Musk Accelerates "Super Intelligence" Ambition: SpaceX (SPCX.US) and Tesla (TSLA.US) Discuss $3.5 Trillion Merger, ASML Rushes Delivery for Terafab

According to Zhitong Finance APP, after the U.S. stock market closed on Tuesday, Musk tied together two originally separate threads during SpaceX's (SPCX.US) third quarter earnings call. Both pieces of news point to the same thing: on the AI computing power supply chain, Musk is simultaneously tackling two problems — "unable to manufacture" and "cannot merge".

One is about chips: he said that ASML (ASML.US) and other key equipment suppliers have begun accelerating their equipment deliveries for Tesla's (TSLA.US) and SpaceX's Terafab project. "These timelines are very aggressive, and we greatly appreciate ASML and all our other suppliers for expediting equipment delivery to achieve this outcome." The other is capital operation: Both SpaceX and Tesla have each established a special committee composed of independent directors to explore the possibility of a merger—this is the most complete public statement about the potential deal so far.

Terafab Timeline: Test wafers out by May 2027, mass production in Q3 2028

This time, Musk gave the first set of specific internal milestones for Terafab, and proactively added a caveat—"the margin for error is large." According to him, the goal of the research-oriented fab in Austin is to successfully run the first test chip, "which is similar to an SRAM chip," to verify the production line by May or June next year; the next goal is to achieve mass production at the main Gibbons Creek plant in Q3 2028.

Placing these two milestones back in the context of the project itself reveals their significance. Terafab was officially launched in Austin on March 21, 2026, pushed jointly by SpaceX, Tesla, and xAI, aiming for an annual output of 1 terawatt (1 TW) of AI computing power—SpaceX stated in the project announcement that "the combined chip demand of SpaceX and Tesla is expected to exceed 1 TW, significantly greater than current global supply." The plant is located on the site of the former Gibbons Creek coal-fired power plant in Grimes County, Texas, covering more than 6,000 acres, about 90 miles northwest of Houston; a groundbreaking ceremony was held on October 6, land clearing began in August, formal construction was set for December 1, and full completion is scheduled by the end of 2028.

The investment structure is also stratified: Phase one investment totals $16.8 billion, as confirmed in SpaceX's August 6 announcement and by Texas Governor Abbott. According to incentive application documents submitted to Texas, the total value of the project could reach as high as $119 billion upon completion. The Texas Enterprise Fund is providing a $30 million subsidy for phase one, and in June the County Commissioners Court approved a 100% property tax exemption by a vote of 4 to 1.

The real manufacturing capability is not in Musk's hands. Intel (INTC.US) joined the project on April 7, promising to cover design, manufacturing, and packaging across the full vertical chain. Terafab mass production will adopt Intel’s 14A process. On October 2, the market rumored that TSMC (TSM.US) might participate, but on October 7, Musk clarified on X — "We will build and operate this factory ourselves," and TSMC can at most lease a portion. Analyst Adrien Sánchez of research firm Yole Group summarized it more bluntly: Terafab "is more like an expansion of Intel’s fab, with Tesla, SpaceX, and xAI as anchor customers."

ASML's Role: The Only Bottleneck and the Most Direct Beneficiary

Musk's specific thanks pointed to a company whose capacity is already booked out to 2030.

On October 14, ASML announced record Q3 results: net sales of €11.662 billion, up 25.0% quarter-on-quarter and 55.2% year-on-year, but slightly below Wall Street’s average estimate of €11.735 billion; full-year net sales guidance was raised to €44–45 billion, with gross margin guidance increased to about 56%. CEO Christophe Fouquet had already confirmed direct communication with Musk on Terafab, describing Musk as "very serious," while adding the caveat—only if ASML itself is not the supply bottleneck does this project constitute an opportunity, and the company has not yet met this condition for existing customers.

However, equipment may arrive before the fab itself is ready. According to chip industry observer Tim Culpan, citing informed sources, the first ASML EUV lithography machine for Terafab may be delivered by mid-next year; but according to SemiAnalysis, the main structure of the Austin research fab won’t be finished until the end of 2027, with equipment moving in later, meaning this costly machine could sit idle for quite some time. For traditional chipmakers, that would be a heavy financial burden, but for Terafab, not so much.

Why Build It Themselves? A Financial Report Setting AI as a Core Business

The urgency of Terafab can be quantitatively shown in the same night’s earnings report.

SpaceX Q3 revenue was $14.575 billion, up 161% year-on-year ($5.576 billion last year); operating profit $2.791 billion, net profit $4.404 billion, both returning to the black; adjusted EBITDA $8.809 billion, up 306% year-on-year; GAAP diluted EPS $0.35. All three exceeded S&P Global Market Intelligence analyst consensus (revenue approx. $12.8 billion, net profit about $2 billion, adjusted EBITDA about $6.8 billion).

The shift in structure matters more than absolute numbers: The AI business, dubbed "Super Intelligence," recorded Q3 revenue of $7.56 billion, surging 830% year-on-year, surpassing Starlink as the biggest revenue source for the first time at about 52% of total revenue, and achieved $431 million in operating profit. The driver is cloud computing rental contracts—Anthropic agreed in May to pay SpaceX’s Colossus data center nearly $45 billion over three years, while Google signed to pay $920 million monthly from October 2024 to June 2029 (AFP).

The cost is seen in capital expenditure: Q3 capex was $19.593 billion, of which $16.4 billion went to the Super Intelligence business, bringing the nine-month total to roughly $48 billion. Computing power scale as of September 30 was 1.7 GW, up from 0.6 GW a year earlier; Musk claims this will reach 2.5–3 GW by year-end, with confidence of adding another 10 GW by 2027, even a "slight possibility" of hitting 20 GW, and expects SpaceX’s annualized recurring revenue (ARR) by end-2027 will reach $500 billion to $1 trillion.

How aggressive is this? Q3 revenue of $14.575 billion annualizes to about $58.4 billion; reaching $500 billion in 15 months means ARR must increase at least 8.5 times. Looking at this alongside Terafab, the logic is self-consistent—when the slope of computing power demand is this steep, the supply chain itself is the biggest bottleneck, and Musk’s solution is to directly build factories.

Merger Signals: From “Can’t Discuss on Earnings Call” to Special Committees

The second matter is what really made the market hold its breath.

In July this year, when asked by Wells Fargo analysts about a possible merger, Musk replied, "This is not something that can be discussed on an earnings call. It requires proper procedures." On October 20, he gave the procedure itself: “Both SpaceX and Tesla have set up special committees, at shareholders’ request, to discuss this idea. These special committees are comprised of independent directors. Because I am an interested party, I cannot participate in the discussions. It’s a bit like a weird kabuki dance—the two sides are talking while I cannot be involved.”

He also stressed three boundaries: the special committees have not agreed on any deal terms, "there is obviously no guarantee" they will; the deal would require shareholder approval, and "for Tesla, it needs a vote in favor by 50% of all shares, not just the voting shareholders"—calling this "a fairly high hurdle"; and, "some people outside seem to think the merger is a done deal. I think that’s a bit of an overstatement."

But he also made his own views clear: "Do I think it’s a good idea? Yes." The rationale is a synergy checklist—super intelligence software, chip design, Terafab, robotics, StarMind constellation launch capability, and "connectivity" as president Shotwell immediately added. If successful, the scale will be unprecedented: the two companies are valued at about $2 trillion and $1.5 trillion respectively.

The governance tension is more noteworthy. According to SpaceX’s IPO filings, Musk holds over 82% voting power in SpaceX, but about 20% in Tesla; Barclays analyst Dan Levy calculated that if SpaceX acquires Tesla with an all-stock deal, Musk will hold about 37% economic interest and control about 56% of voting rights post-merger—a “minority economic, majority control” transaction.

The Market Is Pricing It In, but Disagreement Remains Clear

Sell-side projections are already quite specific. Barclays’ Dan Levy believes the deal could be announced before Tesla’s annual shareholder meeting (already scheduled for December 3), with SpaceX acquiring Tesla at a premium using all stock. Based on his cited $2.1 trillion SpaceX and $1.5 trillion Tesla valuations, Tesla shareholders could receive about 0.7 SpaceX shares per Tesla share; if not this window, it could still be announced in the next 6–12 months. Wedbush’s Dan Ives gives a probability of over 80% for completion before the end of 2027. RBC Capital has already factored a 25–30% SpaceX acquisition premium into Tesla’s target price.

Skeptics are just as direct. Gary Black of The Future Fund cautions bulls "stop hyping the TSLA-SPCX deal," arguing a share-swap merger will be priced at Tesla’s lower multiples and the premium for SpaceX shares will be arbitraged away before completion; Morningstar earlier this year also flagged potentially deep valuation disagreements, especially on the Tesla side. Sell-side target prices don't differ much but are overall cautious: the 12-month average from 38 analysts is $397.83, with JPMorgan at $415, Morgan Stanley at $400, Goldman Sachs at $360, and UBS at $391—all mostly neutral or hold ratings.

Prediction markets provide a different set of numbers: Polymarket’s contract for an "official merger announcement before the end of 2026" is priced around 23.5% (up 6 points in 24 hours); "by the end of 2027," about 64.5%—this is traders’ pricing, not confirmation of an agreement.

Price Tag on Ambition: $48 Billion Capex Meets Tightening Financing Environment

Looking at the two threads together, Musk is offering a plan to "trade the balance sheet for time."

The hard schedule is still clear: SemiAnalysis has calculated that the research fab needs 15 months to build, plus about 6 months to install equipment before any chips can be made; Intel 14A process transfer itself requires extensive equipment certification; ASML’s High-NA EUV capacity in 2028 will still prioritize existing customers. Musk’s own caveat is "the margin for error is large."

The financial constraint is real too. SpaceX’s capex for the first nine months was around $48 billion, and Q4 guidance is $35–40 billion. Meanwhile, AI infrastructure financing is weakening: CoreWeave and Nebius each fell about 28% in October, IREN fell 35%, and Australian data center operator Firmus withdrew its IPO plan citing market volatility.

The most easily underestimated uncertainty is procedural. As of now, the existence of the special committees comes only from Musk’s verbal statement on the results meeting—neither SpaceX’s financials nor any Tesla filings mention it, and neither company has announced committee members or any transaction plan (Drive Tesla Canada). Moreover, the deal must pass two rounds of shareholder voting, antitrust review, and potentially a national security review triggered by a defense contractor absorbing a car company with deep exposure to China.

In other words, the signals Musk released that night were clear and coherent—equipment has been expedited, committees are in place, and computing power goals are written into the guidance; but between "signals" and "facts," lies the very threshold he himself repeatedly emphasizes: the shareholder vote.