Brian French
Imagine taking a pile of money so large it could end homelessness in America, cure several diseases, or buy every person on Earth a decent sandwich—and then deciding the best use of that capital is to build a server farm that floats in the vacuum of space where nobody can fix it when it breaks. Welcome to the theoretical SpaceX/xAI merger: a financial weapon of mass destruction disguised as innovation.
This proposed combination would create a corporate entity with a market capitalization somewhere between “more than the GDP of Poland” and “less than the GDP of France”—let’s call it $500 billion to $1 trillion. That’s real money, the kind that could theoretically be exchanged for goods and services. Instead, it would be deployed to prove that the second-hardest problem in physics (making money in space) can be combined with the first-hardest problem in finance (making money with AI) to create a synergy that loses money faster than both individually.
Let’s explore this masterclass in capital destruction with the respect it deserves: absolutely none.
The Current Cash Flow Situation: Or, “Where Did All The Money Go?”
SpaceX: The Expensive Habit of Throwing Things Into Space
SpaceX currently generates approximately $9-10 billion in annual revenue, which sounds impressive until you realize this is roughly the same as Chipotle. Except Chipotle doesn’t have to defy gravity every time it wants to make a burrito.
The Revenue Breakdown:
- Starlink: $6-7 billion (from beaming internet to people who live where Comcast won’t go)
- Launch services: $3-4 billion (the “taxi service to space” business model)
The Cash Flow Reality: Despite pulling in $10 billion, SpaceX probably generates somewhere between $1-2 billion in actual free cash flow. Where does the rest go? Oh, you know, just the usual:
- Launching 2,000-3,000 satellites per year to replace the ones that inevitably fall back to Earth ($2-3 billion)
- Building Starship, a rocket so ambitious it makes the Apollo program look fiscally responsible ($2 billion annually)
- General “space stuff” that seems reasonable until you realize you’re spending billions on stuff that has to work in an environment actively trying to kill you
The Valuation Comedy: At a $350 billion valuation, SpaceX trades at roughly 175-350x free cash flow. For comparison, Apple—which actually makes products people can hold and use without requiring orbital mechanics—trades at about 25x free cash flow. But sure, rocket company that burns money faster than its own engines should be worth more. This makes perfect sense if you don’t think about it.
xAI: The Art of Burning Money To Teach Computers To Chat
xAI represents humanity’s latest attempt to pour billions into making autocomplete really, really good. Current estimates suggest they’re pulling in maybe $100-200 million in annual revenue while spending $2-4 billion. That’s a profit margin of “deeply negative infinity percent.”
The GPU Bonfire: xAI’s Memphis supercomputer contains 100,000 NVIDIA H100 GPUs, which is tech-bro speak for “we spent $4 billion on graphics cards to run a chatbot.”
Let’s break down this beautiful disaster:
- GPU purchase: $4 billion (or roughly one Twitter acquisition in Elon-dollars)
- Total infrastructure: $6-8 billion (because you can’t just plug 100,000 GPUs into a wall outlet and hope for the best)
- Annual electricity: $130-200 million (enough to power a small city, or one very chatty AI)
- Annual maintenance: $50-100 million (IT support for the world’s most expensive Tamagotchi)
The Business Model:
- Spend $300-400 million per year on operations
- Make $200 million in revenue
- ???
- Somehow profit
The recent $6 billion funding round at a $50 billion valuation provides approximately 2-3 years of runway, which is venture capital speak for “we’ll worry about making money later, probably never.”
The Combined Entity: When Two Wrongs Make A Catastrophic Wrong
The Merged Financial Profile:
- Combined market cap: $400-500 billion (roughly the GDP of Austria, but with worse fundamentals)
- Combined revenue: $9-10 billion (decent money, if you’re not trying to colonize space and teach robots to think)
- Combined free cash flow: -$1 billion to +$1 billion (so basically zero, rounding up for optimism)
- Combined annual capital requirements: $8-12 billion (to maintain the illusion of growth)
The Math That Makes Accountants Weep: Market cap of $450 billion divided by free cash flow of $0 equals infinity. This is not a business valuation; it’s a philosophical statement about the nature of reality. Traditional finance suggests this company should be worth $50-80 billion based on actual cash generation, but we stopped caring about “actual cash generation” sometime around 2015 when tech valuations became more art than science.
Why The Valuation Exists Anyway: The market cap is held up by four pillars of modern financial delusion:
- FOMO: Everyone wants to own the Space + AI stock before their brother-in-law does
- The Greater Fool Theory: Confident there’s someone dumber who will pay more later
- Monopoly Dreams: Maybe if they own space AND AI, they’ll own everything? Somehow?
- Zero-Rate Nostalgia: Investors still pricing assets like it’s 2021 and money is free (narrator: it’s not 2021)
The Interest Rate Problem, Or Why Math Is The Real Enemy: If this entity borrows $50 billion for expansion:
- At 3% (2021 rates): $1.5B annual interest—painful but survivable
- At 6% (2024 rates): $3B annual interest—exceeds all free cash flow
- At 8% (recession rates): $4B annual interest—hello bankruptcy court!
This is a company that can only survive in a world where money is free, which is like saying your boat only floats in environments without water.
The Space Data Center Idea: A Monument To Hubris
Now we arrive at the crown jewel of terrible ideas: building data centers in orbit. This is the business equivalent of deciding regular data centers are too easy and profitable, so let’s make them impossible and ruinous instead.
The Thermal Problem: Physics Is Such A Buzzkill
In space, you can’t cool computers with fans because there’s no air. This is the kind of detail that really should have come up earlier in the planning process. Instead, you need massive radiators—think “International Space Station-sized” radiators for every decent-sized server rack.
The Radiator Economics:
- 100,000 GPUs generate 70 MW of heat (enough to roast approximately 1 million turkeys simultaneously)
- Required radiator surface: 700,000 to 1.4 million square meters (roughly 200-300 football fields of radiators)
- Cost to manufacture space-grade radiators: $10,000-20,000 per square meter (because everything is more expensive when you add “space-grade” to it)
- Total radiator cost: $7-28 billion
- Launch cost (even with Starship’s “affordable” rates): $10-15 billion
- Grand total for cooling: $20-40 billion
For context, that’s enough money to air-condition the entire state of Nevada. But instead, we’re using it to cool servers floating in space where there’s no air to condition.
The Power Problem: The Sun Is Great, But Also Not Great Enough
Solar Panel Math That Makes You Question Everything:
- A real data center needs 100+ megawatts
- Space solar panels generate about 300W per square meter
- Required panel area: 333,000 square meters (82 acres of solar panels, in space)
- Panel cost: $1.7-3.3 billion
- Launch cost: $3-5 billion
- Structural support: $5-10 billion (you can’t just let 82 acres of solar panels flap around)
- Power subsystem total: $10-20 billion for 100 MW
To match a serious terrestrial data center (1 GW): $100-200 billion just for solar panels.
At this point, you could instead build 20 nuclear power plants on Earth, but where’s the prestige in that? Nuclear power is so 1970s. We need orbital power generation that costs 100x more and works 10x less reliably.
The Radiation Problem: When Space Tries To Murder Your Data
Cosmic rays flip bits in computer memory at a rate of about one per gigabyte per day. This is nature’s way of saying “you really shouldn’t be doing this up here.”
Your Options:
- Radiation-hardened chips: Cost 10-50x more and take 2-5 years to develop (and perform like hardware from 2015)
- Heavy shielding: Add 50-100% more mass and launch costs (wrap everything in lead, brilliant)
- Triple redundancy: Run three copies of everything and vote on the answer (democracy, but for computers)
Conservative estimate: Multiply all hardware costs by 3-5x.
So that $8 billion GPU cluster? Now it’s $24-40 billion. But at least it works, sometimes, when the sun isn’t angry.
The Full Price Tag: Numbers That Require Scientific Notation To Discuss
Total Capital Requirement for One Orbital Data Center (100 MW scale):
| Component | Cost | What You Could Buy Instead |
|---|---|---|
| Computing hardware (rad-hard) | $24-40 billion | 80 Twitter acquisitions |
| Thermal management | $20-40 billion | Every house in Cleveland |
| Power generation | $10-20 billion | Tesla’s market cap on a bad day |
| Structure & framework | $10-20 billion | The entire GDP of Iceland |
| Redundancy & spares | $30-50 billion | The UN’s annual budget × 10 |
| Assembly systems | $20-40 billion | Bezos’s net worth |
| Launch costs | $40-80 billion | The Apollo program, inflation-adjusted |
| Ground operations | $5-10 billion | An aircraft carrier |
| TOTAL | $159-300 billion | More than the Marshall Plan |
For a competitive 1 GW facility: $500 billion to $1 trillion (or roughly the cost of fighting a major war, except the only casualty is shareholder value).
The Debt Death Spiral: A Financial Horror Story
Let’s say our intrepid merged entity borrows $250 billion to build this orbital monument to human ambition. Here’s what happens:
The Interest Coverage Table of Doom:
| Interest Rate | Annual Interest Owed | Company Free Cash Flow | Coverage Ratio | Bankruptcy Status |
|---|---|---|---|---|
| 4% (optimistic) | $10 billion | ~$1 billion | 0.1x | Extremely yes |
| 6% (realistic) | $15 billion | ~$1 billion | 0.07x | Super yes |
| 8% (recession) | $20 billion | ~$1 billion | 0.05x | Yes with honors |
For reference, a healthy company maintains coverage above 2.5x. Below 1.5x, your creditors get nervous. Below 1.0x means you’re paying interest with money you don’t have. At 0.1x, you’re not even in the same galaxy as solvency.
The Timeline That Breaks Hearts and Portfolios
Years 1-3: Design, procurement, get excited ($100B spent)
Years 4-7: Launch stuff, assembly, realize how hard this is ($150B spent)
Year 8: First services go live, charge 10x more than terrestrial competitors
Years 8-12: Wonder why nobody uses your expensive space internet
Year 13+: Hope to break even (spoiler: you won’t)
The Compound Interest Surprise Party: Remember that $250 billion loan? At 6% for 13 years while you’re building this thing:
- Total interest if you could pay it: $195 billion
- Total debt if you can’t pay it: $534 billion
- Your face when you realize this: priceless
The company now owes $534 billion for a data center that still costs 10x more to operate than ones on Earth.
The Terrestrial Competition: Or Why This Was Doomed From The Start
While our orbital heroes spend 13 years and $500 billion building their space fortress, Microsoft/Google/Amazon will:
- Build a 1 GW data center in 18-24 months for $10-15 billion
- Put it next to a hydroelectric dam (electricity at $0.02/kWh)
- Use normal hardware they can upgrade next year
- Hire technicians who can actually drive to work
- Achieve 99.99% uptime immediately
The Pricing Death Match:
- Terrestrial AI inference: $0.10-0.50 per million tokens
- Your orbital facility must charge: $2-5 per million tokens (to service debt)
- Market share captured: 0%, except maybe CIA black sites
The only customers willing to pay 10x more for space-based computing are governments with classified needs, worth maybe $1-2 billion annually. Against $15-30 billion in annual debt payments.
The math works out if you’re bad at math.
The Endgame: How This Actually Concludes
After 10-15 years and $400-600 billion invested:
✓ The orbital data center exists and technically functions
✓ Operating costs exceed revenue by $3-5 billion annually
✓ Debt service requires $15-30 billion annually
✓ Cannot raise more capital (prior investors have entered witness protection)
✓ Cannot sell the asset (no buyer exists outside psychiatric facilities)
✓ Cannot shut it down (governments won’t allow “space junk” optics)
Final Resolution:
- Nationalized as “strategic infrastructure” (taxpayers inherit the debt)
- Or bankruptcy, with total losses exceeding $500 billion
- Either way, founders write memoirs titled “What I Learned About Business” and get speaking fees
Awards Won:
- Largest deliberate capital destruction in human history
- Most expensive lesson in thermodynamics ever purchased
- Guinness World Record for “Most Dollars Turned Into Orbital Garbage”
Why Smart People Will Still Invest
Despite everything above, this will attract billions in funding because:
- The Techno-Optimist Defense: “They said reusable rockets were impossible too!” (Yes, but those actually generated revenue)
- The Monopoly Dream: “Winner takes all in Space + AI = $10 trillion market cap!” (Or, alternatively, loser takes all the losses)
- The Exit Strategy: “I’ll sell to a greater fool before this collapses” (Musical chairs with billions)
- The Inflation Hedge: “When the dollar collapses, at least I’ll own space servers!” (Owning worthless things in space is still owning worthless things)
- The Ego Factor: “I want my name on something that goes to space” (Your name on a bankruptcy filing achieves this more cheaply)
- The FOMO: “What if I’m wrong and it works?” (You’re not wrong. It won’t.)
Conclusion: A Masterclass in Destroying Shareholder Value
The proposed SpaceX/xAI merger pursuing orbital data centers represents the rare business plan that violates multiple laws of physics, finance, and common sense simultaneously. It’s the business equivalent of building a submarine to explore the desert—technically you could do it, but why would you?
The Core Problem:
- Current cash generation: ~$1 billion/year
- Required capital: $300-500 billion
- Annual debt service: $18-30 billion
- Competitive advantage: None (actually negative)
- Probability of success: Approaching zero from the wrong direction
This isn’t a “risky investment”—those have a chance of working. This is a financial performance art piece exploring the question “How much money can we make disappear while everyone applauds?”
The physics won’t negotiate. The thermodynamics won’t pivot. The interest won’t compound in your favor. And when $500 billion collides with basic arithmetic, arithmetic wins every single time—usually while laughing.
But hey, at least it’ll look cool in the promotional materials. And really, isn’t that worth half a trillion dollars?
Investor’s Note: If you’re considering investing in this, I’m a Nigerian prince with a bridge in Brooklyn to sell you. It’s also in space. Very disruptive. Ground floor opportunity. Can’t miss.
The author holds no positions in SpaceX or xAI and deeply regrets having learned enough about orbital mechanics to write this article.
About the Author
