Starship's Orbital AI: Cheaper Compute, Infinite Space Power #shorts

演讲者提出，如果星舰（Starship）投入运营，轨道增强现实（AR）计算的成本将远低于地球上的计算。预计这一转变将满足对计算“近乎无限”的需求，而太空中的计算成本将比地球上便宜得多。 演讲者表示，这种情景将天然地有利于SpaceX及其特定的计算堆栈和决策流程，使其相对于现有的超大规模运营商和其他竞争对手处于优势地位。因此，与SpaceX的轨道产品相比，地球上（特别是超大规模运营商）产生的输出令牌的成本预计将变得“经济上失衡”。 然而，这里引入了一个关键的细微之处：一旦SpaceX的轨道计算达到规模，任何留在地面的计算基础设施都将变得“极其、极其有价值”，就像“钻石”一样。演讲者强调，这种价值取决于规模。像“10千瓦钻石”这样的小规模地面计算，被视为微不足道（称之为“一文不值的碎石”），在商业上毫无意义。相反，只有运行在“数百兆瓦到吉瓦”规模的地面计算设施才真正有价值，可与“希望之星”（Hope Diamonds）相媲美。

The speaker posits that if Starship becomes operational, orbital AR compute will drastically undercut Earth-based compute in terms of cost. This shift is predicted to meet an "essentially infinite" demand for compute, which will be significantly cheaper in space than on Earth. This scenario, according to the speaker, will inherently favor SpaceX and their specific compute stack and decision-making processes, putting them at an advantage over existing hyperscalers and other competitors. Consequently, the cost of an output token generated terrestrially, particularly by hyperscalers, is expected to become "economically lopsided" when compared to SpaceX's orbital offerings. However, a critical nuance is introduced: once SpaceX achieves scale with its orbital compute, any compute infrastructure that remains on the ground will become "incredibly, incredibly valuable," likened to "diamonds." This value, the speaker emphasizes, is contingent on size. Small-scale terrestrial compute, such as a "10 kilowatt diamond," is dismissed as negligible ("a little pebble of shit") and commercially irrelevant. Instead, only terrestrial compute facilities operating at "hundreds of megawatts to gigawatts" will be truly valuable, comparable to "hope diamonds."