May 23, 2026 · Tesla · SpaceX · xAI · Energy Strategy

Share

Elon Musk Is Giving Up
on Earth Solar —
and Pitching Orbit Instead.

In January 2026, on a Davos stage with BlackRock CEO Larry Fink, Elon Musk made a quiet pivot that the residential solar industry has been bracing for since the 2016 SolarCity acquisition. He told Fink that space-based solar is “five times more effective” than the kind on a roof. Four months later, on May 20, 2026, SpaceX’s IPO filing made it concrete — Starship-launched orbital data centers, powered by panels Musk says belong above the atmosphere, not on it.

Tesla’s numbers tell the same story from a different angle. The Energy Generation & Storage segment posted $2.408B in Q1 2026 revenue — down 12% year-over-year. Solar Roof installations, once promised at 1,000 a week, peaked around 23 a week in 2022 and now go unreported. The Megapack utility-battery business, by contrast, hit 46.7 GWh of deployments in FY2025 and a record 39.5% segment gross margin. Two products. One growing fast. The other winding down quietly.

This is the story of a corporate strategy bifurcating in plain sight: Tesla Megapack scaling toward the grid, Tesla Solar quietly exiting the rooftop, and SpaceX repackaging the solar bet for an orbit where Musk’s own Starship economics — if they land where projected — finally make space-based photovoltaics pencil.

• −12%YoYTesla Energy segment Q1 2026 vs Q1 2025 · $2.408B vs $2.730B
• +26.6%FY2025Tesla Energy segment full-year revenue $12.8B · driven by Megapack
• 97.7%below targetSolar Roof peak ~23/wk vs Musk’s 2020 commitment of 1,000/wk
• $93.66/kg LEOStarship projected after 6 reusable flights · Falcon 9 today ~$2,720/kg

On January 23, 2026, Elon Musk joined BlackRock chief executive Larry Fink at the World Economic Forum’s annual meeting in Davos. Fink’s questions tracked the usual investor agenda: AI compute demand, energy supply, capital allocation across Tesla, SpaceX, and xAI. The answer that landed hardest in the renewable-energy press — and that TechCrunch’s Tim De Chant returned to in his May 23 piece — came when Fink steered the conversation toward solar.

“When you have solar in space you get five times more effectiveness, maybe even more than that, than solar on the ground.”

Elon Musk to BlackRock CEO Larry Fink · World Economic Forum · January 23, 2026

Musk framed the case in three steps. First, terrestrial solar suffers a duty-cycle problem: the sun does not shine at night, weather degrades irradiance, and dust degrades panel performance. Second, geostationary orbit gives a panel near-continuous sunlight at full intensity. Third — and this is the part that depends on him — the only thing keeping space-based solar out of the cost curve is launch. Bring launch costs to $100 per kilogram or below, and a kilowatt put into orbit could outcompete a kilowatt installed on a Houston rooftop. He told Fink that AI data centers in space could be operating “in two to three years.”

Two and three years is an aggressive number. But the structural argument — that terrestrial photovoltaics hit a hard ceiling somewhere between the duty cycle and the cost of land — is one the academic space-solar community has been making in peer-reviewed work for two decades. What is new is that the world’s most-funded launch provider is now making it in front of the world’s largest asset manager, on a public stage, with a follow-on IPO filing four months later.

Strategy statements get tested in 10-Q filings. Tesla’s Q1 2026 Form 10-Q, filed with the SEC for the quarter ending March 31, 2026, reported Energy Generation & Storage segment revenue of $2.408 billion — down 12% from $2.730 billion in Q1 2025. Energy-storage deployments came in at 8.8 GWh, a 38% sequential decline from the prior quarter. Segment gross margin nevertheless set a record at 39.5%, reflecting a continued mix shift toward higher-margin Megapack hardware and away from lower-margin residential installation work.

The full-year 2025 picture, by contrast, was strong. Tesla’s Q4 2025 update reported $12.8 billion in Energy segment revenue, up 26.6% year-over-year. Energy-storage deployments hit 46.7 GWh, up 49%. Megapack 3 production has begun at the new Megafactory Houston, with stated capacity up to 50 GWh per year — on top of the existing Lathrop, California facility and the Shanghai plant that came online in 2025. Tesla’s global Megapack manufacturing capacity is approaching 133 GWh per year.

What the Q1 2026 print actually tells you is that the Megapack ramp is lumpy — large utility projects ship in batches — while the residential side keeps falling. Solar Roof did not appear as a discrete line item in the Q1 2026 release. It has not appeared as a discrete line item since Q1 2024, when Tesla removed the per-period solar deployment figure from the quarterly update. Electrek’s Fred Lambert reported on May 14, 2026 that the program is “on life support,” with Tesla quietly pivoting back to conventional panels for the residential customers still in the pipeline.

In October 2019 Musk unveiled Solar Roof v3 at Tesla’s Hawthorne studio. By 2020, in earnings-call commentary aimed at shareholders, he was pledging that Tesla would be installing 1,000 Solar Roofs per week — first by year-end 2019, then by mid-2020, then with the date increasingly vague. Independent measurement was difficult because Tesla never broke Solar Roof installations out separately. Industry trackers and Wood Mackenzie residential analysts, working from utility interconnection filings and Tesla’s legacy SolarCity database, estimated the peak weekly run-rate in Q2 2022 at around 23 installations per week— about 2.3% of the public commitment.

The 1,000-per-week number was never plausible at the unit economics Tesla was carrying. Solar Roof tiles cost more to manufacture than conventional shingles plus a separate roof-mounted PV array, and the installation labor required a specialized crew — in many markets, a crew Tesla had to train itself. The legacy SolarCity panel business, the part of the 2016 acquisition that was running at scale, contracted year over year through the high-interest-rate environment of 2023 and 2024. By April 2024, Tesla announced workforce reductions exceeding 10%, and Drew Baglino, the Senior Vice President for Powertrain & Energy Engineering who had run the segment since October 2019, resigned on the same day.

From Q1 2024 forward, Tesla stopped giving the market a clean look at Solar Roof or residential PV unit volumes. Investor presentations focus on energy storage. The TechCrunch May 23 piece is blunt: the rooftop bet that justified the SolarCity acquisition has been quietly walked back.

The roots run back to November 2016, when Tesla acquired SolarCity in an all-stock deal valued at approximately $2.6 billion. SolarCity had been founded a decade earlier, in 2006, by Musk’s cousins Lyndon Rive and Peter Rive. Musk himself was the company’s chairman and largest shareholder. At the time of the acquisition, SolarCity was reportedly running deeply negative free cash flow and faced an upcoming debt wall it could not easily refinance.

Tesla shareholders sued, arguing the merger amounted to a Musk-controlled bailout of a failing entity at the expense of Tesla’s minority shareholders. The case worked through Delaware’s Court of Chancery and ultimately the Delaware Supreme Court. In April 2022, Vice Chancellor Joseph Slights III ruled for Musk on the merits. In June 2023, the Delaware Supreme Court affirmed, finding the acquisition was “entirely fair” under Delaware’s heightened fairness standard for transactions involving a controlling stockholder.

That legal vindication is one fact. The business reality is a separate one. Ten years on, the integrated rooftop business Tesla bought has shrunk; the new product Tesla built on top of it — Solar Roof — never reached three percent of the run-rate Musk publicly promised in 2020; and Tesla’s 2023 Master Plan Part 3— which described the company’s overarching purpose as helping to “expedite the move from a mine-and-burn hydrocarbon economy towards a solar electric economy” — reads, in 2026, like a document about a different company. The current company is selling Megapacks to gas-fired and nuclear-firmed data centers and pitching orbital photovoltaics to the IPO road show.

“The overarching purpose of Tesla motors…is to help expedite the move from a mine-and-burn hydrocarbon economy towards a solar electric economy.”

Tesla Master Plan Part 3 · 2023 · quoted in TechCrunch, May 23, 2026

One reason the orbital pitch lands the way it does is that the on-Earth fact set runs in the opposite direction. TechCrunch’s May 23 article documents that xAI — the AI company Musk founded in 2023 to compete with OpenAI and Anthropic — is currently operating dozens of natural-gas turbines at its Memphis Colossus data center, and has planned an additional $2.8 billion in fossil-fuel infrastructure to power the next phase of training. Tesla Megapacks back up the load — xAI’s spend on Megapack hardware over two years runs around $697 million— but the energy itself, at the scale Colossus demands, is overwhelmingly gas-fired.

The Memphis grid did not have spare interconnect capacity sized for a hyperscale AI build. The turbines are how xAI bridged the gap. They are also exactly the kind of fossil infrastructure that Musk, fifteen years earlier, argued Tesla existed to retire. TechCrunch’s framing is direct: Musk is, at this moment, the chief executive of one of the largest new natural-gas power consumers in North America and the largest financial backer of a launch company whose investor pitch deck argues that the future of AI compute belongs in space, powered by panels in geostationary orbit.

On May 20, 2026, SpaceX filed confidentially with the U.S. Securities and Exchange Commission for what is expected to be the largest initial public offering in American history. According to early reporting in The Wall Street Journal and Reuters, the registration statement leans heavily on three pitches: (1) a Starlink consumer-broadband business now generating recurring revenue at scale, (2) Starship as the launch vehicle that finally collapses the per-kilogram cost curve, and (3) a forward-looking line item that did not exist in any prior Musk-company pitch deck — orbital data centers, powered by space-based photovoltaics.

The filing does not commit to a delivery date for the orbital-compute constellation. It does not have to. The bet investors are being asked to size is launch economics: if SpaceX makes Starship fully reusable at the cadence Musk has publicly promised, the per-kilogram cost of putting silicon — whether photovoltaic, computational, or both — into low Earth orbit collapses by roughly two orders of magnitude compared to Falcon 9, and by roughly three compared to expendable launch. Every other line item in the orbital-compute business case keys off that one number.

“Hopefully this year we should prove full reusability for Starship, which would be a profound invention, because the cost of access to space would drop by a factor of 100 when you achieve full reusability.”

Elon Musk · World Economic Forum · January 23, 2026

SpaceX has separately filed with the Federal Communications Commission for a next-generation Starlink architecture that contemplates up to one million satellites in low Earth orbit, laser-linked and solar-powered — the same orbital constellation that, in the All-In Podcast discussion in early January 2026, was framed by Chamath Palihapitiya and David Sacks as the eventual home of frontier-model AI training. The investor pitch is now coherent across three properties: Tesla Megapack firms terrestrial AI loads where they have to live; xAI runs the actual models on gas-firmed compute today; SpaceX moves the next generation of compute, and the photovoltaics that power it, to orbit.

Musk has been pre-positioning the orbital-compute narrative on X for the better part of a year. The May 12, 2025 post about a “terawatt of compute” lifted on ten Starships frames the launch-cost argument as a literal AI bottleneck, not an abstract space-policy debate.

Musk is not the only actor turning away from terrestrial residential PV. The federal policy environment did the same thing on a different timeline. The residential Clean Energy Credit under Internal Revenue Code Section 25D — the 30% federal investment tax credit homeowners use to defray the cost of a rooftop solar system — was scheduled to sunset on December 31, 2025 under the 2024 reconciliation package. In parallel, the White House moved aggressively on nuclear: executive orders in May 2025 accelerated advanced-reactor licensing, and a December 2025 order directed agencies to assess lunar and orbital reactor research programs.

The residential solar tax credit sunsetting at year-end 2025, combined with high mortgage rates, structurally lengthened the payback period on a rooftop installation in most U.S. markets. SunPower — once the second-largest American residential solar installer after Sunrun — filed for Chapter 11 in Delaware on August 5, 2024. Complete Solaria acquired the SunPower estate’s assets for $45 million in September 2024 and rebranded the combined entity as SunPower in April 2025. Sunrun reported Q1 2026 revenue of $722.2 million, up 43% year-over-year — market share concentration, not market growth.

Musk’s “five times more effective” line is not, in itself, a controversial claim. The duty-cycle advantage of geostationary orbit over the Earth’s surface has been well-characterized since at least the 1970s, when Peter Glaser proposed the first detailed space-solar-power architecture. What matters is whether the engineering closes — whether the panels, the wireless power transmission, the structural assembly, and the launch all pencil at the same time.

Caltech’s Space Solar Power Project, led by Principal Investigator Harry Atwater (Howard Hughes Professor of Applied Physics and Materials Science) with co-PIs Ali Hajimiri (electrical engineering — wireless power) and Sergio Pellegrino (aerospace — deployable structures), demonstrated the first-ever wireless power transmission from space to Earth in 2023, via the MAPLE experiment aboard the Space Solar Power Demonstrator 1 (SSPD-1). DOLCE, the deployable structure prototype on the same platform, demonstrated that ultra-lightweight modular reflector geometries can be unfolded in orbit. The science is real and is moving.

The European Space Agency reached a more cautious conclusion. The SOLARIS programme, led by Sanjay Vijendran, ran a multi-year feasibility study into a European SBSP demonstration. In August 2024, ESA concluded that space-based solar power was “not yet mature enough to advance to a demonstration mission” on the SOLARIS timetable and shelved the next-phase commitment. Vijendran subsequently departed ESA to co-found TerraSpark, a venture-backed orbital-prototype startup that closed more than €5 million in pre-seed funding and targets a flight-demonstration vehicle in 2027.

“Space-based solar power is not yet mature enough to advance to a demonstration mission.”

European Space Agency · SOLARIS programme conclusion · August 2024

Read together, the Caltech and ESA records describe SBSP as a technology where the physics has been proven, the launch-cost barrier is the binding constraint, and the engineering integration risk is large but bounded. None of that is in dispute. What Musk added on the Davos stage was a commercial timeline — “two to three years” for AI data centers in space — and a launch vehicle his company is actively building. Whether Starship hits its full-reusability cadence on his timeline is a different question. The IPO filing, in effect, asks public-market investors to underwrite the bet that it will.