What is it that can be seen from SpaceX’s listing—SpaceX’s most profitable business?

Author: Silicon Valley's Alan Walker

Today is May 21st, 7:30 a.m. in the morning, and the Zombie Café on California Avenue just opened its doors. Alan Walker from Silicon Valley ordered a double shot espresso latte and spread out the IPO prospectus he downloaded last night on the wooden table. Outside the window, the early morning glow in Palo Alto still has an orange hue — quite matching the color of the cover of this document.

On May 20, 2026, SpaceX officially filed an S-1 with the SEC, stock ticker "SPCX". Roadshow begins on June 4, pricing on June 11, Nasdaq debut on June 12, with a target valuation of $1.75 trillion and a fundraising of $75 billion — this deal, by size, is 2.6 times larger than Saudi Aramco’s record fundraising in 2019, making it the largest IPO in human history. The underwriting lineup is packed with all the big names on Wall Street you can name: Goldman Sachs, Morgan Stanley, BofA, Citi, JPM leading as joint bookrunners, followed by Barclays, Deutsche Bank, UBS, Wells Fargo, Allen & Company… a total of 23 investment banks jointly sponsoring.

This is the most high-profile IPO of the past decade, bar none. But honestly, after reading the prospectus, what Alan most wants to talk about isn’t how dazzling the rockets are — but a fact you’ve never even considered: SpaceX’s most profitable business isn’t rockets at all.

Here are eight angles peeling back this IPO for you.

This isn’t an ordinary IPO — it’s Musk’s "Super Fusion"

SpaceX was founded in 2002, now in its 24th year. Before, everyone thought it would remain a "private company" — with enough primary market funding, and Musk’s dislike of SEC hassles. But this time, it’s different. Three details reveal this is a carefully designed Endgame:

First, timing. On February 2, 2026, SpaceX first merges with xAI (called "xAI Merger" in the prospectus), bundling in the X (formerly Twitter) acquired in March 2025; then on May 4, 2026, a 5-for-1 stock split; on May 20, the S-1 is filed. Less than three and a half months in total, a series of moves complete the "rocket + satellite + AI + social media" mega-assembly. This rhythm isn’t spontaneous — it’s a meticulously scheduled timetable.

Second, structure. Class A shares have one vote per share, Class B shares have ten votes per share — similar to Google, Meta, Snap’s dual-class structure. Musk, via Class B, firmly locks in majority voting rights. The company explicitly discloses it will operate as a "controlled company" post-IPO, exempt from many corporate governance rules (such as not needing a majority of independent directors). Simply put: money can flow in, but Musk’s voice remains unchallenged.

Third, underwriting lineup. 23 investment banks jointly sponsor, with five being first-tier. Having done VC for nearly twenty years, Alan has seen this scale only twice before: Alibaba in 2014 and Saudi Aramco in 2019. Each gets a slice of the pie, indicating Musk is deliberately maximizing "interest binding" — motivating the big banks to push the price upward and feeding their big clients.

Note, the document also mentions a phrase many overlook: the xAI Merger and X Merger are explicitly listed as "transactions between entities under common control" — a related-party transaction under common control. Musk is "internally shifting" among his controlled companies, making the final listed SpaceX look like a complete "AI + space" giant. This is compliant, but you need to understand this move.

Revenue breakdown: Starlink is the real cash cow

See the data. Here’s the full-year 2025 revenue and profit breakdown for the three business segments:

Starlink accounted for 61% of total revenue in 2025, contributing nearly all operating profit (a positive $17.5k), while all other businesses are in the red.

Even more impressive is the growth rate. The Connectivity segment, where Starlink resides, grew revenue by 49.8% year-over-year in 2025, operating profit by 120.4%, EBITDA by 86.2% — this isn’t startup growth; it’s a "giant dancing" with a $11.4 billion revenue scale, still accelerating.

What about rockets? The Space segment only generated $4.09B in revenue in 2025, and still lost $657 million — because Starship R&D is extremely costly (Space segment R&D investment was $3B in 2025). Musk himself states plainly in the document: the purpose of the rocket business is "to enable our other businesses" — launching satellites for Starlink, transporting cargo for future AI compute satellites.

So that "most profitable business" in the headline — is that white box on your roof, on your yacht, or pressed into the airplane cabin ceiling.

Looking at user data: as of March 31, 2026, there are about 9,600 satellites in orbit, 10.3 million subscribers, covering 164 countries/regions. The satellite-to-phone direct connection service (Starlink Mobile) is live in over 30 countries, with 7.4 million active devices monthly. These satellites now account for 75% of the total mobile satellite constellation — in the sky, three out of four active satellites belong to Musk.

This is the story Musk can truly tell Wall Street — not "how cool rockets are," but "I am now the world’s largest satellite internet operator, with subscription cash flow, steady monthly income, and increasing gross margins."

AI: the cash-burning beast — $7.7 billion quarterly capex, burning Starlink’s cash flow

If Starlink is the cow, then AI is the hungry little monster next to it. Look at the capex data for Q1 2026:

One quarter, AI spent $7.7 billion — nearly three times the combined Space and Connectivity segments.

For the full year, 2025 AI capex was $12.73B, with a loss of $15B — a pure cash burner.

But Musk isn’t worried. Why? Because he has Starlink as a cash pump, plus a mysterious "external compute order" flywheel that no one fully understands. The IPO prospectus reveals three major external contracts, which are the biggest hidden bombs of this IPO:

1. Anthropic’s compute contract — a game-changer.

   Signed in May 2026, a Cloud Services Agreement, where Anthropic pays SpaceX $1.25 billion per month to rent COLOSSUS and COLOSSUS II compute power, with a term until May 2029. Total about $45 billion over three years, averaging $15 billion per year. What’s the scale? SpaceX’s entire AI segment revenue in 2025 was only $3.2 billion — meaning this single Anthropic contract’s annual value is nearly five times the entire AI revenue in 2025. When Anthropic’s capacity fully ramps in 2027, AI segment revenue could leap from $3.2 billion to over $20 billion.

2. The "enemy vs. ally" nature of this money.

   Anthropic is a core competitor of OpenAI, backed by Google (over $2 billion invested) and Amazon (over $4 billion). Yet, Anthropic signed a three-year compute order with Musk’s xAI — meaning Google and Amazon’s money, channeled through Anthropic, flows continuously into Musk’s pocket. Even more painfully, Anthropic is preparing for an IPO in late 2026 with a valuation target of $70k. A large chunk of its funds will go to pay SpaceX. This forms a perfect "enemy capital feeding itself" closed loop.

3. Cursor’s option agreement.

   In April 2026, SpaceX provided GPU compute to Cursor (the hottest AI programming startup in San Francisco), and signed an "acquisition option" at a $60 billion valuation. If SpaceX backs out, it owes Cursor $1.5 billion termination fee plus $8.5 billion deferred service fee — a "killer clause" almost locking out other options for Cursor.

4. Terafab’s chip ambitions.

   Collaborating with Tesla and Intel to produce chips, with a long-term goal of "producing 1 terawatt of compute power annually." What’s a terawatt? About twice the total power consumption of today’s global data centers. This is Musk’s plan to bypass Nvidia’s chip manufacturing.

These three contracts (plus inevitable future orders from Meta, Mistral, Cohere, and national AI projects) will turn SpaceX’s AI business into a true "AI cloud like AWS." Note: McKinsey estimates the global compute infrastructure market will reach $70 trillion in the next five years. If SpaceX/xAI captures just 10%, that’s $7 trillion in revenue — not counting the space premium once orbital AI compute takes off.

Thus, SpaceX’s current capital strategy is: profits from Starlink flow into the company, the rocket business handles "launch" expenses, AI takes the largest capex share to stockpile compute, then wholesales the compute to other AI firms (including competitors) via long-term contracts, and the cash recovered fuels further capex. It’s a self-sustaining "internal financing + external feeding" flywheel — other AI companies need outside funding, Musk does internal financing at home, with external labor from Anthropic.

Why include xAI? — the most brilliant valuation move

Many ask: isn’t xAI just a different company from SpaceX? Why merge before going public? The answer involves Musk’s goals: maximize valuation, optimize capital use, secure control, and tax planning. This merger accomplishes all four.

First, valuation multiple upgrade. If SpaceX goes public alone, what valuation would the market assign? "The world’s largest satellite internet + largest rocket operator" might get 15-20x P/S (based on 2025 revenue of $15.4 billion, roughly $230-300 billion). But if xAI is merged in? Its last funding round at the end of 2025 valued it at $200 billion. Post-merger, SpaceX becomes a "quadruple play" of "AI infrastructure + satellites + rockets + social media" — a narrative premium that pushes valuation beyond simple P/S multiples.

Second, leveraging Starlink’s cash flow as "credit backing" for xAI.

xAI alone valued at $200 billion, but losing over $6 billion annually. Investors buy it for the "dream." Merging into SpaceX, where Starlink’s $4.4 billion operating profit offsets much of xAI’s $6.3 billion loss, makes the whole financial picture look healthier. It’s a form of "cash flow credit transfer" — good business cash flows backing the high-burn business valuation.

Third, locking in control.

If xAI went public alone, Musk’s stake could be diluted by VCs. Merging into SpaceX, with Class B shares giving 10:1 voting rights, Musk’s control remains intact. Early xAI employees’ equity becomes SpaceX Class A shares with no voting power, further consolidating control.

Fourth, the appearance of compliant related-party transactions.

Both the xAI Merger and X Merger are listed as "transactions between entities under common control" — meaning Musk is "internally transferring" among his private companies, accounting as a "merger under common control." This avoids goodwill impairments and keeps financials clean. Likely, it also avoids the need for independent fair value assessments of xAI — which would leave a valuation anchor open to scrutiny.

What is SpaceX really? Rockets are just the entry ticket; the real story is on the moon and in orbit

Today’s SpaceX is actually a fusion of six companies:

• The world’s largest rocket operator
  — Since 2023, responsible for over 80% of global orbital launches annually, with 650 Falcon launches and over 99% success rate.

• The world’s largest satellite internet operator
  — 9,600 satellites in orbit, 75% of the global mobile satellite constellation, 10.3 million paying subscribers.

• One of the world’s top AI training compute operators
  — About 1 GW of compute (COLOSSUS + COLOSSUS II), claiming to be the first gigawatt-scale "coherent" AI training cluster on Earth. First phase of COLOSSUS built in 122 days, COLOSSUS II in 91 days (industry average for a 100 MW data center is two years).

• One of the top three social media platforms
  — X + Grok combined MAU of 550 million, with 350 million daily original content posts. 117 million monthly active users utilize Grok’s AI features.

• Future space data center operator
  — Planning to deploy orbital AI compute satellites starting 2028, with a long-term goal of 100 GW of compute annually in orbit, requiring "thousands of launches" and "about a million tons of payload into orbit" each year.

• Future lunar developer
  — Explicit plans for "lunar mass drivers," lunar energy production, lunar manufacturing, and eventually Mars cities. Musk calls this the path toward a "Kardashev Type II civilization."

Looking at these layers, a chilling realization emerges — what Musk is building isn’t just a company, but a new kind of sovereignty.

Why can valuation reach such heights? Because he’s painting himself a $28.5 trillion market — roughly the entire US GDP in 2024.

How the capital markets see it: valuation at $1.75 trillion — how is that calculated?

First, the conclusion: the IPO target valuation is set at — $1.75 trillion, raising $25k. The pricing on June 11, Nasdaq debut on June 12. It’s the largest IPO in human history, bar none — 2.6 times larger than Saudi Aramco in 2019 (raising $290 billion), and three times Alibaba’s 2014 IPO ($25 billion).

Looking at SpaceX’s valuation trajectory, it’s mind-boggling:

In 24 months, valuation soared from $285k to $1.75 trillion — an 8.3x increase. Such a steep slope is rarely seen in public markets; only Tesla in 2020 has matched it.

What does $1.75 trillion mean?

• Equal to 9.2 Reliance Reliance Industries (India’s largest conglomerate).
• On IPO day, SpaceX will surpass Tesla ($1.57 trillion), becoming the 9th largest company globally, behind Apple, Microsoft, Nvidia, Google, Amazon, Meta, Saudi Aramco, Berkshire Hathaway.
• P/S ratio based on 2025 revenue ($285k): about 94x; based on 2026 forecasted revenue ($22-24 billion): about 73-79x. For comparison, Tesla today trades at about 12x P/S, Nvidia around 25x.
• This valuation assumes: SpaceX’s revenue exceeds $290k by 2030, with profit margins over 30%, driven by continued 50% Starlink growth, AI annual revenue over $20 billion, Starship commercialization, and orbital AI compute.

But several red flags:

• Profits are still small.
  — 2025 GAAP net loss: $4.9 billion; Q1 2026 operating loss: $240k. Investors buy the story and cash flow, not profit.

• Related-party transaction risks.
  — S-1 discloses Tesla owns 17.5k shares of SpaceX Class A (worth about $2.5 billion at IPO valuation). Valor Equity Partners’ AI lease obligations exceed $17.5k; SpaceX paid $29B in 2025, $857 million in early 2026. This "internal Musk ecosystem" structure will be scrutinized by SEC and media.

• Dual-class shares + 85.1% voting control by Musk.
  — Retail investors buying SPCX have little governance power, relying on "trust Musk."

• Retail quota of 30%.
  — Reports say 30% of IPO shares are reserved for retail — much higher than usual 5-10%. Musk is counting on retail investors to buy in, aware that institutional investors are cautious about a $1.75 trillion valuation.

My personal view: a $20 trillion opening isn’t surprising, but whether it stays above $15 trillion in 12 months depends on Starship’s first commercial flight in late 2026. If Starship fails, valuation could drop 30%; if it succeeds, combined with Anthropic’s compute revenue, $20-25 trillion is a reasonable expectation.

Where will liquidity be drained from the market upon SpaceX’s listing? OpenAI under immense pressure

A VC term: "sucking sound" — a mega IPO will siphon liquidity from other assets. SpaceX will likely draw two types of capital: first, AI-focused funds (reducing Nvidia/MSFT/GOOG/META holdings), second, space concept stocks (Rocket Lab, AST SpaceMobile). But the biggest impact will be on OpenAI. Here’s why.

Eight specific ways SPCX impacts OpenAI

1. "The crown of most expensive AI company" is lost forever.
   — In October 2025, OpenAI’s $500 billion valuation tender made it the most expensive AI firm. But just two months later, SpaceX’s December tender hit $800 billion; by the S-1, the target is $1.75 trillion — 3.5 times OpenAI’s valuation. The "AI leader" label now belongs to SpaceX, with OpenAI falling to second or third place (depending on whether you count Google+DeepMind).

2. Anthropic’s defection is a blow to OpenAI’s morale.
   — Founded in 2021 by former OpenAI execs (Dario Amodei, Daniela Amodei, Tom Brown, Sam McCandlish), claiming to build "more responsible AI." Yet, in May 2026, they signed a three-year, $17.5k compute contract with Musk’s xAI — effectively paying Musk’s company instead of OpenAI. It’s a symbolic blow, more painful than the business loss.

3. Hardware race: Stargate vs. COLOSSUS II — a direct confrontation.
   — OpenAI announced Stargate infrastructure in January 2025, aiming for $50 billion in AI compute, with some deployment in Texas by 2026. Meanwhile, Musk’s COLOSSUS factories built in 122 and 91 days, respectively, with industry averages of two years for a 100 MW data center. When Stargate’s Abilene site goes live, SpaceX will already have a 1 GW coherent cluster running for months. The market trusts operational machines more than announced plans.

4. Microsoft’s "application layer" attack on OpenAI.
   — OpenAI’s key pitch (2025-2026): "agentic AI" — autonomous AI agents that operate software and workflows. But the S-1 states that Macrohard (Microsoft’s "new" company) is designed to fully simulate digital workflows, enhance human operations, and create an AI-run software company. Essentially: "We aim to replace Microsoft + Salesforce + OpenAI enterprise." Backed by Tesla’s supply chain, SpaceX’s compute, X’s real-time data, and xAI’s Grok — what does OpenAI have to counter?

5. OpenAI’s IPO will be squeezed out of the best window.
   — Market expects OpenAI to IPO in late 2026 or 2027, with a valuation of $750-15.7k. But once SPCX hits the market, the AI sector’s valuation anchor is set at — 58-65x P/S. To match that, OpenAI must demonstrate comparable revenue growth, moat, and expansion potential. But OpenAI lacks launch capability, Starlink, Tesla synergy, real-time data, or Macrohard. It only has models and ChatGPT. So, its IPO valuation will likely be forced down to $500-1.94B, well below internal expectations.

6. Microsoft’s "exclusive partnership" with OpenAI is devalued.
   — From 2023-2025, OpenAI’s most valuable asset was "Azure exclusive compute + global distribution." Now, SpaceX has its own compute (COLOSSUS), own app portals (X + Grok + Macrohard), and hardware (Terafab). A fully integrated competitor emerges, turning Microsoft’s "exclusivity" into a chain. Microsoft has been diluting dependence on OpenAI, investing in Mistral, and developing its own models, foreseeing this future.

7. Talent war: OpenAI begins to lose ground.
   — Once SPCX goes public, Musk can use liquid Class A shares to poach talent — employees can sell options immediately. OpenAI’s "profit-cap units" require IPO to cash out. Musk is openly recruiting OpenAI researchers, offering packages 1.5-2x better than OpenAI’s, all in liquid SPCX shares. The window: Q3-Q4 2026 — a talent hemorrhage.

8. Musk vs. Altman — a 11-year personal feud escalates.
   — Musk co-founded OpenAI with Altman in 2015, left in 2018 over disagreements. In 2024, Musk sued OpenAI for violating "nonprofit roots." In early 2025, Musk offered to buy OpenAI for $97 billion, which Altman rejected publicly. Their conflict isn’t just business — it’s about "who defines the AI era." After SPCX’s listing, Musk’s war chest includes $75 billion cash, a top-10 global market cap stock currency, and a fully integrated AI infrastructure. Altman’s only counter is to rush an IPO before Musk’s Macrohard commercializes, locking in enterprise clients. It’s a countdown war.

A prescription for OpenAI: — either aggressively IPO in late 2026 at a "discounted" $550 billion valuation to raise cash, or abandon public markets, deeply partner with Microsoft, and focus on a narrow "best model API" moat. Hesitate further, and OpenAI risks becoming the "Yahoo of AI" — once a king, overtaken by a more vertically integrated rival.

Impacts on other AI firms, ranked:

• Anthropic — appears "loss-making" (paying Musk $45 billion), but in reality, forced to cooperate, with a "Switzerland" AI positioning, IPO valuation could still go up to $350 billion or more.
• Microsoft — neutral to slightly negative. Azure’s AI compute continues, but SPCX becomes a second "hyperscaler" with vertical integration, narrowing its moat.
• Google / Meta — limited impact. They have their own TPU/MTIA paths; the main effect is on X’s competition with YouTube Shorts/Threads.
• Nvidia — short-term benefit (before Terafab ramps up, SpaceX needs to buy cards); long-term caution — once Terafab is operational, one of the top three AI clients could turn into a competitor.
• Oracle / Softbank — Stargate partners, deeply tied to OpenAI, will face pressure together. Oracle’s stock is a "reverse indicator" to watch.

Another overlooked detail: Cursor’s $20k options. If Musk exercises, it’s like using SPCX stock to buy one of the hottest AI startups — signaling that SpaceX will become the biggest buyer in AI applications. All AI startups, like Perplexity, Mistral, Cohere, Character, Suno, ElevenLabs, are lining up for bids. OpenAI’s "application ecosystem" will be gradually bought out by Musk with cash.

Back to the core question: SpaceX’s ultimate destiny is to privatize "civilization itself"

Closing the last page of the S-1, I stared at the early morning commuters on California Avenue. I suddenly understood — the most profitable business of SpaceX isn’t rockets — but the "industry map" Musk has drawn across 30 years. Let’s unfold this map in four layers:

(Note: these four layers are not parallel but sequential — each layer depends on the capabilities of the previous one. Musk isn’t doing diversification; he’s doing progressive vertical integration.)

Layer 1 (2026–2028): Earth infrastructure sovereignty

Replace and consolidate all human "information infrastructure" on Earth. Starlink replaces telecom operators like AT&T, Verizon, China Mobile; Starlink Mobile replaces cell towers; X replaces Meta’s social media; Grok replaces ChatGPT; Macrohard replaces Microsoft + Salesforce + Workday; Terafab aims to take a chunk of TSMC, Nvidia, Intel’s market. In short: SpaceX’s goal is to build "global telecom + global social media + global AI assistant + global enterprise software + global chip manufacturing". The foundation must be laid before 2028.

Layer 2 (2028–2032): Orbital economy pioneer

The S-1 clearly states the goal — deploy 100 GW of compute annually in orbit, requiring "thousands of launches, about a million tons of payload into orbit." Orbital AI compute satellites in sun-synchronous orbit, with solar efficiency five times that of Earth, can forgo terrestrial power grids. Musk’s answer to "AI energy ceiling": go to space. This layer also includes: space solar power transmission, microgravity high-purity materials manufacturing, Starship point-to-point Earth travel (e.g., 30 minutes from Paris to Tokyo), commercial space stations. By 2032, the "orbit GDP" should surpass $15k.

Layer 3 (2032–2040): Lunar economy operator

The S-1 mentions a rarely noticed term — "lunar mass driver". This is an electromagnetic launcher to send materials from the Moon’s surface, leveraging its 1/6 gravity and no atmosphere, making launches vastly cheaper than from Earth. Once built, the Moon becomes a logistics hub in the solar system. Coupled with lunar helium-3 (fusion fuel), lunar water ice (rocket fuel raw material), and lunar solar power (14 days of continuous daylight), Musk’s vision is to turn the Moon into a "closest industrial base to Earth." All materials destined for farther out — Mars, asteroids, outer solar system — will depart from the Moon.

Layer 4 (2040+): Interplanetary civilization private contractor

Mars cities, asteroid mining, Kardashev Type II (Dyson sphere-level energy use) — sounds sci-fi, but it’s all in the S-1. Musk’s KPI: "extend the light of consciousness to the stars." Translated for investors: by the 2040s, SpaceX aims to be the first "interplanetary company" — operating assets, employees, tax bases, and autonomous energy on Earth, Moon, and Mars. An unprecedented entity.

Stacking these four layers reveals a chilling truth — Musk’s grand project isn’t just a company, but a new form of sovereignty.

Why can valuation reach such heights? Because he’s drawing a $28.5 trillion "market" — roughly the entire US GDP in 2024.

How the market values it: $1.75 trillion — how is that calculated?

First, the conclusion: the IPO valuation target is — $1.75 trillion, raising $75 billion. The price set on June 11, debut on June 12. It’s the largest IPO in history, 2.6x Saudi Aramco, 3x Alibaba.

Looking at SpaceX’s valuation leap, it’s staggering:

In 24 months, from $210 billion to $1.75 trillion — an 8.3x increase. Such a steep trajectory is rare; only Tesla in 2020 has matched it.

What does $1.75 trillion mean?

• Equal to 9.2 Reliance Industries.
• On IPO day, surpass Tesla ($1.57 trillion), becoming the 9th largest company globally, behind only Apple, Microsoft, Nvidia, Google, Amazon, Meta, Saudi Aramco, Berkshire Hathaway.
• P/S based on 2025 revenue ($17.5k): about 94x; based on 2026 forecast ($22-1M): about 73-79x. Tesla today trades at ~12x P/S, Nvidia at ~25x.
• This valuation assumes: SpaceX’s revenue exceeds $150 billion by 2030, with profit margins over 30%, driven by continued 50% Starlink growth, AI annual revenue over $20 billion, Starship commercialization, and orbital AI compute.

But several red flags:

• Profits are still small.
  — 2025 GAAP net loss: $4.9 billion; Q1 2026 operating loss: $1.943 billion. Investors buy the story and cash flow, not profit.

• Related-party transaction risks.
  — S-1 discloses Tesla owns 19 million shares of SpaceX Class A (worth about $2.5 billion at IPO valuation). Valor Equity’s AI lease obligations exceed $20 billion; SpaceX paid $885 million in 2025, $17.5k in early 2026. This "internal Musk ecosystem" structure will be scrutinized.

• Dual-class shares + 85.1% voting control by Musk.
  — Retail investors have little governance power, relying on "trust Musk."

• Retail allocation of 30%.
  — Reports say 30% of IPO shares reserved for retail — much higher than usual 5-10%. Musk is counting on retail investors to buy in, aware of institutional caution.

My judgment: a $20 trillion opening isn’t unlikely, but whether it stays above $15 trillion in 12 months depends on Starship’s first commercial flight in late 2026. If Starship fails, valuation could drop 30%; if it succeeds, combined with Anthropic’s compute revenue, $20-25 trillion is a reasonable expectation.

Where will liquidity be drained from the market? OpenAI under immense pressure

A VC term: "sucking sound" — a mega IPO will draw liquidity from other assets. SpaceX will likely pull capital from two sources: first, AI funds (reducing Nvidia/MSFT/GOOG/META holdings), second, space concept stocks (Rocket Lab, AST SpaceMobile). But the biggest impact will be on OpenAI. Here’s why.

Eight specific ways SPCX impacts OpenAI

1. "The crown of most expensive AI company" is lost.
   — In October 2025, OpenAI’s $500 billion valuation tender made it the priciest AI firm. But two months later, SpaceX’s December tender hit $800 billion; by the S-1, the target is $1.75 trillion — 3.5x OpenAI. The "AI leader" label shifts to SpaceX, with OpenAI falling behind (or third, depending on inclusion of Google+DeepMind).

2. Anthropic’s defection hits morale hard.
   — Founded in 2021 by ex-OpenAI execs, claiming "more responsible AI." Yet, in May 2026, they signed a three-year, $450 billion compute deal with Musk, paying him instead of OpenAI. A symbolic blow, more damaging than the business loss.

3. Hardware race: Stargate vs. COLOSSUS II — a direct showdown.
   — OpenAI announced Stargate infrastructure in January 2025, aiming for $50 billion in AI compute, with some deployment in Texas by 2026. Meanwhile, Musk’s COLOSSUS factories built in 122 and 91 days, respectively, with industry averages of two years for a 100 MW data center. When Stargate’s Abilene site goes live, SpaceX will already have a 1 GW coherent cluster running for months. The market trusts operational hardware more than announced plans.

4. Microsoft’s "application layer" attack on OpenAI.
   — OpenAI’s key pitch (2025-2026): "agentic AI" — autonomous AI agents that operate software and workflows. But the S-1 states Macrohard (Microsoft) is designed to fully simulate digital workflows, enhance human operations, and create an AI-run software company. Essentially: "We aim to replace Microsoft + Salesforce + OpenAI enterprise." Backed by Tesla’s supply chain, SpaceX’s compute, X’s real-time data, and xAI’s Grok — what does OpenAI have to counter?

5. OpenAI’s IPO will be squeezed out of the best window.
   — Market expects OpenAI to IPO late 2026 or 2027, with a valuation of $750-830 billion. But once SPCX hits the market, the AI sector’s valuation anchor is set at — 58-65x P/S. To match that, OpenAI must demonstrate comparable revenue growth, moat, and expansion potential. But OpenAI lacks launch capability, Starlink, Tesla synergy, real-time data, or Macrohard. It only has models and ChatGPT. So, its IPO valuation will likely be forced down to $500-600 billion, well below internal expectations.

6. Microsoft’s "exclusive partnership" with OpenAI is devalued.
   — From 2023-2025, OpenAI’s most valuable asset was "Azure exclusive compute + global distribution." Now, SpaceX has its own compute (COLOSSUS), own app portals (X + Grok + Macrohard), and hardware (Terafab). A fully integrated competitor emerges, turning Microsoft’s "exclusivity" into a chain. Microsoft has been diluting dependence on OpenAI, investing in Mistral, and developing its own models, foreseeing this future.

7. Talent war: OpenAI begins to lose ground.
   — Once SPCX goes public, Musk can use liquid Class A shares to poach talent — employees can sell options immediately. OpenAI’s "profit-cap units" require IPO to cash out. Musk is openly recruiting OpenAI researchers, offering packages 1.5-2x better than OpenAI’s, all in liquid SPCX shares. The window: Q3-Q4 2026 — a talent hemorrhage.

8. Musk vs. Altman — a 11-year personal feud escalates.
   — Musk co-founded OpenAI with Altman in 2015, left in 2018 over disagreements. In 2024, Musk sued OpenAI for violating "nonprofit roots." In early 2025, Musk offered to buy OpenAI for $97 billion, which Altman rejected publicly. Their conflict isn’t just business — it’s about "who defines the AI era." After SPCX’s listing, Musk’s war chest includes $75 billion cash, a top-10 global market cap stock currency, and a fully integrated AI infrastructure. Altman’s only counter is to rush an IPO before Musk’s Macrohard commercializes, locking in enterprise clients. It’s a countdown war.

A prescription for OpenAI: — either aggressively IPO in late 2026 at a discounted $550 billion valuation to raise cash, or abandon public markets, deeply partner with Microsoft, and focus on a narrow "best model API" moat. Hesitate further, and OpenAI risks becoming the "Yahoo of AI" — once a king, overtaken by a more vertically integrated rival.

Impacts on other AI firms, ranked:

• Anthropic — appears "loss-making" (paying Musk $45 billion), but in reality, forced to cooperate, with a "Switzerland" AI positioning, IPO valuation could still go up to $350 billion or more.
• Microsoft — neutral to slightly negative. Azure’s AI compute continues, but SPCX becomes a second "hyperscaler" with vertical integration, narrowing its moat.
• Google / Meta — limited impact. They have their own TPU/MTIA paths; the main effect is on X’s competition with YouTube Shorts/Threads.
• Nvidia — short-term benefit (before Terafab ramps up, SpaceX needs to buy cards); long-term caution — once Terafab is operational, one of the top three AI clients could turn into a competitor.
• Oracle / Softbank — Stargate partners, deeply tied to OpenAI, will face pressure together. Oracle’s stock is a "reverse indicator" to watch.

Another overlooked detail: Cursor’s $600 million options. If Musk exercises, it’s like using SPCX stock to buy one of the hottest AI startups — signaling that SpaceX will become the biggest buyer in AI applications. All AI startups, like Perplexity, Mistral, Cohere, Character, Suno, ElevenLabs, are lining up for bids. OpenAI’s "application ecosystem" will be gradually bought out by Musk with cash.

Back to the core question: SpaceX’s ultimate destiny is to privatize "civilization itself"

Closing the last page of the S-1, I stared at the early morning commuters on California Avenue. I suddenly understood — the most profitable business of SpaceX isn’t rockets — but the "industry map" Musk has drawn across 30 years. Let’s unfold this map in four layers:

(Note: these four layers are not parallel but sequential — each layer depends on the capabilities of the previous one. Musk isn’t doing diversification; he’s doing progressive vertical integration.)

Layer 1 (2026–2028): Earth infrastructure sovereignty

Replace and consolidate all human "information infrastructure" on Earth. Starlink replaces telecom operators like AT&T, Verizon, China Mobile; Starlink Mobile replaces cell towers; X replaces Meta’s social media; Grok replaces ChatGPT; Macrohard replaces Microsoft + Salesforce + Workday; Terafab aims to take a chunk of