The SpaceX AI revenue projections put out by Morgan Stanley and Goldman are intentionally fraudulent and they harm retail investors.
They believe the system allows this due to the Spitzer case which supposedly "separates" analysts claims from the S-1. The S-1 is the CYA document. We all know however retail investors rely on the same firm's sell-side analysts who are paid to model reality. They don't know these people can make false claims without recourse.
Earlier Gemeni acknowledged the Google GPU lease rate was 8x market. I asked it why Google did it and it said it was because Google owned SpaceX and it helped the IPO Valuation.
Here is the Gemeni's analysis of Goldman and Morgan Stanley's 2030 AI revenue projections:
Goldman Sachs projects $322B and Morgan Stanley $190B in SpaceX AI revenue by 2030—a 60–100x jump from $3.2B in 2025. This segment underpins the vast majority of SpaceX’s trillion-dollar IPO story.
The GPU Math: Temporary premium lease rates must eventually decline to market inference rates (~15,000–21,000/GPU/year). To hit analyst targets at these market rates, SpaceX would need 10 to 18 million GPUs—a 45–80x expansion from its 220,000-GPU fleet today.
Power and Grid:
Running 10–18M GPUs requires 20 to 40 GW of power. This exceeds entire national grids and faces 4-to-7-year interconnection delays, making it physically impossible to build by 2030.
Capital & Supply:
AI Build costs of $1 trillion to $2 trillion dwarf SpaceX’s $75B IPO. Furthermore, rival hyperscalers have already locked up NVIDIA allocations, CoWoS packaging, and HBM memory through 2027.
Building this AI capacity is physically impossible even with $1 trillion in capital, and the analyst understand the manufacturing and power constraints well.
My claim: Goldman Sachs and Morgan Stanley are committing fraud via a system they believe legally insulates them from liability. They are intentionally using nuances created by the Spitzer case to harm retail investors.