a lawyer so I'm not sure how they would play in court. On the gig of Mexico talks and tariffs Rohan Patel said the risk of Trump placing import tariffs on cars from Mexico made by American companies is near zero. It would significantly harm US companies think GM and Ford who are currently making vehicles in Mexico and importing to the states. It would also very likely be illegal based on his own updated NAFTA agreement. This is just his usual bluster to scare Chinese companies away from setting up in Mexico. Much more likely is some national security related excuse to exclude specifically Chinese vehicles. But if we're looking at this from first principles I think Elon blaming Trump's political chatter is only part of the reason Gigamexico may actually be on pause.

埃隆在电话中确认了这一点，他说特斯拉已经从X AI中学到了很多。不过，我认为这对特斯拉的投资者来说是需要保持清醒头脑的，因为这起码在法律层面上为特斯拉、埃隆和X AI开启了更多可能的攻击途径。并不是说这些攻击是正当的，只是因为我不是律师，所以不确定这些在法庭上会如何处理。 关于墨西哥超级工厂的谈判和关税问题，罗汉·帕特尔表示，特朗普对由美国公司在墨西哥制造的汽车征收进口关税的风险几乎为零。这将严重损害通用和福特等目前在墨西哥制造并向美国进口车辆的美国公司的利益。而且，根据他自己更新的北美自由贸易协定，这也很可能是非法的。这只是他用来吓唬中国公司不要在墨西哥设厂的惯常伎俩。更有可能的是，他会用一些国家安全相关的借口来排除特定的中国车辆。 但如果我们从最基本的原理来看，我认为埃隆将墨西哥超级工厂暂停归咎于特朗普的政治噪音只是原因之一。

The other and likely more accurate reason is that Tesla currently has nearly 3 million units of production capacity but they're only producing about 1.8 million per year which just means Tesla still has over 50% growth potential at its current factories before needing to invest any money in new manufacturing lines at new sites. Elon did reiterate that the Robotaxi and Optimus production version 2, the high volume one, will both be built at Gigatexas. That's not to say Gigamexico isn't going to happen at all just that Trump likely won't be the reason that it doesn't and it may be 3, 4, 5 years before Gigamexico ever actually enters production. I'm not trying to say Elon's comments on Trump's policy chatter are not valid but on a prior conference call we were told that Gigamexico was intended for South America and some other markets in that the US was not going to be one of the main export regions. It's just that Elon's a smart guy, he knows most people aren't that far into the weeds and they'll hear that argument about why Gigamexico is on pause and think okay yeah that makes sense. And it does, I'm just saying I think there's more to the story.

另一个更可能且更准确的原因是，特斯拉目前拥有约300万台的生产能力，但他们每年只生产约180万台，这意味着特斯拉在现有工厂中还有超过50%的增长潜力，在需要投资新生产线之前无需投入新地点的新建工厂。埃隆强调，Robotaxi和Optimus第二版（量产版）都将在德克萨斯超级工厂生产。这并不是说墨西哥超级工厂不会存在，只是说特朗普大概不是导致它无法实现的原因，墨西哥超级工厂可能在3到5年后才会真正进入生产。我并不是说埃隆对特朗普政策言论的评论没有道理，但在之前的一次电话会议中，我们被告知墨西哥超级工厂是为了供应南美洲和其他一些市场，美国不会是主要的出口区域。埃隆是个聪明人，他明白大部分人不会这么深入思考，他们听到关于墨西哥超级工厂暂停的理由会觉得“嗯，那说得通”。确实如此，但我认为背后还有更多故事。

Touching on Tesla energy just briefly, this was a screenshot from the 10Q that was released this morning. As of the end of Q2, the total transaction price allocated to performance obligations that were unsatisfied or partially unsatisfied for contracts with an original expected length of more than one year was $5.71 billion. Of that amount, we expect to recognize $2.56 billion in the next 12 months and the rest over the remaining performance obligation period. To make this as simple as possible, let's say Tesla shakes hands on a new megapack contract for $100. Up front, they may recognize a portion of that revenue. Then when the megapacks are actually set up, deployed and connected to the grid, Tesla receives a bit more of that $100. But then there's a remaining portion, these unsatisfied performance obligations that Tesla will only be able to recognize after the project is turned on. Thus, over time, as Tesla signs more contracts for the megapack, these UPOs will call them will continue to build.

简要提及一下特斯拉能源，这是今早发布的10Q报告中的一张截图。截至第二季度末，未满足或部分未满足的履约义务总交易金额为57.1亿美元，这些合同的原始预期期限超过一年。在这部分金额中，我们预计在接下来的12个月内确认25.6亿美元，其余部分将在剩余的履约期内确认。为了尽量简化这个过程，我们假设特斯拉签订了一份价值100美元的新Megapack合同。在初期，他们可能会确认一部分收入。当Megapack实际安装、部署并接入电网后，特斯拉会收到更多的那100美元。然而，还有一部分未满足的履约义务，特斯拉只有在项目启动后才能确认收入。 因此，随着特斯拉签署更多的Megapack合同，这些未满足的履约义务（我们可以称之为UPOs）将继续积累。

Now, here's the kicker, and before I say this, I'm trying to actually confirm this with a few different accountants, but it's very tough to tell for sure. There is however a chance that these numbers that we're looking at on the screen will flow down largely to Tesla's gross profits when this revenue is actually recognized. That's because there's a good chance from an accounting perspective that when these megapack projects do actually go live and are turned on, Tesla is recognizing the majority of the cost at that point. If not some, even before that, during the production process. So if that's true, a large portion, if not all of this amount, could be flowing right down to Tesla's gross profits. Which yes, over time, as this is recognized, would indeed play a large role in actually boosting the gross margins and the percentages we're seeing on the Tesla energy business side. Matt Smith from Rebellionair shared some charts on this very data for our visual learners.

现在，这才是关键所在。在我说这之前，我正在尝试与几个不同的会计师确认这一点，但目前还很难确定。不过，有可能我们现在在屏幕上看到的这些数字在实际确认收入时会大部分流入特斯拉的毛利润。这是因为从会计的角度来看，当这些Megapack项目实际启动并运行时，特斯拉很可能会在那个时候确认大部分成本。如果不是这样，可能在生产过程中就已经确认了一部分成本。所以，如果这是真的，这个金额的大部分，甚至全部，可能都会直接流入特斯拉的毛利润。是的，随着时间的推移，当这些收入被确认时，确实会在很大程度上提升特斯拉能源业务侧的毛利率和百分比。Rebellionair的Matt Smith为我们的视觉学习者分享了一些关于这些数据的图表。

The blue bars are the total unsatisfied performance obligations you can see them increasing with each passing quarter. The orange is the total energy revenue that's actually recognized in that quarter. Again, gross oversimplification, but these blue bars or UPOs are kind of like a piggy bank that Tesla can access at certain times in the future. And listen, even if these UPOs don't flow through as nearly pure profit, it's also going to be much higher than zero, future revenue in profits that Tesla will recognize. This shows the total balance at $5.7 billion and the orange is the current portion which again is roughly $2.6 billion to be recognized over the next 12 months. Plus, Elon did say that the Shanghai megapactory set to come online Q1 2025 may actually triple Tesla's megapack output rather than just double it, which is what we were expecting. Tesla energy will definitely be bumpy quarter to quarter, however, it's going to play a large role in helping Tesla to actually grow into its current valuation even after today.

蓝色柱状图代表的是尚未满足的总业绩义务，你可以看到它们随着每个季度的推移在不断增加。橙色柱状图代表的是该季度实际确认的总能源收入。虽然这是一个非常粗略的简化，但这些蓝色的柱状图（UPOs）就像是特斯拉未来可以在特定时间取用的储蓄罐。即便这些UPOs不会直接转化为纯利润，它们带来的未来收入和利润也远高于零。 现在我们看到的总余额是57亿美元，橙色部分是当前的一部分，大约是26亿美元，将在未来12个月内确认收入。此外，埃隆·马斯克曾表示，计划于2025年第一季度上线的上海超级工厂，可能会让特斯拉的超级电池产量增加三倍，而不是我们预期的两倍。尽管特斯拉能源的季度表现可能会不稳定，但它将在帮助特斯拉增长以实现当前估值方面发挥重要作用。

On the energy side, I'm glad Elon and the team took some time to explain why the megapack really is so revolutionary for the grid and why Tesla is now building out its megapack backlog into 2026. The TLDW is that megapacks give existing power plants the ability to produce nearly two times the energy as they are right now because it allows these plants to operate in a more continuous manner because with the battery storage, they can actually store that excess energy. As it's been for the past five years, Elon reiterated that the expected demand for megapack is still likely to be underestimated by orders of magnitude.

在能源方面，我很高兴Elon和他的团队花时间解释了为什么Megapack对电网来说如此具有革命性，以及为什么特斯拉计划将Megapack的订单积压延续到2026年。简单来说，Megapack让现有的发电厂能够产生几乎是现在两倍的能源，因为它允许这些电厂以更连续的方式运行。有了电池储能，电厂可以实际储存多余的能源。过去五年来，Elon一再强调，人们对Megapack的需求预期可能仍然被大大低估。

Just to quickly touch on the auto side, we'll talk a lot more about this in the weeks to come, but something's not really adding up for 2025. As it stands now, Tesla would be rolling out a Model Y refresh, a more affordable vehicle sometimes supposedly in the first half of next year, and Roadster production is now supposed to enter production next year. And Tesla Semi is supposed to enter production by the end of next year. That's basically three and a half new vehicles for Tesla next year, and then you could add another half because the Cybertruck will still be ramping next year as well. Given that Tesla said the Robotaxi was still going to pursue the fully unboxed method on the slide deck, I'm not expecting any Robotaxi production until 2026 at the very earliest. I'd love to be wrong, but Tesla's 2025 is already quite booked up. That doesn't even consider the fact that Optimus production version 1 is supposed to start production next year followed by version 2 shortly after in 2026. All that to say, at this point, I am most certainly expecting some delays on those timelines.

关于汽车方面，我就简单提一下。在接下来的几周里，我们会详细讨论这个问题，但2025年的情况有些不太对劲。现在来看，特斯拉计划在明年上半年推出Model Y的改款和一款更便宜的车型，而Roadster也计划在明年进入生产阶段。此外，特斯拉的Semi卡车预计将在明年底前投入生产。基本上，明年特斯拉会有三个半新车型，如果再算上仍在提升产量的Cybertruck，可以算作四个半车型。 考虑到特斯拉在演示文稿中提到Robotaxi仍将采用完全拆箱的方法，我认为最早也要到2026年才能看到Robotaxi投产。希望我是错的，但特斯拉的2025年排期已经相当紧张了。这还不包括Optimus的第一个版本预计明年开始生产，第二个版本将在2026年紧随其后。总而言之，我非常确信这些时间表上会有一些延迟。

I think part of the reason that Wall Street isn't more excited about this more affordable Tesla is of course one we don't have really any information about the vehicle. Tesla's obviously avoiding the Osborne effect, but it's true this more affordable vehicle will likely have slimmer margins and when it comes to the FSD attach rate, we're still not really getting any insight. They did say the FSD attach rates improved materially after the price cut, but Vibov also said they were coming from a very low base. I also believe if the attach rates were good, Tesla would be sharing them. That's also why Tesla is pushing FSD demos so hard because they're seeing when people actually try it out they tend to keep using it, but it's just most people aren't even trying it out. Thus, in a sense regardless of how good Tesla's FSD actually is, the market is saying right now at that price point I'm really not that interested. It would be nice if we could have even a breakdown of these FSD buys, how many are buying it outright, versus how many are going the subscription route. But for now we're stuck with very little.

我认为华尔街对这款更实惠的特斯拉并没有那么兴奋，部分原因当然是我们几乎没有关于这辆车的任何信息。特斯拉显然是在避免“奥斯本效应”（指宣布新产品导致现有产品销量下滑的现象），但事实是这种更实惠的车型可能利润率较低。而关于完全自动驾驶（FSD）的安装率，我们仍然没有什么具体的见解。他们确实说价格下调后FSD的安装率有明显提高，但Vibov也提到这些增长是从一个非常低的基数开始的。我还相信，如果安装率真的很高，特斯拉会分享这些数据。这也是为什么特斯拉如此努力地推广FSD演示，因为他们发现一旦人们实际尝试了FSD，往往会继续使用，但问题是大多数人甚至没有尝试。因此，不论特斯拉的FSD到底有多好，市场目前的反应是：在这个价格水平上，我不太感兴趣。如果我们能有一个FSD购买情况的细分数据就好了，例如有多少人是直接购买的，有多少人是通过订阅方式购买的。但目前我们只能得到很少的信息。

A few quick ones in a new Tesla software update, you're now going to be able to control the brightness of the ambient lights in the Cybertruck and the new Model 3. Before, the brightness of the lights was directly linked to the brightness of the display. There's also a new Night Only option which turns on the accent lights when conditions outside get dim so you don't waste any of the lifetime of those LED lights during the day when they're not as visible. Additionally, when you're using the Charge Your Other EV option in the app, the options to sort and filter chargers have been updated. The first Model 3 performance deliveries in Canada are right around the corner as VINs are being assigned. The Irvine Police Department has been talking about it but they just confirmed on their official X account that the Cybertruck will be joining their police fleet soon.

在最新的特斯拉软件更新中，有几个快速变化。你现在可以控制Cybertruck和新款Model 3环境灯的亮度了。之前，灯光的亮度是直接与显示屏亮度关联的。另外，还有一个新的“仅夜间”选项，当外部光线变暗时，该选项会自动开启氛围灯，这样在白天可见度不高时就不会浪费LED灯的寿命。此外，当你在应用程序中使用“为你的其他电动车充电”选项时，充电站的排序和筛选选项已经更新。加拿大首批Model 3性能版的交付即将到来，车辆识别号码（VIN）正被分配。尔湾警察局一直在讨论这个问题，但他们刚刚在其官方X账户上确认，Cybertruck即将加入他们的警察车队。

Elon was at the Capitol building today and he said at no point did I say I was donating $45 million a month for Trump, that was a fiction made up by the Wall Street Journal. We haven't dropped into the macro world lately but it's worth pointing out if you haven't been paying attention there's currently over a 90% chance that we get a rate cut at the September meeting. The sooner and more often we can get these rate cuts the better it will be for Tesla and the less they'll actually have to bear the brunt of offering those lower APRs. On that note, Gary Black was able to confirm with Tesla IR, the present value of the discounted loan rate is accounted for upfront at the time of sale so discounting a Tesla by $5,000 compared to offering a low cost loan impacts Tesla earnings about the same. As Gary said, the low APR offers don't impact residual values and Tesla can end those at any time but as we said heading into quarter to earnings don't forget about the margin impact that those low APR rates were likely to have.

埃隆今天在国会大厦，他说：“我从未说过我要每月捐款4500万美元给特朗普，那是《华尔街日报》杜撰的。” 我们最近没有关注宏观经济，但值得一提的是，如果你没注意的话，目前有超过90%的机会在9月会议上降息。降息越早越频繁，对特斯拉就越有利，他们就不必承受提供低APR（年度百分率）的压力。 关于这一点，Gary Black确认了特斯拉投资者关系部门的消息，贴现贷款利率的现值将在销售时一次性计入，因此相比于提供低利率贷款，直接降价5000美元对特斯拉收益的影响是差不多的。正如Gary所说，低APR的贷款并不会影响特斯拉汽车的残值，特斯拉可以随时结束这些优惠，但如我们在第二季度财报前所说的那样，不要忘记这些低APR利率可能对利润率的影响。

On the financials, it's true that Wall Street has been modeling a bit more for Tesla energy and getting excited about Robotaxi potential but it's plain as day that this line right here, auto gross margin X credits is still largely driving the stock. Going into quarter to many analysts were expecting this line to have bottomed in quarter one and that was certainly not the case. Operating margin is a line we should be looking at every quarter because that's what Tesla has said they're currently optimizing for. The headline number was 6.3% which was below expectations but if we remove that $622 million restructuring charge that was from the layoffs, the operating margin would have been 8.7% the best reading in the past four quarters.

关于财务状况，确实，华尔街对特斯拉能源的预期有所提升，并对Robotaxi的潜力感到兴奋，但很明显，这一行，即“剔除信贷的汽车毛利率”，仍然在很大程度上驱动着股票价格。在进入第二季度时，许多分析师期望这一指标在第一季度触底，但事实并非如此。每个季度我们都应该关注运营利润率这一指标，因为特斯拉表示他们目前正在优化这一点。标题数字是6.3%，低于预期，但如果我们去掉6220万美元的重组费用（来自裁员），运营利润率将为8.7%，这是过去四个季度中的最佳表现。

I know people will argue that the regulatory credits were a lot higher than expected so those two can effectively be awash because $350 million of that restructuring charge was expected heading into Q2. But the reason I don't look at it that way is because the restructuring charge really is a one off that should be backed out. The regulatory credits have been going on now for years and based on what legacy auto is doing, in this case not doing, I think those regulatory credits will proceed into the future. There certainly could be policy changes but Tesla said on the call they're planning their business as if all of these IRA benefits and regulatory credits could go away. This is not a number that Tesla reports but one that I like to track, regulatory credits as a percentage of operating income this quarter it was up at 55.5%.

我知道人们会争辩说，监管信贷远超预期，所以这两者基本可以相互抵消，因为进入第二季度时，预计有3.5亿美元的重组费用。但我不这样看，是因为重组费用实际上是一次性的，应该被排除在外。而监管信贷已经持续多年，基于传统汽车制造商的表现——在这方面是没什么表现的——我认为这些监管信贷在未来还会继续。当然，政策变化是可能的，但特斯拉在电话会议上表示，他们的业务规划是基于所有这些《通胀削减法案》（IRA）福利和监管信贷可能消失的假设。这不是特斯拉报告的一个数字，但我喜欢追踪本季度监管信贷占营业收入的比例，这个比例上升到55.5%。

Again though even if you take the restructuring costs out of this equation that number still would have been 40% a new high for Tesla and if you can't tell we want this number to be low. While these regulatory credits may be around for another few years at some point they are indeed likely to go away so this number shows you kind of the true health of Tesla's business without those red credits. We learned the Cybertruck is expected to reach profitability by the end of this year which means for the rest of the year the Cybertruck will indeed boost Tesla's average selling price but it's also going to be a drag meaning it'll be increasing Tesla's cost of goods sold. It's not profitable now they're selling it for around $110,000 so the Cogs per unit is potentially north of $100,000.

不过，即使你把重组费用从这个方程式中剔除，那个数字仍然会是40%，这是特斯拉的一个新高。如果你还没明白，我们希望这个数字是低的。尽管这些监管信用可能还会存在几年，但某个时候它们的确可能会消失，因此这个数字展现了特斯拉业务的实际健康状况，不包括那些监管信用。我们了解到，Cybertruck预计将在今年年底实现盈利，意味着在今年剩余时间里，Cybertruck确实会提升特斯拉的平均销售价格，但它也会成为一个负担，因为它将提高特斯拉的销售成本。现在它还不盈利，他们以大约11万美元的价格出售，所以每辆车的生产成本可能超过10万美元。

So when it comes to the sexy lineup there really isn't that much more room to take off any more Cogs which means barring a demand spike and price increases we really are in this waiting period for next generation vehicles and products. While Tesla energy has largely been the star of the show as of late one thing to keep in mind on much greater energy revenue and profit for the quarter the gross margin came in flat quarter over quarter which likely means we're seeing the beginning of the impact of Tesla reducing the mega packed price dating back to last year. Over that time the pricing for the mega pack has effectively been cut in half so it's going to be interesting to watch this energy gross margin number balancing the price cuts with the recognition of these UPOs. Tesla stock closed the day at $215.99 down 12.33% while the Nasdaq was down 3.64%.

所以，当谈到这个诱人的产品阵容时，实际上已经没有太多空间可以进一步削减成本了，这意味着除非出现需求激增和价格上涨，否则我们真的只能等待下一代车辆和产品的到来了。虽然最近特斯拉能源几乎成了焦点，但需要注意的是，尽管本季度的能源收入和利润大幅增加，毛利率季度环比却持平，这可能意味着特斯拉去年开始降低的Mega Pack价格的影响已经开始显现。期间，Mega Pack的价格实际上被削减了一半左右，因此观察这一价格削减与这些未完成订单收入确认之间的毛利率平衡将会很有意思。特斯拉股价收于215.99美元，下跌了12.33%，而纳斯达克指数下跌了3.64%。

We'll wait and see how Wall Street updates their Tesla EPS expectations for this year to look at the forward PE ratio but annualizing quarter two gap and non-gap has Tesla's forward PE over 100. That number could certainly run up to three four five hundred in the future it's been in the thousands before but based on Tesla's current earnings per share growth and the expectations I would say that's very unlikely. I did buy some shares today but I think there could be some weakness for Tesla stock in the weeks ahead a retest of 200 is certainly on the table and if it breaks through 170s are back in play. Not financial advice just sharing where I'm coming from I did use some of my dry powder today after this move but definitely not all of it as I think there could be better opportunities ahead.

我们会静观其变，看看华尔街如何更新对特斯拉今年每股收益 (EPS) 的预期，以便了解其未来的市盈率 (PE)。但按季度二的数据显示，按年度计算，特斯拉的未来市盈率超过了100。这个数字未来可能确实会涨到三四百甚至五百，过去也曾达到几千，但根据特斯拉当前的每股收益增长和预期来看，我认为这很不可能。我今天确实买了一些股票，但我认为在未来几周内，特斯拉的股票可能会表现疲软，重新测试200美元的水平是有可能的，如果跌破这个水平，170多美元也是有可能的。这不是财务建议，只是分享我的看法。我今天确实动用了一些现金进行投资，但绝对不是全部，因为我认为未来可能会有更好的机会。

To clarify one more thing Elon said he said that Tesla's self-driving capability that's been deployed in Europe and other regions is significantly behind that in the US. All he was saying was that in geographies outside of North America they have dumbed down versions of Tesla's FSD whether it's just autopilot or just enhanced autopilot. He was not saying the full version of Tesla's FSD is significantly behind for other regions because it for some reason would need to be trained on their specific data. Then the team went on to explain how adding new video data from new geographies won't really impact FSD that much it's built to work really anywhere on the planet or even another planet. And Elon did say he thinks that they'll get regulatory approval to release FSD in Europe, China and other countries before the end of this year but we'll see.

为了澄清Elon所说的另一件事，他提到特斯拉在欧洲和其他地区部署的自动驾驶功能明显落后于美国。他只是说在北美以外的地区，他们使用的是简化版本的特斯拉全自动驾驶系统（FSD），不论是单纯的自动驾驶还是经过增强的自动驾驶。他并不是说其他地区需要基于当地特定的数据进行训练，因此完全版的特斯拉FSD在这些地方会显著落后。团队还解释说，添加来自新地理区域的新视频数据对FSD的影响不大，因为它的设计是可以在地球上的任何地方甚至另一个星球上工作。Elon确实表示他认为他们将在今年年底之前获得在欧洲、中国和其他国家发布FSD的监管批准，但具体如何还要再看。

Tying back to earlier in the video though if that happens Tesla would likely need a lot more training compute once they get to wide release with FSD and Europe and China. Just be prepared for the potential storm that is looming if Tesla gets to a point where they're scrambling for Nvidia Compute.

回到视频前面提到的内容，如果这种情况发生，当特斯拉的全自动驾驶（FSD）在欧洲和中国广泛发布时，可能需要更多训练计算资源。请准备好应对潜在的挑战，如果特斯拉到时候急需英伟达的计算资源，可能会迎来一些麻烦。

As Vibov said the future is still as bright as ever but patience will be required. Hope you guys have a wonderful day please like the video if you did you can find me on X-linked below and a huge thank you to all of my patreon supporters.

正如Vibov所说，未来依然光明，但我们需要保持耐心。希望大家有美好的一天。如果你喜欢这个视频，请点赞。你可以在下方找到我的X-link。衷心感谢所有的Patreon支持者。