SpaceX's AI Factory to Drive Trillions in Revenue

The facility aims for a total capacity of 2 to 2.5 gigawatts by the end of 2024, leveraging mega packs to ensure a smooth and consistent electricity supply for heterogeneous machine training and inference.

InvestAnswers estimates that renting out this compute capacity could generate over $100 billion in annual recurring revenue (ARR) by year-end, significantly boosting SpaceX's revenue streams.

With weekly installations of large quantities of GPUs, the company is outpacing its internal projections and targeting a 2.5 gigawatt AI infrastructure capacity by the end of the year, Aaron Burnett noted.

The company is installing massive quantities of GPUs into its facility on a weekly basis, demonstrating aggressive scaling to meet its ambitious capacity targets.

With 90% of total compute allocated for inference rather than the 10% used for model training, the monetization of user request processing remains the primary driver of economic viability for SpaceX's infrastructure.

Inference, defined as processing user requests token-by-token through a saved model, presents a highly lucrative monetization opportunity.

Revenue potential is estimated at $30 to $50 billion per gigawatt for external rentals, and internal usage of Grok could potentially scale this monetization value to $100–200 billion per gigawatt.

Meta's 'Muse' product quickly reached 1 million users, reportedly exhausting its available compute capacity, indicating a rapid saturation of resources at major tech firms.

This swift capacity saturation serves as a counter-argument to the notion of an AI bubble, suggesting instead that demand for AI intelligence is effectively infinite.

InvestAnswers contends that constraints on inference demand are primarily due to corporate budget cycles rather than any lack of practical utility or interest.

These key segments include rocket launches, global communications via Starlink, AI compute services, and government/defense contracts; Starlink is identified as the shortest path to a high internal rate of return (IRR) over the next three years.

Burnett emphasizes that while the engineering challenges are significant, they are not constrained by physical laws, distinguishing them from true physics limiters.

Space-based infrastructure offers distinct geographical advantages, eliminating regional terrestrial cable and tower limitations and providing a global network.

The company is concentrating on space-based connectivity to bypass the limitations and bottlenecks of terrestrial cable and tower infrastructure.

Relocating critical infrastructure to space establishes a global, stationary network that inherently reduces overall connectivity costs and expands access.

Terrestrial cell towers are often inefficient and require complex handoffs between various geographic locations, leading to less reliable service.

In contrast, Starlink terminals can connect to approximately 25 satellites simultaneously, with the capacity to scale up to 200, according to Aaron Burnett.

This satellite network design eliminates the need for devices to constantly jump between physical towers, enabling consistent, high-bandwidth switching and superior latency.

Starlink provides the essential connectivity foundation for future global operations, enabling a seamlessly integrated ecosystem.

Space offers an alternative location for data centers, circumventing terrestrial land-use permits and energy consumption constraints, according to Aaron Burnett.

The integration of Tesla's Optimus robots with Starlink's global communications capability represents a massive total addressable market (TAM), leveraging cross-company synergies.

This convergence of AI, hardware, and space infrastructure is creating a uniquely hyper-efficient and expansive technological ecosystem.

This involves increasing power generation per satellite from tens of kilowatts to 100-200 kilowatts, necessitating highly efficient cooling systems.

Utilizing sun-synchronous orbits helps maintain radiator panels pointed away from direct sunlight, optimizing their heat rejection capabilities.

Engineers must optimize heat transfer pathways from the chips to the radiators to maximize radiation efficiency and prevent overheating.

This expansion includes a massive increase in solar panel arrays compared to traditional communication satellites, providing the necessary power for intensive AI computation.

Dedicated radiator panels are required for the AI1 satellites because the standard satellite bodies are incapable of dissipating the high heat generated by AI compute operations.

Kessler syndrome, a theoretical scenario of cascading debris collisions making space unusable, is a recognized concern, but the vastness of space makes the risk lower than commonly perceived, Aaron Burnett explains.

Current satellites actively perform constant collision avoidance maneuvers, sometimes up to three times daily per satellite, to mitigate risks.

Active debris management represents an underfunded engineering opportunity rather than a technical impossibility, suggesting that solutions exist but require greater investment and development.

Monetizing compute capacity internally for SpaceX presents a higher opportunity than external revenue generation, according to Aaron Burnett.

Utilizing a gigawatt of compute for internal operations could yield $100 billion to $200 billion per gigawatt, assuming a 50% utilization rate of the compute rack.

SpaceX strategically balances its internal compute requirements against external demand to maximize overall value and resource allocation.

Demand for compute remains effectively infinite, driven by the proliferation of sovereign AI initiatives and diverse industry applications.

Despite the low adoption, users who engage with MCP demonstrate 15 times higher usage compared to traditional reading methods, indicating its efficiency.

The host attributes this low adoption rate to the initial learning curve and effort required to set up and become familiar with the protocol.

Conservative estimates suggest a stock price of $210, while more aggressive timelines aligned with Elon Musk's projections could see valuations exceeding $2,600.

InvestAnswers notes that the combined business units represent a total addressable market (TAM) of approximately $83.5 trillion, underscoring the immense potential for growth.

The model further projects a $924 target by 2030, indicating a substantial 6.18x growth, according to InvestAnswers.

Aaron Burnett validates the application of probability distributions for these valuations due to the 'long right tails' of potential outcomes, suggesting that valuing SpaceX at $1.75 to $2 trillion is akin to investing in a predictable 'Starlink bond' with significant upside optionality.