Musk Lifts Tesla Stake to 20% After SpaceX Listing, Clearing Path for Historic Merger

A June 2026 options exercise, a public SpaceX, and an xAI combination already done leave Tesla as the last piece of a roughly $4 trillion ecosystem deal.

Two-thirds of the Musk industrial stack already sits under one roof. SpaceX is public, xAI is inside it, and Tesla is the only major piece still sitting outside. The question is no longer whether a full combination makes industrial sense. It is whether Tesla’s outside shareholders, Delaware courts, and a fair-exchange price can survive the hardest related-party deal in modern markets.

Key Takeaways

• On June 17, 2026, Musk converted about 304 million shares from his 2018 CEO options at $23, lifting Tesla voting power from roughly 13% to 20%.
• He now holds around 700 million Tesla shares - about a fifth of the company - after selling 17 million shares for roughly $7 billion to cover strike cost and tax.
• SpaceX began public trading on June 12, 2026; five days later the Tesla stake move hit, and Motley Fool openly floated a SpaceX merger next.
• Tesla’s November 2025 pay package, approved with over 75% support, targets up to about 424 million shares and a path toward roughly 25% voting control if milestones hit.
• SpaceX priced a first investment-grade bond in June 2026: $25 billion raised against a roughly $90 billion order book, coupons from about 5.35% to 6.65%.
• Committed outside compute revenue at SpaceX tops $80 billion through 2029, including Google near $920 million a month and Anthropic around $45 billion over multi-year terms.
• TerraFab, the Tesla-SpaceX-xAI chip JV near Austin, saw all-in cost talk climb from about $25 billion in March 2026 toward as much as $120 billion by May.
• Tesla’s $2 billion January 2026 investment in xAI is already in litigation, which is why the private Nevada path for SpaceX-xAI came first and the public Tesla leg last.
• A combined Tesla-SpaceX-xAI entity would land near $4 trillion on paper, with valuation split between Tesla and SpaceX holders as the real fight.

---

The Stake Move That Unlocked the Table

On June 17, 2026, Musk exercised options from the old 2018 Tesla CEO award - roughly 304 million shares at $23 - and converted them into restricted stock with immediate voting rights. That single step took his Tesla voting power from about 13% to 20%. He sits on around 700 million shares. Paying for it was not free: strike price plus tax, funded by selling 17 million shares for about $7 billion. The converted shares stay locked until January 2028.

SpaceX had only started trading as a public company five days earlier, on June 12. Mainstream finance outlets immediately connected the dots. I do not treat that sequence as coincidence. Voting control is the precondition for any cross-company deal that has to clear both boards and, eventually, courts.

Why Twenty Percent Still Is Not Enough

Tesla shareholders already approved a new pay package in November 2025 with more than 75% support. Twelve tranches, about 424 million shares, worth up to roughly a trillion dollars over a decade if the milestones land. Those milestones are extreme: an $8.5 trillion market-cap target, 20 million vehicles, a million Optimus robots in the field. As tranches vest, ownership walks toward about 25% voting control.

That 25% figure is intentional. In January 2024 Musk said he wanted enough influence to steer strategy without enough control that he could never be overturned. On the SpaceX side he already holds something like 82% to 85% of the vote through super-voting shares. In any Tesla-SpaceX transaction he sits on both sides of the table. More votes help pass a deal. They also tighten the legal standard after the fact, because a controlling insider on both sides invites the hardest fairness review courts run.

SolarCity Proved the Gauntlet. Nevada Dodged It.

We have seen this structure before. In 2016 Tesla bought SolarCity for about $2.6 billion with Musk on both sides. That deal went to a full Delaware trial, survived an entirely-fair ruling in 2022, and was affirmed in 2023. Survival required a genuinely independent board process, a minority shareholder vote that excluded Musk’s own shares, and a price that could be defended on the merits. That is a brutal gauntlet.

The SpaceX-xAI combination walked around most of it by staying private and incorporating in Nevada, where the bar is closer to known misconduct than Delaware’s entire-fairness standard. Tesla is public, heavily owned by outside shareholders, and sued constantly. Those same shareholders are already in court over Tesla’s $2 billion investment into xAI from January 2026. Friction decided the order: merge the quiet private pieces first, save the lawsuit-prone public one for last.

SpaceX Becomes the Balance Sheet

In June 2026 SpaceX priced its first investment-grade bond. The company targeted $20 billion, upsized to $25 billion, and saw an order book around $90 billion - oversubscribed by roughly three and a half times. Five tranches, coupons from about 5.35% up to 6.65%, maturities stretching to 2056, interest cost near $1.5 billion a year. Lead banks were the usual giants: Bank of America, Citi, Goldman, JPMorgan, Morgan Stanley. Ratings came back investment-grade across the board.

The money refinances expensive debt left from the 2022 Twitter acquisition and costly xAI loans from 2025, plus general corporate use. SpaceX also disclosed a cash pile of about $100 billion. Ecosystem debt now routes through a public, investment-grade SpaceX balance sheet. Any Tesla combination would lean on that foundation, not the other way around.

Colossus Turns Into a Neocloud Landlord

The other inheritance was compute. After the merger, SpaceX owns Colossus - the cluster that trained Grok - and flipped it into a commercial multi-tenant cloud, what the industry calls a neocloud. In June 2026 SpaceX signed Reflection AI to rent compute at $150 million a month, about $6.3 billion total through the end of 2029. Reflection is an open-source AI startup founded by ex-DeepMind people, NVIDIA-backed, pitched as an American answer to DeepSeek.

That sits on top of Anthropic and Google deals worth multiple tens of billions. Google is in at roughly $920 million a month. Anthropic signed for around $45 billion over multiple years. Committed outside compute revenue now tops $80 billion through 2029. That backlog helps justify SpaceX’s roughly $1.8 trillion valuation and helps service the new bond debt. It also makes SpaceX a credible AI infrastructure partner for the two things Tesla burns compute on hardest: autonomy and Optimus.

TerraFab, AI5, and the Walls Already Showing Cracks

TerraFab, announced in March 2026 near Austin, is reported as a joint venture across Tesla, SpaceX, and xAI. The three-way structure itself is a tell: corporate walls already block simply building the plant as one project. Cost estimates climbed from about $25 billion initially to as much as $120 billion all-in across phases by May 2026. The fab is supposed to produce custom chips for FSD, Cybercab, and Optimus, plus radiation-hardened silicon for orbital data centers.

Tesla has lived this movie with Dojo. The in-house training supercomputer was killed in August 2025 when leadership said all paths converge on AI6 and called Dojo 2 an evolutionary dead end. The program lead and about 20 staff walked. Months later the effort was revived. Tesla still targets small-batch AI5 by late 2026 and volume in 2027. Morgan Stanley projects meaningful output closer to 2028-2030. AI5’s design taped out in April 2026. Tesla claims inference roughly 10 times cheaper than NVIDIA hardware. Megapod, a June 2026 Tesla trademark, points at containerized AI data centers powered by Megapack batteries. Grok already ships in millions of Tesla cars as of July 2025 and is the conversational brain for Optimus Gen 3. StarMind, SpaceX’s orbital inference layer, had its first satellite unveiled in June 2026. Four legal entities. One silicon, model, and data-center stack already plugged together.

Inference Factory Meets Inference Consumer

Step back and the industrial logic is simple. Tesla acts on intelligence: FSD turns inference into steering, Optimus into hands on a box, robotaxi into a left turn through traffic. Every car and robot is an edge node that spends a finished model as physical motion. SpaceX and xAI produce intelligence: Colossus trains Grok, StarMind runs inference in orbit, Starlink beams it down, Starship drops the cost of putting raw compute anywhere.

When those costs sit in separate companies, every unit of intelligence that crosses the boundary gets taxed - IP licensing, related-party reviews, duplicate overhead, lawyers on both sides. Put them in one entity and that friction collapses. One trained model amortizes across robotaxis, millions of Optimus units, satellites, and outside neocloud customers. One custom line from AI5 into AI6. One energy stack with Megapack and solar powering both the factories that make intelligence and the fleets that burn it. Apple’s chip-OS-device stack, extended until the device is a car, a humanoid, and a satellite.

I have watched this constellation since 2012. The industrial case for combination is the strongest I have seen in that stretch. Inevitable for the product stack still does not mean fair for every shareholder class.

The $4 Trillion Fight Over Price and Courts

Combine the pieces and you get a roughly $4 trillion entity. Tesla’s GAAP profit is positive but small relative to that whole. SpaceX investors and Tesla investors - many people own both - will fight over relative value. Higher Tesla relative value helps Tesla holders. Lower helps SpaceX holders. Given voting power and execution trust, the clearing price will likely be close to what Musk decides is right.

That does not end the story. Winning a shareholder vote and surviving post-deal litigation are different problems. SolarCity shows survival is possible when process and price hold up. Tesla’s public float, constant litigation, and the existing xAI investment lawsuit raise the bar. I still think the combination gets done. Too many shared chips, models, fabs, energy assets, and compute contracts only make sense inside one P&L. When it closes, it will be one of the largest industrial mergers ever attempted - and one of the most legally contested.