Logic doesn't lie. That's the first rule of any technical due diligence. Yet when ARK Invest published its analysis claiming SpaceX's future growth is 90% AI-dependent, the market didn't question the math—it salivated. As a Due Diligence Analyst who spent years dissecting blockchain whitepapers and yield farm code, I see the same pattern: a company with a mature product (rockets, Starlink) suddenly repackages itself as an AI infrastructure play, hoping investors will overlook the assumptions that prop up the entire narrative.
Context
SpaceX, fresh off its largest-ever IPO, is now telling investors that over 90% of its future growth will come from AI—not from launching satellites or Starlink subscriptions. ARK's report positions the company as a vertically integrated AI infrastructure monopolist: control the rockets, control the data centers, control the AI chain. The core thesis: SpaceX can launch computing hardware into orbit at under $100 per kilogram (versus current ~$1,500/kg), build orbital data centers at 25% lower cost than ground-based ones, and lease this compute to giants like Anthropic and Google.
This is not a new story. In crypto, every cycle produces a project that pivots its narrative to the hottest sector: first DeFi, then NFTs, then Layer-2, then AI agents. The mechanism is identical— Read the code, ignore the roadmap. Here, the “code” is the engineering economics: can Starship really hit $100/kg? Can orbital data centers survive radiation without constant maintenance? Can you really run high-power GPUs in a vacuum?
Core: Systematic Teardown of the ARK/SpaceX AI Narrative
Let’s break down the assumptions, one by one, using the same forensic lens I applied to the Terra Luna collapse and the DeFi Summer re-entrancy bugs.
1. The $100/kg Launch Cost Assumption ARK's entire thesis hinges on this number. Today, Falcon 9 costs around $1,500/kg for commercial customers (their internal cost is lower, but not by an order of magnitude). Starship is supposed to slash this to under $100/kg at scale. That’s a 15x reduction from an already industry-leading cost. For comparison, no aerospace program in history has achieved a sustained 15x cost reduction. The Space Shuttle promised reusability to cut costs but failed. Starship is still in testing; its last orbital attempt ended in a fireball. Volatility is just unpriced risk — here, the risk is not just engineering failure but schedule delays. If Starship takes 5 more years to reach operational reliability, SpaceX’s AI narrative collapses.
2. Orbital Data Center Economics ARK claims orbital data centers cost 25% less to build and have “near-zero” energy costs (solar power). This is dangerously naive. Space-qualified electronics cost 10–100x more than terrestrial ones due to radiation hardening, vibration tolerance, and thermal management. A single H100 GPU consumes 700W; in orbit, you must radiate that heat away, which requires heavy radiators. Solar panels provide ~200W/m² in LEO, but to power a cluster of GPUs, you need dozens of square meters of panels, plus batteries for eclipse periods (45 minutes every 90 minutes). The mass and cost of this power system more than offset any savings from “free sunlight.” I’ve audited projects that claimed low-cost mining farms using “free geothermal energy” — they always ignored the infrastructure cost. This is a textbook case of selective data presentation.
3. Customer Validation The report mentions Anthropic and Google as leasing compute. This sounds impressive until you ask: at what scale? At what price? Are they just running a few test workloads for PR, or is this their primary training infrastructure? In crypto, it’s common to see a project announce a “partnership” with a top-tier exchange, only to find it’s a simple listing agreement. Similarly, Anthropic may be renting a handful of server racks on the ground (not in orbit) as a hedge against AWS dependency. Without contract details, this “validation” is noise.
4. The 90% Future Growth Claim If 90% of future growth comes from AI, that implies the traditional business (Starlink, launch services) will contribute only 10% of incremental revenue. Starlink already generates over $10B/year and is growing. For AI to dominate growth, SpaceX would need to generate tens of billions in new AI revenue within a few years. Compare that to AWS, which took a decade to reach $10B from zero, while leveraging existing cloud infrastructure. SpaceX is starting from scratch, with no software ecosystem, no developer tools, no MLOps. This is a narrative-driven valuation pump — classic crypto style.
5. The “Vertical Integration” Mirage Vertical integration is touted as a defensive moat. In practice, it means you must excel in every link of the chain: rocket design, satellite manufacturing, ground stations, data center cooling, AI chips, model optimization, and customer support. No company in history has done all of these at world-class level. SpaceX is great at rockets and satellites; it has zero track record in AI chips or cloud software. Compare with Google, which designs TPUs, operates data centers, and runs Kubernetes. Space X’s vertical integration is a double-edged sword: control comes with massive execution risk. Logic doesn't lie — the probability of success is the product of each independent probability. If each component has an 80% chance of success, the combined probability is 0.8^6 = 26%.
Contrarian Angle: What ARK Got Right
To be fair, ARK identified a genuine long-term opportunity. If launch costs truly fall to $100/kg, space-based computing could unlock niche use cases: ultra-secure data vaults (no physical access), global low-latency edge computing for finance algorithms (if orbital latency is lower than transoceanic fiber), and scientific simulations that benefit from microgravity. The vertical integration does create a unique cost structure that no cloud provider can copy without building their own rockets. Also, the “AI infrastructure” narrative is not entirely baseless—the global IDC market is $500B, and even capturing 1% would be a $5B business.
The contrarian view: ARK may be early, but not wrong in direction. The mistake is in the timeline and the exclusion of risk. They present the best-case scenario as the probable outcome. In crypto, we call this “pump the narrative before the tech is ready.” The smart move is to monitor the signals—not the hype.
Takeaway: Accountability and the Due Diligence Mindset
Every analyst, whether covering crypto or aerospace, must ask: Where is the code? For SpaceX, the code is the engineering data—Starship flight test results, orbital data center prototypes, published power consumption metrics. ARK’s report is a roadmap, not a verification. The market is pricing in a 90% AI-growth future based on a roadmap that has not delivered a single orbital data center.
Read the code, ignore the roadmap. Until SpaceX publishes the actual cost per launch for Starship or deploys a test compute node on orbit, this is a narrative—and narratives can reverse faster than a Falcon 9 landing burn.
As a due diligence analyst, I’ve seen this movie before. Last cycle, it was “blockchain for supply chain” projects promising 90% efficiency gains. This cycle, it’s aerospace companies promising 90% AI revenue. The underlying mechanism is the same: investors buy stories, not facts. But facts always catch up.
Volatility is just unpriced risk. The risk here is not volatility—it’s the probability that the $100/kg target never materializes, that orbital data centers prove uneconomical, or that AI giants don’t move their training to space. When those risks materialize, the narrative will collapse, and only those who read the code will remain.