SpaceX Signs New Computing Power Lease Agreement with Reflection, Paying $150 Million Monthly to SpaceX
High-end AI computing power is in long-term shortage, with most AI companies relying on leasing to enter the market; Tesla's own computing cluster supports rapid iteration of FSD.
In market dynamics, the surge in demand for computing power leasing is driving diversification of revenue for aerospace companies like SpaceX, with capital leaning towards AI infrastructure leasing platforms, while building their own clusters becomes a core competitive advantage for AI companies.
Source: Public Information
ABAB AI Insight
SpaceX has previously signed similar agreements with companies like Anthropic, and this collaboration with Reflection continues its expansion in the AI computing power business, akin to Microsoft's Azure providing computing power for OpenAI.
From a capital perspective, the $150 million monthly revenue secures long-term cash flow, with funds directed towards the expansion of SpaceX's data centers, while Tesla's own cluster case highlights the advantages of vertical integration in AI training speed.
Similar to Google's TPU building its own cluster to accelerate Gemini iterations, the current AI industry is at a watershed moment transitioning from reliance on leased computing power to self-built infrastructure, with leasing agreements providing stable income for infrastructure providers.
Essentially, this reflects technological substitution and capital concentration, with the AI computing power shortage driving cross-industry collaboration between aerospace and technology, shifting pricing power from GPU suppliers to computing cluster operators, and accelerating vertical integration of AI infrastructure.
ABAB News · Cognitive Law
Computing power is the lifeblood of AI, self-building is about control, leasing is the entry ticket, and vertical integrators lead in speed. Demand expansion creates scarcity, leasing agreements lock in cash flow, and infrastructure providers reap cyclical dividends. AI companies that cannot build clusters will lease, while Tesla's self-building accelerates iteration, with pricing power determined by cluster owners who can match supply and demand.