SpaceX is building its AI compute infrastructure exclusively on NVIDIA’s Vera Rubin platform, running workloads both in ground-based data centers and in orbit through its Starmind AI satellite program. Elon Musk confirmed the commitment during SpaceX’s Q2 2026 earnings call, making SpaceX one of the most operationally demanding industrial AI infrastructure deals<\/a> NVIDIA has landed. NVIDIA reported $81.6 billion in Q1 fiscal 2027 revenue, up 85% year over year, with data center revenue at $75.2 billion. Consensus estimates now project $387 billion in fiscal 2027 revenue.<\/p>\n The SpaceX deal matters less as a revenue event and more as a category signal. NVIDIA is no longer just the hyperscaler’s GPU supplier; it’s becoming the default compute fabric for organizations running AI at the physical edge, meaning environments where latency, power constraints, and hardware reliability are engineering problems, not just IT ones. If your AI workloads are still entirely cloud-resident, this is the moment to assess whether that’s a strategic choice or an inherited assumption.<\/p>\n What NVIDIA is building with SpaceX isn’t a reference architecture for most enterprises, but it sharpens the vendor consolidation pressure that CTOs are already feeling. The Vera Rubin platform, NVIDIA’s next-generation GPU and systems stack succeeding Hopper, is designed to run unified workloads across locations without rewriting the software layer. That’s the actual strategic threat to AMD and Broadcom: not GPU performance benchmarks, but the switching cost embedded in NVIDIA’s CUDA software ecosystem, which developers have spent a decade building on top of. AMD’s ROCm platform is a credible alternative on paper, and AMD’s data center revenue grew 107% year over year, but “credible alternative” and “production default” are still different things for most engineering teams.<\/p>\n NVIDIA’s forward price-to-earnings ratio sits at 19.52, below the sector average of 21.26, which is a notable compression for a company projecting 90% earnings growth in fiscal 2027. Zacks, whose analysis skews toward equity positioning, flags this as undervaluation, and the framing isn’t wrong on the numbers. The more useful read for a CTO is that the market is pricing in execution risk on NVIDIA’s ability to sustain this growth rate as custom silicon from hyperscalers and Broadcom’s AI chip revenue, up 143% year over year, starts absorbing a larger share of the workload. I’d revisit that vendor diversification question if NVIDIA’s enterprise software lock-in starts eroding in fiscal 2028 estimates.<\/p>\nWhat this means for your business<\/h2>\n