CNBCs interview with early SpaceX investor Gavin Baker [HHcqWATIJKO]
Most people think the future of SpaceX depends on one thing: Starship. If Starship launches on time, the company wins. If reusability is delayed, the valuation falls apart. Technology investor Gavin Baker thinks Wall Street may be watching the wrong machine. In this full CNBC interview, Baker argues that the most important part of the SpaceX story over the next several years may not be happening in orbit at all. It may be happening inside enormous terrestrial data centers, where gigawatts of electricity are converted into AI models, tokens and revenue. His claim is striking. If SpaceX can bring new power online faster and more cheaply than its competitors, the company could monetize each gigawatt at an enormous premium. Under one hypothetical scenario discussed in the interview, three additional gigawatts could represent roughly $150 billion in revenue that may not be reflected in conventional Wall Street estimates. But why would a rocket company have an advantage in AI infrastructure? That question leads directly into the economics of artificial intelligence. When a cheaper Chinese open-source model appears, markets tend to assume that the entire AI industry will suddenly need less money, less infrastructure and fewer expensive chips. Baker believes that may be exactly backward. A cheaper model does not produce free intelligence. Every token still requires GPUs, electricity, capital expenditure and operating costs. And when intelligence becomes cheaper, people usually do not consume the same amount and pocket the savings. They consume more. So could cheaper AI models actually increase demand for Nvidia chips, data centers and electricity? Baker argues that open-source AI may not destroy the economics of the industry. It may simply transfer more of the profits from the model companies to the infrastructure beneath them. That creates another puzzle. If companies can use free or inexpensive open-source models, why would they continue paying a premium for systems from OpenAI, Anthropic or xAI? The answer may be token efficiency. A weaker model can appear cheaper per token while requiring two or three times as many tokens to reach the same result. What matters is not simply the price of intelligence. It is how much useful intelligence a company receives for every dollar it spends. Baker describes a future in which businesses use an entire orchestra of AI models. A frontier model such as Claude or Grok may act as the conductor, while cheaper specialized models perform smaller tasks in the background. But if every company uses several models at once, who controls the orchestra? The interview explores whether the real competitive advantage will sit inside the model itself, or inside the products and software built around it. Tools such as Claude Code and Grok’s surrounding ecosystem may create switching costs that are far harder to escape than the underlying model. Baker also discusses Nvidia’s surprisingly capable open-source AI models, why the company may be deliberately avoiding a direct confrontation with its largest customers, and why Nvidia could potentially produce an American open-source model near the frontier whenever it chooses. Then there is Grok. Baker argues that the market overreacted to the arrival of cheaper Chinese models while underreacting to Grok’s position on the intelligence-per-dollar frontier. It may not be the most intelligent model available, but it could be among the most economically efficient. And that may reveal Elon Musk’s real advantage. Model supremacy receives most of the headlines, but Baker believes the speed at which a company can secure power, construct data centers, energize GPUs and begin producing tokens may matter even more. Every additional day spent building is another day of labor costs, lost revenue and idle equipment. In that world, the AI race is not only a competition between researchers. It is a race between electricians, cooling systems, power grids, semiconductor companies, data center operators and billionaires trying to turn entire gigawatts into intelligence before somebody else does. Starship still matters. Bigger models still matter. Grok’s next training run still matters. But the deeper question is whether Wall Street is valuing SpaceX as a rocket company when it should be looking at something much stranger: a vertically integrated power, compute and artificial intelligence empire that also happens to launch spacecraft. This full interview with Gavin Baker explores SpaceX’s valuation, Starship delays, Chinese open-source models, Nvidia, Grok, Anthropic, token efficiency, AI infrastructure and the increasingly important battle to produce the most intelligence for the lowest possible cost. #SpaceX #ArtificialIntelligence #ElonMusk