Orbital Compute Is Not a Data Center: Google’s 15-Minute TPU Test and the Crypto Compute Mirage

On October 1, a refrigerator-sized satellite built by Planet Labs will ride a SpaceX Falcon 9 into low Earth orbit from Vandenberg Space Force Base. Inside it: four Google Trillium Tensor Processing Units, a kilowatt of solar power, and a 15-minute compute window. That is not a typo. The TPUs will run for approximately 15 minutes before thermal shutdown forces a cooldown. The satellite is called MVP. It is the first physical artifact of Google’s Project Suncatcher, announced in November 2025. The launch is part of SpaceX’s Transporter-18 rideshare mission. This is not an operational data center. It is a survivability test. For a crypto hedge fund analyst, the signal is not the satellite. It is the vertical integration. Google owns the silicon, the cloud, and now the orbital platform. That is the compute landlord thesis extending from terrestrial GPU factories to orbit. In 2017, I audited 45 ICO whitepapers. None of them mentioned thermal vacuum chambers. That is why most of them are dead. Project Suncatcher is a research moonshot. The immediate objective is to gather in-orbit data on how Trillium TPUs handle launch vibration, sustained acceleration up to 10 times the force of gravity, radiation from solar events and cosmic rays, and thermal extremes. In a vacuum, there is no airflow. Heat can only be dissipated via radiators. That is a fundamentally different cooling architecture than anything terrestrial data centers require. Ground testing at UC Davis’s Crocker Nuclear Laboratory exposed Trillium TPUs to proton beam irradiation while running AI workloads. The chips survived a total ionizing dose greater than what they would receive during a five-year space mission. Thermal vacuum chamber tests confirmed the heat-pipe-and-radiator approach works in simulated conditions. But some things, as Google’s team acknowledges, can only be tested in orbit. Google’s own blog post, published September 24, frames the mission as “measured, deliberate steps” toward a long-term goal. The long-term goal is orbital data centers. SpaceX is separately targeting approximately 1 gigawatt per year of orbital compute by the end of 2027 via its own Nvidia-powered AI1 satellites. Starcloud and Axiom are pursuing similar solar-powered orbital AI data center concepts. The driving force is the same across all of them: terrestrial data centers are hitting physical and political limits. Power availability, cooling requirements, land use, and permitting delays are constraining the expansion of AI compute infrastructure on the ground. In low Earth orbit, satellites can access near-constant sunlight, generating up to eight times more solar power than on Earth’s surface. The physics are attractive. The engineering is brutal. The economics are unknown. Let us analyze the 15 minutes. A 15-minute compute window followed by thermal shutdown implies a duty cycle. If cooldown takes 45 minutes, the duty cycle is 25%. If cooldown takes 90 minutes, it is 14%. A 1-kilowatt solar array in LEO, after power conversion and battery charging, might deliver 600 to 800 watts to the compute payload. Trillium TPUs are efficient, but a four-chip board still generates heat. In vacuum, that heat must be radiated. Radiator area scales with heat rejection. At 1 kilowatt, you need roughly 2 to 3 square meters of radiator at 300 Kelvin. At 1 megawatt, you need thousands of square meters. That is not a satellite. That is a space station. At 1 gigawatt, it is a fantasy. Google’s own modeling suggests cost parity with terrestrial data centers could arrive in the mid-2030s, contingent on launch costs falling below approximately $200 per kilogram. Current Falcon 9 launch costs are around $2,700 per kilogram. Starship promises $10 to $100 per kilogram, but it is not operational at that price. The gap is not incremental. It is a 90% or greater cost reduction. Gartner, in February 2026, called the concept “pie-in-the-sky.” I agree with the assessment, but not the conclusion. The point is not whether orbital data centers are economical today. The point is that Google is spending capital to test the hypothesis. That is a signal. In crypto, we have a similar pattern: DePIN compute networks. Akash, Render, io.net. They tokenize GPU and CPU capacity. They use cryptographic attestation to verify hardware. They sell compute on-chain. The pitch is decentralized compute. The reality is that most of these networks rely on a small number of large providers. The hardware attestation is weak. The token incentives subsidize supply, not demand. Now imagine orbital compute. You cannot physically access a satellite. You cannot swap a failed GPU. You cannot audit the hardware without launching a repair mission. Trust shifts entirely to the satellite operator. That is not decentralization. That is vertical integration. The entity that controls the chip, the launch, and the orbital platform controls the stack. Google’s Project Suncatcher is a vertical integration play. The payload is Google’s own silicon. The satellite bus is Planet Labs. The launch is SpaceX. The compute layer is Google’s. This is the compute landlord thesis. We have traced this pattern through SpaceX’s $119 billion Terafab initiative, through Nvidia’s $10 billion anchor in Anthropic’s IPO, and through the $517 billion in compute commitments Anthropic accumulated in 11 months. The entity that controls the chip, the cloud, and now the orbital infrastructure controls the stack from bottom to top. A 2027 mission will test high-bandwidth free-space optical laser inter-satellite links between two satellites. That is the precursor to distributed orbital compute. The long-term vision involves clusters of satellites carrying dozens of TPU chips each, connected by laser arrays. That is a network. But it is not a blockchain network. It is a private network. The data does not settle on-chain. The compute does not settle on-chain. The only on-chain component might be payments. That is not a DePIN revolution. That is a centralized cloud in space. The ledger never lies, only the narrative does. The 15-minute test is not proof of orbital data centers. It is proof that TPUs can survive launch and radiation for at least 15 minutes. That is a long way from a 1-gigawatt orbital compute farm. The bottleneck is thermal physics. Heat in vacuum requires radiators. Radiator mass scales with heat rejection. At 1 kilowatt, it is manageable. At 1 megawatt, it is a structural problem. At 1 gigawatt, it is a fantasy. The other bottleneck is launch economics. $200 per kilogram is not a price. It is a target. Until Starship achieves routine, high-cadence, low-cost launches, orbital compute remains a research project. The third bottleneck is regulation. Spectrum allocation, orbital debris mitigation, export controls on advanced chips, and liability for collisions are unsolved. The fourth bottleneck is latency. For many AI workloads, latency matters. For training, it matters less. For inference, it matters more. Earth’s surface is still better for latency, maintenance, and regulation. The crypto market will likely conflate Google’s R&D with decentralized compute. Tokens with “space” or “orbital” narratives may pump. That is a trap. The ledger never lies, only the narrative does. The 15-minute test is not proof of orbital data centers. It is proof that TPUs can survive launch and radiation for at least 15 minutes. That is a long way from a 1-gigawatt orbital compute farm. Alpha hides in the variance, not the volume. The variance in this case is the thermal data. How fast do the chips heat up? How long does cooldown take? What is the radiation-induced error rate? Those are the variables that determine economic viability. Not the press release. Not the token price. Based on my audit experience, I always cross-reference token supply schedules with project roadmaps. Here, cross-reference Google’s launch schedule with its cost parity timeline. The launch is October 1. The cost parity target is the mid-2030s. That mismatch is the trade. The token narratives will resolve in hours. The orbital infrastructure will resolve in decades. In the 2022 Terra Luna collapse, I reduced exposure to algorithmic stablecoins by 40% based on a pre-crash audit of code dependencies. I spent six weeks analyzing reserve proofs and on-chain redemption delays. The lesson was mechanical: code and liquidity are the only constants. In orbital compute, the constants are thermal limits and launch costs. The variables are hype and token price. Trust is a variable I do not solve for. Due diligence is the only hedge against chaos. The bear market rewards survival, not stories. A 15-minute compute window is a story about engineering. A DePIN token pumping on that story is a liability. Watch the telemetry, not the ticker. On October 1, watch the launch. But do not trade the launch. Watch the post-launch telemetry. If Google reports a 15-minute compute window with a 45-minute cooldown, the duty cycle is 25%. That is a signal. If the 2027 laser link demo succeeds, that is a stronger signal. If launch costs fall below $200 per kilogram, the economics change. Until then, orbital compute is a capital allocation decision by a hyperscaler, not a market. For crypto, the signal is negative for DePIN compute tokens in the short term. It shows that the real compute buildout is happening in private, vertically integrated silos. It is not happening on-chain. The next-week signal: Google’s post-launch telemetry and any FCC filings for orbital data center spectrum. The 15 minutes is the signal. The decade is the timeline. Not a trade. A watch item.