TechGray CA
3.8K posts

TechGray CA
@ca_techiegray
Social and causal thoughts 😎


Six Sigma vs. Lean supplychaintoday.com/six-sigma-vs-l… Lean Manufacturing–Cheat Sheet supplychaintoday.com/lean-manufactu… How Toyota used its Supply Chain Model to beat FORD & GM supplychaintoday.com/how-toyota-use… What is Lean Six Sigma supplychaintoday.com/lean-six-sigma… #sixsigma #leanmanufacturing #ContinuousImprovement









We set out to estimate the AI revenue stack for our internal modelling. 1 GW of AI compute generates $41.6B at steady state, with a 2.5x spread in what SpaceXAI could capture, depending on which layers it occupies. Jensen Huang at GTC 2026 framed AI economics as: revenue equals tokens per watt times available gigawatts. Power is the binding physical constraint but where most revenue is captured is a different question. We built a bottom-up stack model of what one GW of AI compute is actually worth in terms of revenue. We estimate 1GW generates ~$40B in annual revenue at steady state. The stack distributes: → L5 Orchestration: $5.0B (12.0%) → L4 Model: $20.1B (48.3%) → L3 Infrastructure: $9.2B (22.1%) → L2 Chips: $6.6B (15.9%) → L1 Energy: $0.7B (1.7%) I would have included a chart of the above as the image for this post, but the thumbnail I created for this piece is so hot I simply had to pick that. The model layer alone captures more value than chips and infrastructure combined, although it has some of the smallest margins, which we will explore next week on the cost side. SpaceX is positioning across three layer combinations simultaneously: → Wholesale compute (the Anthropic-Colossus structure). L1+L2+L3. $16.5B per GW, 40% of the stack. → Hybrid with application partner (the reported Cursor structure). L1+L2+L3+L4. $36.6B per GW, 88% of the stack. → Full-stack vertical integration (Grok with improved orchestration,which seems to be on the way with new integrations and skills...or an acquisition of Cursor). L1 through L5. $41.6B per GW, 100% of the stack. Pursuing all three in parallel gives SpaceX flexibility to monetize the compute supply it is building as it scales. Same dynamic as Falcon 9 with Starlink: external customer launches absorbed slack capacity as launch scaled, while internal allocation to Starlink trended up over time. Flexibility matters even more in orbit, where compute supply is harder to redirect once deployed than terrestrial capacity. Full breakdown will be public for 33 hours🧐: research.33fg.com/analysis/break…





The quantity of code that devs ship has roughly 10xed. But net developer productivity (value created by unit of time) is only up by a bit, if at all. Part of it is that the additional code is solving more incremental problems. A bigger part is that the new code is creating problems of its own.













Three Operational Changes That Actually Move the Needle on Sustainability #greensupplychain #sustainability hubs.la/Q04g1Gk00







