The figures revealed by Elon Musk indicate that demand for AI infrastructure is far from reaching saturation. So, who stands to benefit the most? 👇
Let’s first look at the figures from the $SPCX balance sheet:
▪️ Total revenue: $14.6 billion (+161%)
▪️ AI revenue: $7.6 billion (+830%)
▪️ AI infrastructure investment (CapEx): $16.4 billion (+245.5%)
What caught my attention even more than the revenue growth are the comments from company management regarding the demand for AI computing power.
1️⃣ $NVDA – Even older GPUs are being leased at a premium!
According to SpaceX management, H100 GPU systems—which hit the market in 2022—can now be leased at higher prices than they were when they first launched.
Just think about that: demand for four-year-old technology remains this strong!
What does this mean for the next generation of "Vera Rubin" systems?
In my view, it serves as a significant signal regarding NVIDIA’s pricing power and the sustained demand for next-generation GPUs.
2️⃣ $CRWV – The demand is there, but what about the financing?
CoreWeave’s second-quarter revenue grew by 112.5% to reach $2.6 billion, with a reported order backlog of $104 billion.
According to Musk, if SpaceX were to accept all incoming requests, it could lease out its current capacity several times over.
Furthermore, CFO Bret Johnsen notes that, at current price levels, the payback period for investments in new AI computing capacity is less than one year.
This is a strong signal of robust demand for GPU cloud providers like CoreWeave. However, this does not eliminate the risks associated with high leverage, financing costs, and customer concentration.
So, what do I take away from this balance sheet?
AI investments are no longer just about future expectations. They are now being evaluated based on actual demand for capacity, revenue growth, and the returns on investment.
But there is an important distinction to keep in mind:
Strong demand for AI infrastructure does not mean that every company in this space will deliver equal gains to its shareholders.
For me, this is the key difference between NVIDIA and CoreWeave: one provides the core technology for the infrastructure, while the other transforms that technology into a service—a process that requires significant capital and financing.
The AI story is far from over. But moving forward, it is essential to look beyond mere growth and examine how that growth is being financed and how much cash it is actually generating.
In your view, who are the real winners of this era—the GPU manufacturers, or the companies that lease out GPU capacity?
Not investment advice.
