← All videos

Could Elon Musk's " One " Company Control Energy, AI and Robots?

Nnutshell | AI, Tech & Business Published Nov 16, 2026 Added 2d ago 14:11 7 views Open on YouTube ↗

Description

What if Elon Musk combined Tesla and SpaceX into one company that could build everything from solar power to intelligent robots? This is the case for a possible SpaceX–Tesla merger—and the biggest obstacles standing in its way.

━━━━━━━━━━━━

▶️ WATCH MORE FROM NNUTSHELL

Cinematic explainers about AI, technology and the companies shaping our future:

https://www.youtube.com/@nnutshellhq/videos

📚 READ NNUTSHELL

Big ideas, in a nutshell.

https://nnutshell.com/

━━━━━━━━━━━━

🎬 THE STORY

SpaceX and Tesla are separate companies. Their boards have formed committees to consider a potential merger, but no deal is agreed. So what would combining them actually create?

This video traces seven layers of a possible Musk technology stack: solar energy, Starship launches, Starlink connectivity, Terafab chips, Colossus computing, xAI intelligence and Tesla Optimus robots.

We examine the opportunities—and the weak links. Can orbital data centers become economical? Can Starship scale? When might Ter

Transcript

Read auto-generated transcript (2582 words)

Kind: captions Language: en Tesla and SpaceX aimed to build the largest chip fab ever. Tesla wrote that aim into its April report for a fab not yet built. Today, SpaceX and Tesla are separate companies. So what follows is my case for what they would build as one. Their boards have each formed special committees to review a possible merger. On October 20, Elon Musk, CEO of both, said the committees have not reached agreement on the terms. A merger may never happen. If you hold Tesla shares, you own part of SpaceX's partner in that fab. If a merger is ever put forward, Musk says it would need the support of 50% of all outstanding Tesla shares, not just those that vote. So what company would SpaceX and Tesla make together? In my framing, a merger would hold seven layers, sunlight to robot. At the base sits a solar target of 100 gigawatts a year, which Musk says both companies are working toward. Above it sits Starship for launch and Starlink for the network. Then Terra fab for chips and Colossus data centers for compute. At the top, xAI for intelligence and Tesla's Optimus robot. SpaceX already owns Starship and Starlink. It also owns Colossus and xAI. Tesla's own layer is the robot and the two partner on the fab. So most of the stack already sits inside SpaceX. Money already moves both ways between them. Tesla's April report records $2 billion of cash for a SpaceX equity investment. SpaceX's quarterly filing, as Benzinga reported it, shows the reverse. SpaceX bought $480 million of mega pack batteries from Tesla in 2026 through September. So Tesla cash went into SpaceX shares and Tesla batteries came back between two companies Musk runs. Batteries store power and power is where Musk's own case begins. But why would a chain that ends in robots start with sunlight? In February 2026, Musk put his case plainly. The availability of energy is issue. On the Dwarkesh podcast that month, he said the gas turbines new power plants need are sold out through 2030. So the argument turns to the sun. Earth, Musk notes, receives half a billionth of the sun's energy. In his framing, the sun is the only source big enough to scale. That puts sunlight at the bottom of the stack. The first test is on Earth at Colossus II. SpaceX's prospectus defines it as its data centers in Memphis and South Haven, Mississippi. On April 14, 2026, the NAACP sued XAI over the power there. Its complaint alleges XAI is illegally operating 27 gas turbines without an air permit in South Haven. The suit is still pending. So if the NAACP's allegation holds, part of that campus runs on gas, not sunlight. The sharpest objection came from the podcast host Dwarkesh Patel. By his own figure, energy is only 10 to 15% of the total cost of ownership of a data center. If power is that small a slice, cheaper sunlight saves little. But Musk's word was availability, not price. My reply is that if turbines are sold out through 2030, a turbine you cannot buy powers nothing at any price. And Musk's fix puts the power in orbit, which means something has to lift it there. Here is the scale of that lift. SpaceX's prospectus sets a goal of deploying 100 gigawatts of compute power to orbit every year. It attaches no date and no cost. SpaceX's FCC filing says each ton allocated to computing carries on the order of 100 kilowatts. Now divide the gigawatt goal by the kilowatts each ton carries. Orbit would need about a million tons lifted a year. SpaceX's own results table shows what it lifts today. In the first nine months of 2026, it put about 1,500 metric tons into orbit. The goal asks for about a million tons every year. That is about 660 times what SpaceX lifted in those nine months. Closing that gap would take a rocket built for a different scale. The rocket meant to lift that load is Starship. SpaceX's June 2026 prospectus says Starship V3 is designed to deliver 100 metric tons to Earth's orbit fully reusable. Full reuse has not been demonstrated yet. Then in the third quarter of 2026, SpaceX completed Starship flight 14. It was the company's first orbital Starship flight. But Musk himself names the catch. Asked on the Dwarkesh Podcast for Starship's single biggest remaining problem, he answered is having the heat shield be reusable. And SpaceX's prospectus warns investors that its growth strategy is highly dependent on the successful development and scaling of Starship. So SpaceX's growth plan rides on a rocket whose biggest remaining problem, by Musk's answer, is a reusable heat shield. The network layer is Starlink, about 9,600 broadband and mobile satellites in low Earth orbit by SpaceX's prospectus count. On January 30, 2026, SpaceX applied to the FCC for permission to launch as many as 1 million satellites. That is an application, not an approval. It asks for about 104 times today's fleet. In the filing's words, the satellites would operate as orbital data centers. So in this plan, orbit would carry the data and compute it. Starlink already earns. In the third quarter of 2026, SpaceX's connectivity revenue was $5.2 billion. In that same quarter, SpaceX put $16.4 billion into AI capital spending. That is about a third of what it is investing in AI, money spread over years. My case is that Starlink alone cannot fund this climb. In orbit, the prize is sunlight and the bill is the launch. Google Research predicted that launch prices may fall to less than $200 a kilogram by the mid 2030s. On Google's analysis, that is where space compute starts to pencil out. So for now, SpaceX builds its compute on the ground in Memphis and South Haven. Back on Earth, power is still the issue, and Musk says turbines are sold out through 2030. Ground compute also needs chips, and Musk's answer is a fab that makes none yet. Musk launched Terafab, the chips layer, on March 21, 2026 at the Sea Home power plant in Austin. He said, we either build the Terafab or we don't have the chips. Tesla's stated goal posted on X is 1 terawatt of compute a year, but SpaceX's prospectus calls Terafab a chip manufacturing initiative with Tesla and Intel. On Tesla's October call, Musk said the Austin research fab begins commissioning machinery in the first half of 2027. No chips are made there yet. Electrek in an opinion piece argued Intel will build the fab for Tesla. So even merged, SpaceX and Tesla would build this fab with Intel. The compute layer is the Colossus data centers. SpaceX measures them in nameplate compute, the installed capacity of its chips, not the power they draw. On September 30, 2025, nameplate compute stood at 0.6 gigawatts. On September 30, 2026, it was 1.7 gigawatts, nearly triple in a single year. So who is all that capacity for? One answer is Anthropic, an AI lab that rents compute capacity from SpaceX. SpaceX's prospectus says that capacity includes approximately 325,000 Nvidia GPUs. On that reading, SpaceX's ground compute works like a landlord business. SpaceX is the landlord and Anthropic is the tenant. SpaceX's June prospectus puts a price on that rent, and it looks long. Anthropic has agreed to pay SpaceX $1.25 billion per month through May 2029. But the same filing says the agreements may be terminated by either party upon 90 days notice. Musk, posting on X in May, described the deal in his own words. The agreement is a 180-day lease with a mutual 90-day cancellation notice thereafter. So the contract that looks long is a short lease. SpaceX rents out that capacity while still permitting reallocation of that capacity for its own internal initiatives. Its prospectus calls this a dual monetization strategy, renting compute out or using it itself. My case is that the gigawatts are built for SpaceX first with rent on the side. So either side can end Anthropic's lease on 90 days notice and SpaceX may move that capacity to its own projects. If SpaceX takes that capacity back, what would it use it for? For its own AI first, SpaceX acquired xAI in February 2026 and it reports an AI segment it calls superintelligence. In the third quarter, that segment brought in $7.6 billion against $5.2 billion for Starlink. So AI now outsells Starlink inside SpaceX. That quarter, SpaceX put $16.4 billion into AI capital spending and investments spread over years. That is more than twice the AI revenue. And for its own machines, the record is thinner. Electric reports that Musk had predicted 10,000 Optimus robots for 2025, a target Tesla missed entirely. In January 2026, Electric reports Musk admitted that zero Optimus robots were doing useful work in Tesla's factories. So the robot layer starts behind its own schedule. Then Tesla began changing a factory. Its April report said the first-generation Optimus line will replace the Model S and Model X lines in Fremont. That line is designed for 1 million robots a year. Musk gave an update on Tesla's October earnings call. In his words, we started what I'd call early production on our high-volume line in Fremont. So in Fremont, a car line is becoming a robot line. What links that robot line to SpaceX's AI? A single statement from Musk. On the Dwarkesh Podcast in February 2026, he said Grok would orchestrate the behavior of the Optimus robots. On Tesla's October call, he added, I remain convinced that Optimus will be the biggest product on Earth. Fremont has only just begun early production and xAI was acquired by SpaceX while Optimus is built by Tesla. So Musk's link runs across a company line. His case for joining them in his words is that we are pretty much in the super intelligent singularity. So who could approve a merger? Not Musk. On SpaceX's October 20 call, he said there's a lot of merit to combining the companies. But the talking happens between two special committees, one for each board. In his words, it's an odd kabuki dance where two groups of people are talking and I can't be in the conversation. >> A hurdle sits with Tesla shareholders. On the same call, Musk said a Tesla merger would need the support of 50% of all outstanding shares, not just those that vote. So every Tesla share that stays home counts as a no. Tesla's April report says a second Optimus line in Texas is designed for 10 million robots a year long term. And the special committees have not reached agreement on the terms of any transaction, Musk added. So holders have no exchange ratio, the swap rate between shares to weigh. >> Musk says combining has merit. And SpaceX has already absorbed one Musk company. It acquired XAI effective February 2, 2026. But that precedent does not carry over to Tesla. SpaceX's prospectus says that after its listing, Musk will hold about 82.4% of its votes. At Tesla, the decision belongs to every outstanding share, not to Musk. So I can see this objection because without Tesla's holders, SpaceX's AI and Tesla's robots stay in separate companies. >> A merged company would also inherit the orbit bet and SpaceX hedges that bet itself. Its June prospectus warns its orbital data centers may not achieve commercial viability. It expects the first orbital compute satellites as early as 2028. In an August 24 post, Musk set a nearer target. SpaceX and Nvidia, he wrote, had designed a Vera Rubin system for orbit in the fourth quarter of 2027. So the filing doubts orbit will pay or Musk pulls the date forward. Suppose as a what if half of all outstanding Tesla shares said yes, then Grok and Optimus would sit under one owner and reclaimed compute could reach robots without crossing a company line. Suppose instead too few said yes, then SpaceX would stand outside customer for Tesla's batteries. SpaceX's 10Q as Benzinga reported shows $480 million of megapacks bought from Tesla in 2026 through September. And on Tesla's October call, Musk said we're actually almost sold out of all megapacks for 2027. So in that what if SpaceX would keep bidding for batteries, Musk says are almost sold out. For the next merger headline, ask three things. Who owns each layer? Who can take capacity back? Who must vote yes? Here SpaceX owns most of the layers and keeps the take back right. The vote belongs to half of all outstanding Tesla shares. My case is that a merged SpaceX and Tesla could be a closed loop from sunlight to robot. It would aim to supply its own AI and Optimus robots with chips from a fab it builds with Intel. Anthropic's lease can end on 90 days notice and SpaceX's filing keeps the right to reallocate that capacity. No merger terms are agreed and the loop exists only if half of all outstanding Tesla shares say yes. I'd change my mind if SpaceX signed long compute leases, it could not cancel. That would make it a landlord after all. Tesla aims to build the largest chip fab ever with SpaceX as its partner. Merged, they would be one company from sunlight to robot. And SpaceX's filing already states the aim at the bottom of that stack to harness the sun to power a truth-seeking artificial intelligence.