SPACEXAI This is a very, very big deal. Space is turning from MONEY LOSING to a MONEY MAKING proposition. ♦️Google is Buying Space on SpaceX for AI Space Compute. This time on Falcon, soon it will be Starship. This is the start of the space economy flywheel. This is the economic proof of the thesis: AI Space Compute increases Demand to Orbit. Increased Demand lowers costs. Lowered costs = more viable businesses. Repeat. The Space Economy is starting. This is investable proof of concept. You heard it here first. What do you think the next signal will be?

Sundar Pichai is being slightly disingenuous asking: Can our TPUs survive and operate in space? He has very good reason to think that TPUs will work in space: Project Suncatcher. Project Suncatcher is Google’s research moonshot on the viability of space compute and if the economics could check out. Radiation isn’t the biggest expected challenge. Engineering work remains on Thermal management and on-orbit reliability. They demonstrated high-bandwidth free-space optical links with off-the-shelf products, and, most importantly, they simulated the effect of space on TPUs. Five years is the anticipated lifespan of AI space compute. The findings were illuminating: ♦️ Long-term radiation, the type that builds up from years of deployment, first showed ill effects at 3x the expected 5-year radiation dose. ♦️ The bit-flip problem remains solvable. First, they didn’t demonstrate hard failures, even with a total ionizing dose of 15 krad(Si) on a single chip. There was some silent data corruption. The solution will be layered. AI models are noisy and can tolerate some noise, especially at inference. The same techniques that mitigate cosmic radiation and bit flips in airliners can be layered on top and expanded, like shielding, ECC, monitoring, scrubbing, restarts, and fail-safes. Alphabet understands AI. They created the transformer, built super-compute at scale, designed their own chip, and have a world-class model. Alphabet has been raising capital through bonds and an $80–85 billion share issuance. Capital will be needed to fund these ambitious projects. Alphabet’s implicit endorsement of space compute makes that capital easier to raise at better terms, not just for Alphabet but @SpaceXAI , @Tesla , and @Starcloud_ too, which accelerates timelines. Taking Project Suncatcher from research moonshot to a physical test is a turning point. It is an investable proof of the AI space compute flywheel.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ Elon Musk says that space compute will trend to 100%. This is the first step.

Starship is being monetized already. The first customer is SpaceX itself. Starship will take over the lift to orbit job from Falcon - at better economics (with just a reusable booster and far better economics with a double catch).