Battery Storage Is Booming

Not sure what’s up buy my energy levels have been totally MIA the past two days. Was a miracle I got a video out yesterday lol…today, no such fortune

FYI: any text in italics denotes it’s my take or my own added context. I’d also recommend: after you click an X hyperlink to login on the web if you’re not already. It will make the experience much better and you should also have the option to open the links directly in the X app if you’d prefer

⚡️ Karpathy’s Interview

I’m guessing many of you have seen this already. If not though, I figured I’d share the main Tesla takeaways and a quick thought or two. If you’re into AI, it’s worth watching in its entirety

TLDW:

• Andrej said he feels like we’ve reached AGI a little bit in self driving. He mentioned he was very bullish on how Tesla is approaching autonomy and AI a few times. He believes Tesla is actually ahead of Waymo despite many people thinking the opposite, arguing Tesla now has a software problem and Waymo has a hardware problem. Andrej thinks the software problem is a lot easier to solve. This is another way of saying Waymo can’t scale…or at least has a very difficult road ahead to do so. Andrej thinks when we look back around 10 years from now, most of the revenue for autonomy will be coming from Tesla. He reminded everyone Tesla does actually use expensive sensors, but it’s in the training environment only. Tesla distills those learnings into a package deployed to cars as a vision only system and he believes this strategy will work out well. Andrej was also quite bullish on how much of Tesla’s FSD learnings will directly transfer to Optimus. Not only that, but how much of the manufacturing expertise will transfer, highlighting how Tesla is uniquely positioned to solve this problem.

• I think the most important point to be drawn from this interview as it relates to Tesla is that competitors trying to copy Tesla starting with an end to end NN are going to face challnges, largely a lack of data. Andrej explained how you really need to build up the E2E NN incrementally, piece by piece (like Tesla did) and that process took Tesla years. There have been a lot of questions about what Tesla’s lead really is after moving away from C++ code to the E2E network and I think this commentary from Andrej is the best answer we’ve gotten. The lead remains intact

⚡️ Utility Scale BESS Booming

In the last 12 months, the US added 8.97 GW of utility scale battery storage. Typically these projects are 4 hour duration, so this is likely ~35.9 GWh. Over the next 12 months, the US is expected to add 17.6 GW. Nearly twice the amount added in the past 12 months

By June 2025, the US should have 37.5 GW / 150 GWh deployed. Just a random fun fact, that’s 150,000 MWh or about 37,500 Tesla Megapacks. Lathrop can produce 10k Megapacks at max capacity each year, so it would take Tesla nearly four years to fill that level of demand

Separately, a solar + battery microgrid on the coast of Australia saved more than $1.5M in diesel fuel costs in its first two years of operation. The proof this is the blueprint of the future continually rolls in

⚡️ Lightning Round

• Tesla’s referral program has now expanded to Germany, Norway, France, the Netherlands, the UK and of course North America

• Cybertruck buyers are now getting the $2,500 voucher deposited into their Tesla Wallets rather than a shop voucher. This way they can use it for things like service or a new Tesla. This is likely so the owners that can’t have PowerShare installed for whatever reason have more options with that credit

• TSMC achieved production yields at its new Arizona facility on par with established plants back in Taiwan. TSMC is using 4-nanometer tech here and their biggest known customers are Nvidia and Apple but two years ago there were reports Tesla was going to tap TSMC to make HW4 in Arizona. We haven’t heard anything official but there’s a chance this facility produces for Tesla at some point in the future. There have been concerns TSMC couldn’t make chips successfully in America (shortage of qualified labor, costs etc) so this is a welcome report to calm some of those fears

• Toyota cut its EV production plans for 2026 by 1/3rd from a previous target of 1.5M down to 1M. They gave some confusing commentary saying these figures were not targets but benchmarks for shareholders. Toyota only sold 104,000 BEV’s last year so the 1M target is still a massive increase. Color me skeptical they hit even the reduced target

• Tesla is going to continue working with OEM’s including EVject to provide safe charging adapters. If you’re not familiar with EVject, it’s touted as a safety breakaway feature to quickly drive away while SuCing (think carjacking or thefts). Here’s a video. Tesla sued them alleging the connectors get dangerously hot and EVject confirmed they had no thermal protection and recalled all of them. EVject said they’re working on a next gen version that will have thermal protection. More here 

• RJ Scaringe was on a podcast this week and he talked about why he made the intentional decision to be different than Tesla ⤵️ (a wise choice, unlike Lucid)

• There was a big sell off in the markets today with Tesla falling 8.45% down to $210.73. It capped the Nasdaq’s worst week since 2022. Some point to US payroll additions coming in short showing the economy slowing stoking recession fears. The debate is whether the Fed will cut 25 or 50 bps in September. Times like this are good reminders of how much stock price action is driven by external factors with little to no relevance to company fundamentals

Hope you all have a great weekend, enjoy!