Sold Out for 2027: Inside Tesla's Grid Battery Business and the Supply Chain That Feeds It

Last Wednesday, October 21, Tesla reported a quarter that most of Wall Street read as a car story. Revenue was $28.05 billion, flat from a year earlier. Operating margin was 2.6%, down from 5.8%.¹ Automotive revenue fell 3% to $20.56 billion, and the shares slid more than 3% after the bell, erasing a 3.4% gain from the regular session.²

Then, on the call, Elon Musk turned to a different business. “We're actually almost sold out of all Megapacks for 2027, and we're ramping production as quickly as possible to meet demand.” His chief financial officer, Vaibhav Taneja, went further: “the business is witnessing extreme demand from data centers, and we are doing our best to increase production.”³

Sold out a year in advance is not a phrase you hear from a car company. It is what a supplier says when its customers are lining up, and the customer here is not a driver. It is the electric grid, and increasingly the data centers that are straining it. This report follows that business: how it grew inside a company valued on cars and robots, who is buying, where the money goes when a product sells out a year ahead, and which kinds of companies tend to get the orders, with an honest section on what could go wrong.

How a Car Company Built a Grid Business

Tesla's annual report, signed on January 28 and filed with the SEC, puts numbers on the business. Energy generation and storage revenue was $12.77 billion in 2025, up 27% from $10.09 billion in 2024, which was itself up 67% from $6.04 billion in 2023. Over the same two years total revenue slipped from $96.8 billion to $94.8 billion, and automotive revenue fell from $82.4 billion to $69.5 billion.⁴ The car business shrank. The battery business doubled.

It is also the better business. Gross profit from energy generation and storage was $3.80 billion in 2025, up from $2.64 billion in 2024 and $1.14 billion in 2023, and the segment's gross margin climbed from 18.9% to 26.2% to 29.8% over those three years, helped by lower raw material costs and lower manufacturing costs for Megapack as the Shanghai Megafactory ramped, partly offset by tariffs. Tesla deployed 46.7 gigawatt-hours of storage in 2025.⁴

The product at the center of it is Megapack, which the filing describes as “an energy storage solution for commercial, industrial, utility and energy generation customers,” sold in groups that can reach gigawatt-hours of capacity. The same filing adds a line that would have looked odd in a car company's report a few years ago: “As AI infrastructure drives rapid load growth, Megapack helps to, among other things, increase utilization of existing generation and transmission capacity.” Tesla also lists the pieces of the supply chain it is pulling in-house, including a lithium refinery in Texas that began operating in January 2026, the ramps of its Megafactories in Shanghai and Lathrop, California, and the construction of a new Megafactory near Houston.⁴

That Houston plant was the story of the summer. On August 6, Tesla said its next-generation unit, Megapack 3, was “officially rolling off the line at our newest Megafactory in Brookshire, Texas,” a plant that went “from groundbreaking to operation in just 16 months.” The factory, in Waller County, cost about $200 million, is designed for 50 gigawatt-hours a year dedicated to the new product, and is expected to employ 1,500 workers by 2028. Each Megapack 3 delivers about 5 megawatt-hours, uses new 2.8-liter large-format LFP cells, and four units paired with a transformer and switchgear form a 20 megawatt-hour block. Tesla says installation time is 23 percent shorter and construction costs fall by up to 40 percent.⁵

The Quarter: Down Sixteen Percent and Nearly Sold Out

The third quarter looked worse on the surface and better underneath. Energy revenue fell 16% to $2.88 billion from $3.42 billion a year earlier, and is down 6% for the first nine months of the year. Tesla's quarterly filing lists the reasons in order: tariff refunds owed to certain Megapack customers, a lower average selling price per Megapack unit, fewer Powerwall deployments and fewer solar retrofits, partly offset by more Megapack deployments. Gross profit for the segment was $954 million, down from $1.07 billion, but gross margin rose to 33.1% from 31.4%.⁶

The tariff story behind those numbers is unusual. In February 2026 the Supreme Court invalidated certain tariffs imposed under the International Emergency Economic Powers Act, making Tesla eligible for refunds. In the third quarter it booked $526 million of those refunds as a reduction to energy cost of revenues, and $339 million as a reduction to energy revenue, because that money is contractually owed back to certain storage customers.⁶ On the call, Taneja said the margin improvement was “primarily attributed to tariff refunds recognized in the quarter and warranty true-downs,”⁷ and that the refund money “is being reinvested to expand production capacity in the US.”³

Volume tells a steadier story. Tesla deployed 8.8 gigawatt-hours in the first quarter and 13.5 in the second,⁸ then 13.7 in the third, up 10% from 12.5 a year earlier but short of the roughly 15.9 analysts expected.¹ That is 36.0 gigawatt-hours through nine months.⁶ Taneja called it the second-best quarter of deployments in the company's history, and the quarterly deck described the storage order rate and backlog as the highest Tesla has reported.⁷

Supply is being built to match. The Houston Megafactory began production in the third quarter, and in the same quarter Tesla shipped the first Megapack with LFP cells made at its Gigafactory in Nevada.⁶ That Nevada cell line is listed at 7 gigawatt-hours a year and is in early ramp.⁹ With the Texas plant's planned 50 gigawatt-hours, about 40 in California and 20 in Shanghai, Tesla's Megapack capacity works out to roughly 110 gigawatt-hours a year.⁷

✓ Trusted partner One analyst has named the small company he believes benefits most as this battery business scales, and lays out why in a free briefing →

The orders keep landing. On October 8, Taiwan's Apollo Power signed with Tesla Motors Taiwan for 260 Megapack 2XL units, roughly one gigawatt-hour, to be delivered in stages through the end of 2027, and tied the purchase to its plans for energy integration services for AI data centers. The price was not disclosed.¹⁰ And on the car side of the same earnings report, Tesla said “battery availability continued to be the limiting factor in increasing auto volumes in the quarter.”¹ The cells that constrain the cars are being pulled toward the grid, where Taneja expects “improvements in ASP and further improvements in margins as we move forward” and Musk expects gross margins “to increase meaningfully.”³

Why the Grid Is Buying

A whole year can sell out because the demand side has changed shape. The Energy Information Administration's October outlook expects U.S. power use to reach a record 4,288 billion kilowatt-hours this year and 4,356 billion next year, after a record 4,195 billion in 2025. Commercial demand, the category that includes data centers, is forecast at 1,549 billion kilowatt-hours, above last year's record of 1,493 billion, and the agency sees wholesale power averaging $52 a megawatt-hour in 2026, 11% more than in 2025.¹¹

Batteries are the fastest thing anyone can add to meet that. In February the EIA reported that developers planned 86 gigawatts of new utility-scale capacity for 2026, including 43.4 gigawatts of solar, 11.8 of wind and only 6.3 of natural gas. Battery storage accounts for 24 gigawatts of the total, after a record 15 gigawatts was added in 2025, and 53% of it, 12.9 gigawatts, is in Texas.¹²

The two biggest grids show what those batteries do. On July 22, Texas set a demand record of 91,308 megawatts, and the same day its batteries discharged a record 11,980 megawatts into the grid; a day earlier solar had set its own record of 34,665 megawatts.¹³ In California, battery capacity rose 79% to 16 gigawatts between April 2024 and April 2026, and battery discharge in the first five months of this year was roughly three times its level two years earlier.¹⁴ A battery that charges on cheap midday solar and sells into the evening peak is the simplest explanation of why Tesla's filing talks about increasing the “utilization of existing generation and transmission capacity.”⁴

Data centers are now a battery customer in their own right. About 10.3 gigawatts, or 31 gigawatt-hours, of storage came online in the first half of 2026, up more than 23% from a year earlier, and data centers accounted for around 75% of the behind-the-meter installations. The Solar Energy Industries Association expects data center demand for on-site storage to grow from about 5 gigawatt-hours this year to 24.4 gigawatt-hours by 2030. Google agreed in February to buy a 30 gigawatt-hour system from iron-air battery maker Form Energy.¹⁵

The largest AI builders are writing the checks directly. SpaceX, which absorbed xAI, disclosed in its second-quarter report that it had spent $329 million on Tesla Megapacks so far in 2026, $295 million of it in the second quarter, purchases that are likely headed for its AI data centers.¹⁶ Fluence Energy, one of the largest independent storage integrators, said in August it had secured about $850 million of data center business through July, including its first large behind-the-meter data center order, on top of a $6.4 billion backlog.¹⁷ When the customer is a hyperscaler that cannot wait for a grid connection, the battery is not an add-on. It is part of the power plant.

Where the Money Goes When a Factory Sells Out

Follow one block from order to operation and the supply chain comes into view. The cells are the largest cost. They go into units, four units go with a transformer and switchgear into a 20 megawatt-hour block,⁵ and dozens of blocks go onto a site that needs land, civil work, cabling, an interconnection agreement and the software to bid the batteries into the market. When a factory sells out a year ahead, the purchase orders do not stop at the factory gate.

When a factory sells out a year ahead, the purchase orders do not stop at the factory gate.

The electrical equipment behind each block is the tightest link. Reuters reported in July, citing Wood Mackenzie analyst Ben Boucher, that generator step-up transformer lead times surpassed 160 weeks by the first quarter of 2026, up from an average of 143 weeks in 2024, and that high-voltage circuit breaker lead times climbed to 125 weeks in the second half of last year, from 77 weeks in 2023. He expects transformer costs to rise about 4% to as much as 10% over the next year. Data centers' share of the electrical equipment market, just under 2% in 2020, could swell to 40% under Wood Mackenzie's accelerated scenario.¹⁸ A product that is sold out for 2027 at Tesla is sitting next to a transformer that, ordered today, arrives in about three years.

The connection to the grid is the other gate. On June 18, the Federal Energy Regulatory Commission issued show-cause orders to the six grid operators under its jurisdiction, PJM, MISO, SPP, CAISO, ISO-NE and NYISO, requiring each to show whether its procedures allow expedited interconnection of large loads, defined as 50 megawatts or more, and co-located resources, or to propose tariff changes, with 60 days to answer. Texas is excluded because ERCOT sits outside FERC's reach, where a separate Batch Zero process governs large-load connections.¹⁹ Every rule that speeds a data center onto the grid pulls forward the battery, transformer and switchgear orders that come with it.

Who Tends to Win: Categories, Not Names

This report does not pick a stock. It describes the categories that the current order book favors, and it leaves the names to the reader's own research.

First, the cell and materials producers with capacity outside China. Since January 1, 2026, Chinese-origin lithium-ion batteries that are not for electric vehicles carry a 25% Section 301 tariff,²⁰ and by March Benchmark reported that the increase had pushed Chinese battery imports to their lowest level since 2021, even as the U.S. storage industry remains heavily dependent on them.²¹ Tesla's own filing warns that tariffs and the provisions of the One Big Beautiful Bill Act “could significantly increase battery cell expenses.”⁴ Its answer has been its own Nevada LFP line⁶ and a Texas lithium refinery.⁴ Any producer that can sell a tariff-sheltered cell, or the lithium, iron phosphate, electrolyte and separator that go into one, is selling into a shortage.

Second, the makers of transformers, switchgear and circuit breakers, where lead times are measured in years and prices are still rising.¹⁸ Third, the builders: the engineering and construction firms, the electrical contractors and the thermal, enclosure and power electronics suppliers that turn a shipment of units into an operating plant.

✓ Trusted partner See the research behind this story →

Fourth, the developers and software operators that own interconnection queue positions and run the batteries once built; a queue position in Texas or PJM is scarce in exactly the way a transformer is scarce. And fifth, the integrators and independent operators that compete with Tesla for the same customers; Fluence's $6.4 billion backlog shows the pool is large enough for more than one winner,¹⁷ even when, as the next section shows, a quarter does not go to plan.

Small suppliers in a sold-out chain are the ones a single order can reprice. A company doing a few hundred million dollars of revenue that lands a multi-year supply agreement from a buyer the size of Tesla, a utility or a hyperscaler changes shape overnight. That is where the research effort belongs.

The Risks, and What to Watch

The first is China. SNE Research counted 461.3 gigawatt-hours of lithium-ion storage cells shipped worldwide in the first half of 2026, up 71% from a year earlier. CATL alone shipped 125 gigawatt-hours, a 27.1% share, and the top seven suppliers were all Chinese; LG Energy Solution had 2.6% and Samsung SDI 1.4%.²² Tesla's new unit, its Nevada cells and the 25% tariff are a response to that dominance, not a reversal of it. A trade truce that lowered the tariff would hand the price advantage back to the Chinese incumbents; a trade fight that raised it would raise costs for every U.S. project.

The second is that the quarter was genuinely soft. Energy revenue fell 16% and the average selling price per Megapack unit was lower,⁶ and deployments missed expectations.¹ The margin jump came primarily from one-time tariff refunds and warranty adjustments, not from pricing.⁷ Tesla's own filing warns that Megapack deployments “can vary meaningfully quarter to quarter depending on the timing of specific project milestones and logistics.”⁴ A sold-out 2027 is a statement about orders, and orders can slip.

The third is that the sector can disappoint even when demand is real. Fluence cut its fiscal 2026 revenue guidance in August to $2.9 billion to $3.1 billion from $3.2 billion to $3.6 billion, citing production issues at a new international contract manufacturing facility and construction delays at a new U.S. one, in the same release that reported the data center orders.¹⁷ Backlog is not revenue until the plant is built.

The fourth is the market itself. Tesla's shares fell after the call,² so investors have not yet decided that the batteries matter more than the cars, and the rules that speed data centers onto the grid are still being written at FERC and in Texas.¹⁹ None of this breaks the thesis. All of it argues for sizing positions as if the next quarter could be lumpy, because the filing says it can be.

Now the catalysts, dated where they can be. First, the fourth-quarter deployment number. Tesla's quarterly production and deliveries release is where the storage figure first appears; the third-quarter one, which reported 13.7 gigawatt-hours, also set the October 21 results date.²³ Last year's fourth-quarter results came on January 28,²⁴ and a date for this year's had not been announced when this report was written. The number to beat is the 46.7 gigawatt-hours Tesla deployed in all of 2025;⁴ after 36.0 through September, it needs about 10.7 in the fourth quarter to match it.⁶

The Houston ramp. The plant is designed for 50 gigawatt-hours and began production in the third quarter;^6, 5 the pace at which it climbs toward that figure, and at which the 7 gigawatt-hour Nevada cell line ramps,⁹ decides whether “almost sold out” means more revenue in 2027 or just a longer waiting list.

Grid rules. The six grid operators had 60 days from June 18 to answer FERC, so the commission's next move on large-load interconnection is the federal catalyst; in Texas, ERCOT's Batch Zero process decides which large loads connect first.¹⁹

The build versus the plan. The EIA's 24 gigawatts of planned 2026 battery additions¹² will be measured against what actually came online; the first half delivered 10.3 gigawatts.¹⁵ And the next Wood Mackenzie lead-time survey will say whether transformers are getting easier or harder to find.¹⁸

The Business That Sold Out

Put the pieces side by side. A business that did $6 billion in 2023 did $12.8 billion in 2025 at a 29.8% gross margin.⁴ Its newest factory took sixteen months to build and is planned for 50 gigawatt-hours a year.⁵ Its customers are utilities and data centers, the two buyers with the least ability to wait, and the chief executive said next year's production is nearly gone.³ Behind every one of those units is a transformer with a three-year lead time, a cell that is either tariffed or made at home, and a site that someone has to build and connect.

Investors have spent years valuing Tesla on cars and robots. The business that sells out may be the one that matters most, and the companies that feed it are where the next orders land. The smaller the supplier, the more one order can change it.