At SoftBank's annual shareholder meeting on June 23, Masayoshi Son publicly dismissed Elon Musk's vision for a space-based data center, asserting that the outcome of the AI race depends solely on ground computing power. Behind this clash of titans lie two entirely distinct industrial logics.

Masayoshi Son crunched the numbers clearly: in the total cost of a data center, electricity accounts for only 7%, while chip hardware costs a staggering 93%. Although the space-based solution could save on electricity bills, the additional expenses from rocket launches, on-orbit maintenance, and communication delays would completely offset these savings. Industry estimates suggest that the cost of building 1GW of space-based computing power is three times that of ground-based capacity, making it completely uneconomical in the short term.

In contrast, Musk's plan extends far beyond a single data center project. SpaceX has applied to the FCC to deploy one million AI computing satellites, leveraging the low-cost transportation of the Starship, unlimited solar energy in space, and vacuum-based cooling. The goal is to achieve 1 gigawatt of orbital computing power by 2027, forming a long-term, closed-loop ecosystem integrating aerospace, chips, and large-scale models.

Their disagreement goes beyond technical approaches; it represents a vast chasm in underlying thinking.

First, Masayoshi Son embodies a pure financial investment mindset, with short-term ROI as the sole yardstick.

Reviewing his past investments, in 2019, he sold all his shares in NVIDIA, directly missing out on over $150 billion in gains, and has publicly regretted this missed opportunity in the AI boom multiple times. Now, having invested tens of billions in OpenAI, his primary role as a financial investor is to capitalize on the current ground computing power boom, with all decisions hinging on short-term returns. Although he once dubbed himself the 'architect of humanity's future,' all his strategies are anchored in capital appreciation.

Second, Elon Musk represents an engineer's long-termism, breaking free from the constraints of existing business frameworks.

His vision of space-based computing power, Mars colonization, and humanoid robots forms a cohesive system, unrestricted by current financial statements. He has repeatedly argued that future constraints on AI expansion will come from Earth's electricity and land resources, making space-based computing power the only viable long-term solution. He has even suggested that future monetary systems may need to be restructured, a narrative that transcends traditional capital accounting logic.

The two have publicly clashed multiple times before: earlier, Son's $500 billion Stargate project was criticized by Musk for funding gaps; when Son announced a trillion-dollar robotics plan, Musk dismissed it as unrealistic. One focuses on the immediate AI boom, while the other plans for decades of interstellar development.

Third, neither approach is inherently right or wrong, but they suit different types of investors.

In the short term (2-3 years), ground computing power will remain the main battleground for AI, with Son's heavily invested data centers and large-scale models continuing to deliver results. Over a decade, however, space-based computing power could overcome Earth's computing limitations, representing an ultra-long-term industrial opportunity.

This debate is not about who is right or wrong; it is a clash of perspectives between capital-driven profit seekers and technology idealists. Ordinary investors should recognize: for short-term plays, focus on ground-based chip and server supply chains; for long-term Layout (translated as 'strategic positioning'), space-based aerospace computing power will be the next major tech trend.

Which track do you favor: ground computing power or space-based computing power? Share your thoughts in the comments.

Source: Investor Network