SpaceX's Ambitious AI Infrastructure Plans and Tesla Merger Implications

Brad Gerstner, founder and CEO of Altimeter Capital, highlighted the extraordinary nature of SpaceX's recent announcements regarding its AI infrastructure ambitions. He noted that the company plans to build 200,000 GB300 chips per month, a figure that just a few months prior represented an entire year's worth of production.

SpaceX aims to achieve 3 gigawatts of computing power by the end of the current year and an ambitious 10 gigawatts by next year. This projection means SpaceX alone would account for 25% of the total 43 gigawatts of computing power forecast to be built in the United States next year. The total capital expenditure required for this infrastructure is estimated to be around $400 billion.

Gerstner posited that Elon Musk understands the world is on the brink of the singularity and entering the age of super intelligence. In this new era, success is dictated by whoever possesses the most scale in terms of capital, computing power, and data.

According to Gerstner, while smaller firms like Coreweave are talented, they lack the necessary scale compared to a merged entity of SpaceX and Tesla. He suggested that one of the primary motivations behind a potential merger of Tesla and SpaceX would be to achieve this critical scale, accelerating growth and securing more favorable capital costs for the enormous investments required.

Brad Gerstner discussed the fiduciary considerations for a special committee evaluating a potential Tesla merger. He noted that typical control premiums for such transactions range from 20% to 30%. However, given Elon Musk's involvement in both companies, a special committee would have a unique fiduciary duty to ensure fairness for Tesla shareholders.

Gerstner suggested that the committee might lean towards a higher control premium—potentially exceeding 30%—to mitigate against inevitable lawsuits. He estimated a potential price per share for Tesla in the range of $420 to $550, arguing that going above the standard market rate for a control change would help protect against legal challenges and would be considered fair given the circumstances.

Gerstner addressed the market's ongoing questions and caution regarding 'Neo-cloud' providers like CoreWeave. Despite Elon Musk's recent comments suggesting extremely high demand, the broader market remains hesitant to fully embrace this view.

He explained that while current demand might appear strong, investors face uncertainty when forecasting demand 12 months into the future. This inability to predict long-term demand accurately means that market participants require a risk premium when financing companies heavily reliant on specific, unproven hardware, especially if those companies have limited collateral beyond GPUs themselves.

Brad Gerstner emphasized the financial advantages of using Nvidia's GPUs. He stated that it is significantly easier to finance an Nvidia GPU because of its high fungibility and broad utility across a wide range of users.

Conversely, custom ASICs (Application-Specific Integrated Circuits), often utilized by data center operators like CoreWeave, are much more challenging to finance due to their specialized nature and more limited market. Gerstner noted that Elon Musk's strategy relies almost exclusively on Nvidia hardware, further underscoring the liquidity benefits of Nvidia's products.

Gerstner highlighted Nvidia's exceptionally high profit margins, which have reportedly reached 86% in some instances, as seen with Micron. He noted that such margins typically act as a 'siren call' for new competitors to enter the market and drive down costs.

However, he pointed out that despite these high margins persisting since 2023, the competitive landscape has not significantly shifted. Gerstner argued that these margins are a direct result of genuine supply constraints and physical bottlenecks in the production process, rather than mere price gouging. He suggested that those who have predicted an 'oversupply' for the past two years have been consistently wrong, missing substantial market upside.

Gerstner contrasted the current AI boom with the dot-com bubble of 2000, pointing out that bearish predictions of an 'oversupply' in AI have been incorrect for two years, causing investors to miss out on significant gains, including Nvidia's 18-fold growth over the last three years. He argued that the present situation is fundamentally different due to inherent supply limitations.

He explained that in 2000, demand was limited, with only 30 million internet users. Today, the demand for 'super intelligence' is effectively limitless. However, there are hard physical constraints on supply, such as the production of wafers and the availability of energy, which limit how quickly AI infrastructure can be scaled. This makes the current crisis a supply-side issue, not a lack of demand.