This video transcription presents a strong argument regarding the future of AI compute and its potential implications for SpaceX.
Here's a summary including every nuance:
The speaker references **Chamath's** assertion that the demand for AI compute, whether terrestrial or orbital, is fundamentally irrelevant because the demand for compute itself is essentially **infinite**. The speaker expresses strong agreement with this, clarifying that this is ultimately a demand for intelligence, which is anticipated to be **infinite for the foreseeable future**.
A pivotal point is made: **if Starship works**, orbital AI compute will provide the **same outcome** at a **much lower cost** than Earth-based compute. This cost advantage is projected to widen significantly.
**Gavin** is cited as pointing out that as Starship continues to **improve and be more refined**, its cost per launch will consistently **decline**. Conversely, Earth-based compute is expected to trend in the **opposite direction**, becoming **even more expensive**. This creates an increasing **divergence in cost** between orbital AI compute and Earth-based compute.
The conclusion is drawn based on several conditions: **if this is true**, **if Starship works**, **if the overall thesis is on point**, and **if demand is near infinite for orbital AI compute**. Additionally, considering that there is **only one company with Starship** (SpaceX) and **only one company pursuing TerraFab** (which is technically a joint venture of two companies), the implications for "SpaceX AI" are deemed **truly profound, astronomical, and financial.**