I spent three weeks reverse-engineering the Terra/Luna collapse. The same structural debt that killed an algorithmic stablecoin is embedded in Starlink’s revenue projections. The only difference is the collateral: instead of LUNA, it’s orbital slots, spectrum licenses, and a constellation that needs to be replaced every seven years.
Elon Musk says Starlink will carry 50% of global internet traffic. David Friedberg, a potential investor, projects $1 trillion in annual revenue with $300 billion in free cash flow. Both statements are so divorced from physical constraints that they read like a whitepaper from 2021. I’ve audited protocols that promised similar exponential curves. The code always had a revert. The logic held until the liquidity dried up.
Context: The Narrative vs. The Infrastructure
Starlink is a low-Earth orbit satellite internet service. It has ~6,000 operational satellites, ~6 million subscribers, and annual revenue in the $60-100 billion range (unaudited estimates). The product is solid: a phased-array antenna, a subscription, and a lifeline for anyone without fiber. But the recent podcast with David Friedberg and Musk’s subsequent tweets inflated this into a vision where Starlink becomes the backbone of the global internet, capturing 40-50% of the $2.5 trillion telecom market.
Friedberg’s math: $1 trillion revenue, $300 billion free cash flow. That implies a 75% FCF margin. For context, the most efficient telecom operators (think cell towers with zero spectrum cost) struggle to hit 30%. Starlink is a heavy-asset, short-lifecycle business. The margin assumption is not aggressive—it’s delusional.
Core: The Systematic Teardown
Let’s start with the capacity. To carry 50% of global internet traffic, you need to move roughly 200 EB per month by 2027 (based on Cisco projections). Each Starlink V2 Mini satellite can handle about 80-100 Gbps. That’s about 3-4 TB per month per satellite. You’d need 50 million satellite-months of capacity. Even with 40,000 satellites (the current FCC license limit), you’d need each satellite to deliver 5,000 TB per month—a 1,500x increase in per-satellite throughput. No antenna physics on the roadmap supports that. Trace the spectrum, find the truth.
Musk claims “no obvious obstacles.” That’s a lie by omission. The obstacles are: (1) physical Shannon capacity limits, (2) spectrum coordination with terrestrial networks and other satellite operators, (3) the need for 10x more ground stations that can handle the backhaul, and (4) orbital debris constraints that make 40,000 satellites a regulatory nightmare, not a technical ceiling.
Now the business model. Friedberg’s $300 billion FCF implies that after all operating expenses and capital expenditures, Starlink keeps 75% of revenue. But the satellites have a 5-7 year lifespan. That means every year, Starlink must replace 15-20% of its constellation. At current launch costs (even with Starship), maintaining a 40,000-satellite fleet would eat $50-100 billion annually in launch and manufacturing costs alone. That’s before any customer acquisition, ground infrastructure, or operating costs. The FCF number is a fantasy unless Musk assumes he can stop building after reaching scale—which contradicts the “50% of traffic” growth story.
I’ve seen this pattern before. In 2022, I audited a DeFi protocol that claimed it could sustain 20% APY on deposits. The math worked only if new deposits grew exponentially forever. The moment growth slowed, the yield collapsed. Starlink’s revenue prediction depends on the same Ponzi-like assumption: that the telecom market will double in size and Starlink will capture half of it, while maintaining pricing power. But telecom is a commodity business. Bandwidth prices fall 10-20% per year. To reach $1 trillion in revenue, Starlink would need to connect 4-6 billion users (or equivalent enterprise bandwidth) at current ARPU. That’s more than the entire current internet user base. The math is absolute.
Deep Dive into the Cost Structure
Let’s stress-test the FCF margin. The average telecom operator’s FCF margin is 12-18%. Starlink has two advantages: vertical integration (SpaceX launches) and no old infrastructure to maintain. But it has two massive disadvantages: (1) the satellites themselves are the infrastructure, and they die, and (2) Starlink must pay for spectrum licenses globally. The cost of launching and manufacturing 40,000 satellites over 7 years at $1 million each (very optimistic) is $40 billion per cycle. That’s $5.7 billion per year in depreciation alone. Add ground stations, customer terminals (subsidized), marketing, and R&D, and you’re looking at $20-30 billion in annual fixed costs before any variable costs. To get $300 billion FCF, revenue would need to be $400-500 billion, implying a 60-75% margin. That’s not impossible for a software company. It’s impossible for a hardware-satellite operator.
Contrarian: What the Bulls Got Right
I’m not here to say Starlink is a failure. It’s a remarkable piece of engineering. The bulls correctly identify that (1) there is genuine unmet demand for connectivity in underserved areas, (2) AI and autonomous systems will increase demand for real-time data transmission, and (3) Starlink’s first-mover advantage in LEO broadband is real. Direct-to-Device (D2D) partnerships with T-Mobile, KDDI, and Rogers open a wholesale channel that could reach hundreds of millions of phones without Starlink owning the customer relationship. That’s a smart strategy.
But the numbers don’t support the “50% of global traffic” or “$1 trillion” narratives. The more realistic scenario: Starlink becomes a profitable niche player serving 20-30 million high-value users (maritime, aviation, government, rural) with annual revenue of $50-100 billion and FCF of $10-20 billion. That’s still a great business. It’s just not a “new internet” or a “trillion-dollar” one.
Takeaway
Entropy always wins if you stop watching. The hype around Starlink’s financial projections is a distraction from the real challenges: orbital replacement costs, spectrum wars, and the simple fact that bandwidth is a commodity with falling prices. The code of the cosmos doesn’t lie. Friedberg’s model will revert when the launch cadence slows or the satellite lifespan is shorter than expected. I’ve read the reverts before the headlines. This one is coming.
Silence is just uncompiled potential energy. The truth is in the unit economics, not the press releases. Trace the satellites, find the truth.
