Most nations' sovereign AI strategies will not involve creating frontier models from scratch. Instead, they will adopt the best open-source models, customize them with local data and values, and run them on-premise for national security.
The strategy of investing in companies that represent bottlenecks in the AI supply chain is becoming saturated and is likely nearing its end. The next market phase requires identifying businesses with sustainable, long-term franchise value that will thrive once supply constraints ease.
Once Starship is fully reusable, orbital computing becomes economically compelling. A terrestrial gigawatt costs ~$60B, with ~$25B for power and cooling which space avoids. Even with a ~$5B launch cost, the total for an orbital data center becomes significantly cheaper.
Beyond its grand vision of space colonization, SpaceX's immediate financial performance hinges on two key variables: the speed of its terrestrial data center build-out and the market adoption of its Cursor AI model. These are the tangible drivers for the next year.
Contrary to the venture ecosystem's belief, public markets often support long-term investment cycles, as seen with Tesla and Amazon's build-out phases. The market is more patient with companies making strategic, long-horizon bets than it's given credit for.
China's AI lag isn't just from US sanctions; it's a strategic error of believing domestic chips are adequate. Their labs excel at distilling Western models, but this parasitic strategy fails completely if frontier models are no longer released openly.
The market's valuation of Meta, when viewed through the lens of Enterprise Value to Net Property, Plant, and Equipment (PP&E), reveals deep skepticism. This metric suggests investors doubt Meta's ability to effectively monetize its vast data center infrastructure for AI.