Live: Tesla Q2 Earnings Call 2026 (TSLA)
Tesla Daily
Opening Remarks and Record Deliveries 0:00
Elon Musk opens the call by noting record second quarter deliveries, with Model Y described as the bestselling car of any kind in the world. He says demand for full self-driving, known as FSD, is so strong in approved regions that many customers are effectively buying the software with a car attached rather than the reverse, and he expects similar demand jumps as FSD gets approved in more countries.
A Massive Capital Spending Year 2:00
Musk describes this as a huge capital expenditure year, covering new production lines for the Cybercab, Optimus, the Tesla semi truck, Megapack 3, a lithium refinery, a cathode refinery, battery cells, and an eventual solar cell and panel production chain running from silicon refining to deployment. He frames this as preparation for the growing electricity needs of transport electrification and artificial intelligence, calling it possibly the most ambitious infrastructure buildout in history.
Robotaxi Safety Over Speed 4:01
Musk explains that Tesla wants to scale robotaxi as fast as possible but is constrained by the need to avoid causing any harm, noting that while tens of thousands of Americans die in car accidents yearly without much press attention, a single robotaxi injury would become global headline news and invite regulatory crackdown. He says the company posts live updates on X as it expands to new cities, having already launched in parts of Florida, Texas, and the Bay Area, with weekly miles driven growing more than 10 percent.
Optimus as the Hardest Product to Scale 6:31
Musk calls Optimus potentially the biggest product Tesla will ever make, but also the hardest to manufacture because, unlike electric cars which could draw on an existing auto supply chain, almost nothing in the humanoid robot has an existing supply base. He says Tesla has had to build much of that supply chain in-house, including a new production line in Fremont on the site of the old Model S and X line, and warns that the early ramp will look flat and slow before following a normal manufacturing S-curve.
Human-Level Dexterity and Digital Optimus 11:31
Musk says Optimus is designed for full human hand dexterity, eventually surpassing it, and distinguishes real demonstrations from the pre-programmed or remote-controlled robot videos common online. He also introduces the concept of Digital Optimus, an AI system that operates a computer screen the way Tesla's software drives a car, taking in real-time video and issuing controls, managed in partnership with SpaceX's Grok model, and describes plans to package these AI4 computers with x86 hardware into mobile Megapod units that can draw power from sites like Tesla's superchargers, which already have around 7 gigawatts of capacity.
Terafab and Chip Development 11:31
Musk says Tesla will soon announce a location for its Terafab chip facility but wants to give that its own dedicated announcement rather than fold it into an earnings call, describing it as essential for solving memory, logic, and packaging constraints that would otherwise limit Optimus production. He mentions equipment orders already placed for a development fab in Austin that will house lithography, mask production, logic, memory, packaging, and chip testing under one roof, which he believes doesn't exist elsewhere and will allow fast, high-risk experimentation on AI chip designs.
Financial Results and Margins 15:30
CFO Vaibhav Taneja reports record second quarter deliveries with sequential growth of 60 percent in the Americas, 27 percent in APAC, and 12 percent in the Middle East and Africa region, plus the largest order backlog since 2023. Automotive gross margin excluding credits fell from 19.2 to 16.3 percent, partly because a 230 million dollar benefit from warranty adjustments and tariff relief in the prior quarter did not repeat, and rising interest rates increased financing subvention costs. Energy storage deployments reached 13.5 gigawatt hours, up 53 percent sequentially, though energy gross margin dropped from 39.5 to 20.4 percent due to a 240 million dollar warranty true-up tied to legacy cell issues, the loss of prior tariff benefits, and falling industrial storage prices, with long-term energy margins expected to normalize in the mid to low 20 percent range.
Spending, Debt, and Cash Flow 18:01
Taneja notes operating expenses rose mainly from R&D spending on new products like the semi truck, Optimus, and Cybercab, plus added compute depreciation and litigation charges, with further growth expected into 2026. Net income got a 1 billion dollar mark-to-market boost from Tesla's SpaceX holdings, partly offset by roughly 300 million in foreign exchange losses and 100 million in Bitcoin losses. Free cash flow turned negative as capital expenditure more than doubled sequentially, with full-year capex expected to exceed 25 billion dollars and keep rising over the next two to three years, prompting Tesla to arrange debt facilities allowing it to borrow up to 30 billion dollars to help fund the expansion.
Robotaxi Safety Record and Growth Rate 23:30
AI vice president Ashok Elluswamy reports that Tesla's robotaxi program has driven more than 380,000 miles unsupervised across six cities in two states with zero notable incidents, arguing this validates Tesla's camera-only approach against claims that lidar, radar, and HD maps are necessary. He says the unsupervised fleet has grown at double-digit weekly rates since the start of the year, expects the time needed to launch new cities to keep shrinking toward eventual statewide operation, and notes the fleet is already running early versions of the V15 self-driving software, with about 40 percent of its planned improvement tracks merged in.
Optimus Training Data Strategy 27:30
Elluswamy explains that Optimus, especially Generation 3, is built to match human form and dexterity so it can learn tasks by observing people, drawing on Tesla's factory workforce, a dedicated small but high-quality demonstration data team, and broader internet video of humans performing tasks. He describes a second data source, called the Optimus Academy, where robots practice tasks themselves and improve through reinforcement learning from their own successes and failures, using the same end-to-end, pixels-in-controls-out approach and engineering team behind FSD versions 12 through 15.
Analyst Question on Supplier Partnerships 31:02
During the analyst question period, Andrew from an unspecified firm asks whether any suppliers are willing to invest alongside Tesla in domestic manufacturing to help scale Optimus production faster in the United States, given the lack of an existing supply chain. Musk begins to respond by saying Tesla's suppliers have been great and are making significant investments to support Optimus and robotaxi, specifically naming Samsung and TSMC as building fabs, with TSMC in Arizona, as the transcript cuts off.
Supplier partnerships and thanks 32:30
Elon Musk, sounding under the weather, thanks Tesla's manufacturing partners for their investment and effort, mentioning Samsung's massive fab expansion for future projects, Panasonic's billions poured into battery cell production, and Micron making room in its allocation for Tesla despite very tight and expensive memory supply. A Tesla executive adds that new specialized parts like metal injection molded components and flexible printed circuits are being sourced for robotics rather than the traditional vehicle supply chain, and that Tesla brings manufacturing in house whenever a suitable outside partner cannot be found.
Regulatory approach to self driving 35:01
Asked about evolving state level rules on sensor requirements for robotaxis, Lars explains that Tesla has made good progress with federal regulators, praising NHTSA's movement on FMVSS rules toward accepting driverless vehicles. He calls some state level developments, referencing New Jersey, discouraging, and argues the best regulations set a goal and let companies find the solution rather than prescribing a fix before the problem is even defined. He says Tesla will keep letting its safety performance speak for itself as the basis for expansion.
Robotaxi scaling and reliability bar 37:00
Elon says Tesla intends to keep robotaxi fully vertically integrated rather than partnering with ride share companies, expecting demand to far outstrip the service's capacity. He describes the real constraint as reaching what he calls the march of nines of reliability, aiming for something like 99.99999 percent reliability, and says that bar, not demand, is what limits how fast the service can grow.
SpaceX and Tesla collaboration 38:34
Asked whether SpaceX and Tesla might combine, Musk points to growing overlap between the companies, especially around a large project called Terafab, but says any such move would need a formal process and hands the question to general counsel Brandon, who notes Tesla deepened its SpaceX relationship this year through an investment and framework agreement covering projects like Terafab and Digital Optics. Musk adds that Grok is already in the car and helping with Optimus, and that Starlink connectivity will be built into Cybercab and eventually all vehicles in markets where Starlink operates, since cellular coverage is unreliable even in parts of Silicon Valley.
Why Starlink matters for robotaxis 40:31
Musk explains that a robotaxi cannot afford to lose connectivity, describing gaps in cellular coverage as bermuda triangles that would strand vehicles, which is why Starlink's anywhere connectivity is important. He adds that riders freed from driving will want to work, watch movies, or stream live sports in 4K, and Starlink offers a much lower cost per gigabyte for that than cellular data. Ashok Elluswamy ties this back to the Cybercab's design, noting its large screen and that Tesla has begun offering Cybercab rides at the Austin factory to preview the experience.
Expanding across many cities at once 42:30
Asked why Tesla is adding new cities instead of scaling up one city first, Ashok explains the goal is to prove the self driving stack is general purpose and works across many cities without heavy per city customization. He notes growth is exponential but still early, and that robotaxi vehicles rack up far more miles than personally owned cars because they operate almost continuously rather than a few hours a day. A colleague adds that spreading deployment lets Tesla work out software and operational kinks in a controlled way before pushing higher volumes, and Lars notes that transport regulation varies city by city and state by state with no single federal framework, so each city's requirements must be met individually.
Cybercab's own driving data need 44:30
Musk explains that because Cybercab is an entirely new chassis, Tesla must first accumulate driving data specific to it using retrofitted test vehicles with steering wheels and pedals, unlike Model 3 and Model Y which already have millions of miles on the road. He says the number of Cybercabs and cities will increase sharply once that calibration confidence is reached.
Federal rules and Starlink hotspot idea 46:01
Responding to a question about NHTSA administrator comments on steering wheels and pedals, Lars says Tesla has an open, cooperative relationship with the agency and has shared its plans for years, describing regulators as trying to stay ahead of a future everyone already expects. Asked whether Cybercabs could double as Starlink mobile hotspots, Musk says yes, noting the same idea could apply to stationary Starlink terminals too.
Autonomous semi truck timeline 47:59
Musk says a serious shortage of truck drivers makes an autonomous Tesla Semi important for both easing that shortage and improving safety and comfort. Because Semi volumes remain a small share of the total fleet, Tesla is prioritizing self driving progress on high volume Model 3, Model Y, and Cybercab first, expecting semi autonomy work to pick up around the end of this year or early next year and become a bigger focus in the following year.
Optimus chip and supply chain strategy 51:01
Musk says Optimus uses heavily specialized, Tesla designed power electronics currently fabricated by outside suppliers, but Optimus 4, to be built in Austin, will be far more vertically integrated, with an aspirational target of 10 million units a year compared to roughly 1 million for Optimus 3, though he cautions that scaling production is extremely difficult. On chips, he describes an upgraded AI4 reaching production around the middle of next year, AI5 also targeted for around then and headed first into Optimus, and an AI6 chip he calls potentially the best edge computing chip in the world, adding that Tesla's chip team is making very fast progress.
Capital spending philosophy 55:58
Musk says Tesla aims to spend on capital projects as fast as possible without being wasteful, favoring speed over squeezing out maximum capital efficiency because finishing sooner produces a better outcome for the company. He argues this may be the fastest industrial scale up in America since World War II, comparing it in proportion to Henry Ford's Model T era or wartime weapons production. A colleague adds that Tesla often ends up acting as its own general contractor because of the sheer volume of simultaneous construction across Optimus, Cybercab, the LFP battery factory, the semiconductor fab, and expanded solar manufacturing in the US.
Batteries for grid balancing and power 59:59
On energy storage, Musk says batteries are used for grid balancing as much as peak shaving, helping smooth out variable wind and solar generation, and predicts the solar plus battery combination will eventually generate most of the world's energy, comparing Earth to a giant satellite powered the same way as Starlink satellites. He says AI compute demand is straining power availability, with training runs causing power draw to swing by around 70 percent within 100 milliseconds, which is why SpaceX has bought many Megapacks to smooth power for its data centers. He adds that the US has roughly 1.2 to 1.3 terawatts of power generation capacity against only about half a terowatt of average usage, meaning batteries alone could roughly double the country's usable energy supply, and expects very high future demand for Megapack.
Robotaxi expansion pace explained 1:03:30
The host focuses on Ashok's comments about the slow expansion of Robotaxi, arguing this makes sense given the context of how Waymo scaled, even though Waymo started earlier and grew more slowly. He explains that Tesla first deployed Robotaxi in a supervised form, then unsupervised, and that this process generates learnings that take time to fold into the next software version. He believes version V14 was built mostly on supervised customer data, while V15 is the first version shaped by lessons from actual unsupervised testing, and Tesla is now waiting for that next iteration along with CyberCab's ramp, which Ashok said follow similarly aligned timelines.
Why new cities matter more than scale 1:07:00
He argues that expanding into new cities and jurisdictions, rather than simply multiplying vehicle counts in one city like Austin, is the smarter path right now, since each new regulatory approval makes the next one easier and builds public comfort with the service. He sees the real bottleneck as regulatory work, first responder training, and city-by-city approval rather than any technical limit on scaling fleet size, and expects Tesla to keep expanding geographically while waiting for V15 to mature before pushing hard on density in existing markets.
Optimus and FSD take rate 1:08:01
On Optimus, he notes that no analyst models expect near-term revenue from it, so expectations remain low while everyone waits to see the V3 version. He calls the FSD take rate the best part of the earnings report, with North American take rate at 55 percent, showing that people are effectively buying FSD rather than just the car, treating the vehicle as a delivery mechanism for the software. He connects this to Tesla relaxing driver monitoring somewhat, suggesting people are recognizing they do not need to watch every moment of every drive, which is boosting demand.
Margins and rising capital spending 1:10:31
He discusses Vaibhav's remarks on one-time margin benefits from last quarter versus headwinds this quarter, saying his own math showed 17.5 percent versus 16.3 percent, which he would not call flat, so more clarity may come in the 10-Q. He points out automotive cost of goods sold has held around 35,000 dollars for four or five years and will not drop meaningfully until CyberCab arrives with a different, fully autonomous business model. He also notes the market reacted to guidance that capital expenditure will stay above 25 billion dollars this year and keep rising over the next two to three years, signaling this is not a temporary increase.
Starlink integration and CyberCab miles 1:14:01
He highlights Starlink being integrated into CyberCab as a SpaceX-related detail he finds especially interesting, tying it to his broader view that once Starship is operational, SpaceX could add Starlink bandwidth at about five percent of current cost, letting CyberCabs effectively act as a mesh of cell towers. He also notes the reported 380,000 unsupervised miles with zero incidents, cautioning that some collisions, such as being rear-ended, are unavoidable regardless of the software, and interprets Tesla's real goal as avoiding incidents caused by Tesla rather than eliminating all incidents. He closes by mentioning V15 is already running in early form in Robotaxi vehicles and signs off, noting plans to cover the upcoming SpaceX earnings call.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
