Transcription
So, on the day Tesla announced its best quarter ever, Elon posted nothing about it. If you own Tesla, what he posted instead, that's the real story of the week. His feed was robots in Fremont and a new Grok model already running inside SpaceX and Tesla. And I think those two posts tell you more about where this company's going than the delivery number does.
Now, I've already covered the delivery story a few times already, so today it's going to be about something different. Where Elon puts his attention has always been, I think, the best early signal for what Tesla does next. If you remember, he camped out at the factory before the Model 3 ramp cracked. He flooded his feed with FSD before Robotaxi went live in Austin. The post come first and then the business follows. So, when he gives us four days of posts, I want to pay attention to them.
So, the plan today, first, the photo from Fremont and why him walking that specific line matters. Then we're going to talk about the Grok 4.5 post, which I'll read to you word for word because there's a detail in the last sentence I haven't seen anyone cover. Then what the two posts look like when you put them side by side. Then the thing he didn't post this week, which might be the loudest signal of all. And at the end, the honest bear case because Michael Burry, well-known short, just took the other side of this exact trade.
Let's start with the photo. On July 1st, Elon posts five words. He said, "Walking the Optimus production line in Fremont." The photo shows him standing with a production team. They've got the hard hats on. They've got the safety vests all around. But look at where he's standing. That's the old Model S and Model X line. The last S and X rolled off in early May, and Tesla spent the summer converting that floor into an Optimus production line. So, the line that built the cars that built this company is now set up to build robots. And Elon flew up to walk it in person and make sure we all saw that photo. He's done this before and it's worth remembering what it meant.
So, when Elon physically plants himself on a factory floor, that program has become the company's front burner. It happened with Model 3 when the company's survival ran through a tent in the Fremont parking lot. It happened with the early Y ramp. The pattern is pretty consistent. He shows up where the next fight is, and the org chart reshuffles around wherever he's standing. Engineers get pulled, budgets get unblocked, and suppliers suddenly pick up the phone. A CEO's physical presence is the one resource that can't be faked in a press release. And think about what it took to get to that to this photo. Tesla killed two of its flagship products to clear this floor. The S and X weren't dead weight either. They were profitable halo cars with 14 years of history. Ending them was a real cost, taken on purpose to free up space and people for robots. Companies only do that for a science project.
And there are real details underneath that photo if you take closer look. In my interview with Lars Moravy, Tesla's vice president of engineering, he said this line is modular, built to be reconfigured as the robot's hardware evolves, with dozens more lines planned that stand up way faster than a traditional car line. That modular part matters more than it sounds. The standard dogma on Optimus manufacturing has always been that you can't mass produce a robot whose design keeps changing every 6 months. A car line is poured in concrete, basically. You change the product, you rip up the line. Lars is describing a line built for a product that's still evolving, which is Tesla admitting the design will keep moving and building for it anyway. That's a manufacturing answer to an engineering problem, and manufacturing is the thing this company's actually really good at. The eventual target for Fremont is 1 million robots a year. That's the target. And I want to be very careful here. Targets and run rates are very different things. Today it's one line, production hasn't started. Elon says it starts late July or August. Which lines up with the other data on the calendar. Optimus version 3, right? The production design is supposed to get revealed in that same window. He said that the robot is walking around already and just needs finishing touches. So the photo, the line, the reveal, and the production start all converge on this next four to six weeks. That's the setup.
My take on that photo is pretty simple. He's marking the handoff from Tesla's first act to a second. In a single image on that exact factory floor where the first act happened. Nobody at Tesla does investor communication that way except Elon. That was for us.
Now, go back three days earlier, June 28th. That was a different post and honestly, maybe the bigger one. I want you to read the whole thing word for word because the details matter here. He says, "Grok 4.5 based on our 1.5 trillion V9 foundation model with cursor data added in supplemental training is now in private beta at SpaceX and Tesla. Early evaluation show performance close to, perhaps exceeding Opus. Reinforcement learning is continuing to significantly improve the model and the Grok build harness gets better every day. Nice work by all those involved. Completely trained from scratch, new models will be released by SpaceX every month this year."
Okay, so let's take that apart because there are three things in there. The first is where the beta's happening. So, Grok 4.5 is being tested now inside SpaceX and inside Tesla before the public ever touches it. His companies are each other's first customers. A frontier AI model gets stress tested by rocket engineers and vehicle engineers as it's shakedown cruise. And whatever they learn flows back into the model before anyone else sees them. Day-to-day that means Tesla engineers running it on real firmware, real manufacturing problems, real vehicle code, and every failure becomes training feedback. Most AI labs would pay serious money for that kind of testing ground. Elon owns two of them. Tesla put $2 billion into XAI and this is what that pipeline looks like in practice. Every model upgrade lands inside Tesla first. And it's already producing products, by the way. The digital optimist project. Grok agents running computers the way the robot handles physical work came directly out of that same investment agreement. No other car company on Earth gets a Frontier Labs models pre-release, let alone joint products built on them.
There's a cursor detail in there worth 30 seconds, too. Cursor is the coding tool that ate the developer market over the last 2 years, the one engineers actually pay for. That data got bolted into Grok's supplemental training. So, the model heading into Tesla is being sharpened specifically for code and hard technical work, which is exactly the flavor of AI car company and a rocket company can actually use.
All right. The second is the benchmark epic. Close to perhaps exceeding Opus. Opus is Anthropic's top model, basically the gold standard for hard reasoning and coding work right now. Notice the hedge in his words, though, perhaps. These are internal valuations, no public benchmarks yet, so file that claim as then unverified. But, the target tells you the ambition. He's measuring Grok against the best model in the world and putting that measurement inside his own factories.
Okay, and the third is the last sentence, the one I haven't seen anyone cover. New models trained completely from scratch, released every month this year by SpaceX. We do ship the models. XAI got folded into SpaceX, so the frontier era AI lab now lives inside the rocket company that just went public, which means the AI cadence, the compute, the Memphis data centers, all of it sits one corporate layer away from Tesla. Connected by that $2 billion investment and by you on himself. And the connections run in both directions, physically. SpaceX bought about $269 million of Tesla Megapacks this spring to power XAI's data centers around Memphis. So, Tesla high-energy hardware powers the training runs, the trained models beta inside Tesla. Tesla's cash sits in the lab making them. Money, power, and intelligence doing laps around the same three companies. When people ask why the merger conversation will die, this loop is one reason.
So, a monthly cadence from like starting from scratch frontier models would be a brutal pace. I'll believe each month when I see it. Elon dates do slip, and we all know this. But even at half that pace, Tesla's AI supplier is iterating faster than any traditional partnership could.
Now, let's hold these two posts next to each other because this is where the week gets interesting. One post is about software, a frontier model moving into his companies. The other is about hardware, a robot production line standing up in Fremont. Grok is the brain coming down the pipe, and Optimus is the body waiting for it. And the factory in the photo is where the two meet. That's the whole thesis in two posts. Tesla makes physical things move around the world, xAI makes the intelligence that will drive them, and SpaceX now holds the AI lab and the compute. Four days of Elon's feed sketched the entire machine. And it matches where the money and the chips are going.
Elon has said the next generation of Tesla AI chips goes to Optimus and to the training clusters before it goes into cars. That would have been unthinkable 3 years ago. The best silicon Tesla designs now skips the car entirely on its first pass. The robot and the training clusters eat first, and the vehicles get the trickle down. Chip allocation is the most honest org chart a hardware company has, and Tesla says robots outrank car. Reports put Tesla's robotics and autonomy spend this year north of $25 billion. The Fremont line, a second Optimus factory planned at Giga Texas, the Cybertruck ramp, a chip fab, you got attention, capital, and silicon all pointed the same direction. Robots first, cars funded.
Compare that setup to everyone else for a second. The legacy automakers rent their AI. They license from Waymo or Nvidia or whoever wins. They wait in line for GPUs with everyone else, and their AI partner's road map belongs to someone else's board. Tesla's version is the lab it invested in, run by its own testing models inside its own factories on chips it designs, powered by batteries it builds. You can hit the corporate angle of it, plenty of governments people do, and that's a fair fight. But as an operating setup for shipping robots, there's no second example like it anywhere.
Zero posts this week hyping vehicles, by the way. From the guy who runs a car company during the biggest delivery week of the year. Which brings up the part of the week that I think got misread the most. So on Thursday, Tesla reported 480,126 deliveries for the second quarter. It was the best second quarter in company history, up 25% and it beat the estimate by around 74,000 cars. Energy had a monster quarter, too. I'm keeping this part short on purpose since I've covered deliveries already, but what I care about here is what Elon did with the news. And the answer, as of this recording, is basically nothing. No victory lap, no record quarter, congrats to the team. None of those were he didn't post any of those on his feed. The old Elon celebrated delivery records loudly. Delivery day used to be a whole event on his feed, reposting the milestone graphics, thanking the team, poking the shorts. This week's Elon was posting about robots while the record dropped.
And his own words back in January already told us why. He said he doesn't expect unsupervised FSD or robotaxi revenue to be super material this year. But that it gets material, probably in a significant way, next year. He told investors to be patient with Optimus and robotaxi in almost the same breath. So the man said the near-term car numbers carry less weight in his own model of the company, and then spent record delivery week showing us robots. The feed matches the words. That consistency is what makes it a single, more of a signal instead of a noise. I don't know, is that me reading tea leaves? Yeah, it's a little. I'll own that. He posts 30 times a day and a delivery post could still show up. But the pattern across the whole week points one way. The metric Wall Street spent delivery day obsessing over is one he's moved past a long time ago. He told investors back in January to be patient with Optimus and Robotaxi, that the payoff builds into next year. He's acting like someone who believes that.
And the market gave us the mirror image of the same message. Tesla fell about 8% on delivery day, on the best delivery quarter ever. The stock had run up 12% into that number. The beat landed and it sold off anyway. So, when a record quarter can't move a stock up, deliveries have stopped being the thing the valuation breeds on. Wall Street and Elon actually agree on that. They just disagree on what replaces it.
Which brings me to Michael Burry, because the same four days gave us the counterparty. So, on June 30th, right in the middle of the week, Burry disclosed in his newsletter that he shorted Tesla at $416. And he added, "Happy it jumped back to this level." And it's bigger than a Tesla bet. Same letter has him short Nvidia, Caterpillar, Applied Materials, and a separate conductor ETF. He's calling the whole AI build-out the beginning of the end. In his view, the CapEx is a bubble. The returns won't show up and Tesla's priced like the robots are already rolling off the Fremont line.
I want to give him a fair hearing here, because dismissing shorts is how investors get sloppy. The valuation is demanding. That's just true. Robotaxi revenue, by Elon's own words, won't be super material this year. Optimus revenue is basically zero today. So, a big chunk of the market cap rides on things that haven't shipped. And Burry's whole career is built on spotting exactly that shape. If the AI trade cracks across the board, Tesla goes down with it, no matter what's on Elon's feet. Stocks trade in herds, and this one runs with the AI herd right now.
But, notice what the week actually gave us. Burry's shorting a spreadsheet. Elon spent the week standing on a factory floor. One of them is betting the AI build-out never pays for itself, and the other one just showed you the robot line, the model running inside his companies, and the monthly cadence. Same four days, both sides of the trade, laid out in public. I don't remember a cleaner split screen than this one.
So, a couple honest footnotes on the short. He didn't disclose the size, so it could be a massive conviction bet, or could be a small hedge. We have no idea. And he's been early or wrong on Tesla many times before. He shorted in 2021, and then he had to cover. Track records cut both ways.
Now, the pushback that matters, and I want to do it properly. The strongest objection to this whole show is that reading Elon's feed is vibes, and vibes aren't a thesis. That's fair. A photo on the production line is just a photo. Private beta means unverified evaluations. Late July or August comes from the same guy whose ramp dates have slipped for a decade. If you marked every Elon timeline to market, you'd apply a healthy discount to all of it, and you should.
There's also a harder version of the objection. Even if the signals are real, the gap between a production line existing and a robot business making money is measured in years. Model 3 taught us that. The line existed long before the profits did. And the stock chopped violently the whole way through. So, let me hedge this the way I'd want it hedged for my own money. The dates will probably slip. The 1 million robots a year is a target. Targets are marketing until they've they're turned into run rates. What I trust on this week is the direction, because attention, CapEx, and chips are all pointed at the same place. And those three things are expensive to fake. Elon can post whatever he wants, but converting your most historic production line into a robot line is a physical, irreversible commitment. That happened. The photo just proved it.
So, let's zoom out and make this practical. If you own Tesla stock, this week told you what the stock actually trades on now. You got a record delivery quarter produced an 8% down day. So, stop grading your position on quarterly delivery beats, because the market stopped. The re-rating events from here are the Optimus 3 reveal, the production start in Fremont, the robotaxi expansion, and each Grok release cycle feeding into the products. Those are the dates that move the needle.
It also gave you a repeatable habit, and I think this is the most useful part. Watch what Elon emphasizes, and watch what he skips. His feed front-ran Austin. It front-ran the SNX shutdown. This week it's pointing at Fremont and at Grok. You don't need to trade on it. You just want your expectation set by his attention and by the calendar he keeps confirming, instead of whatever the headlines decided mattered that day. One practical way to do this, he posts 30 or so times a day, and nearly all of it are replies or memes and politics. You skip those. The business posts, the ones with a photo from a factory or a spec or a date, show up maybe two or three times a week. Those few posts are him talking to investors, whether he frames it that way or not. And this week two of them pulled over 13 million views between them, while the delivery report got a froze from his feed. The ratio is the message. And hold both hands of the trade at once. Very short is a real reminder that the whole AI complex is priced for success. If robots and autonomy slept badly into 2027, this price has room to fall. If the Fremont line ships robots on anything close to schedule, the delivery number everyone argued about this week won't even make the highlight reel.
So, what do you actually do with all this? I think there's five things I'd watch. The first is the Optimus 3 reveal, that should come hopefully late July or early August. That's according to Elon's timeline. Judge the hands, the walk, and whether they show it doing real work. Then, there's a first confirmed production on the Fremont line. An actual robot coming off that line is the milestone separates this cycle from every Optimus promise before it. Third one is the Grok cadence. Elon said, "New from scratch models every month, shipped by SpaceX." Count the months. Every month that lands on time makes the monthly cadence claim real, and every mass tells you how much to discount the rest. Fourth, Q2 earnings in about 3 weeks. Listen for how much of the call is robots and AI versus cars. My guess is it's most of the call. And the margin questions get 2 minutes. And the last one, keep an eye on Burry's next disclosure. If the size of that short ever shows up in a filing, you'll know whether it was a conviction or just a hedge.
So, summary, Wall Street spent delivery day reading a delivery report. Elon spent the week showing us what comes after it. The robot line is real. The model is in the building. And the two posts that matter got less coverage than the number that didn't move anything. Watch the feet. Watch the dates. And keep your expectations honest in both directions. You know, hopefully you enjoyed the show and found it valuable. If you did, please subscribe. It helps a lot. And I'll see you in the next one. Thanks.