Nvidia launches cloud AI playground, moving further into direct cloud computing
Jun 25, 2025
Key Points
- Nvidia launches DGX Cloud (rebranded DGX Lepton) to compete directly with AWS, Azure, and GCP, shifting from chip supplier to cloud provider to capture infrastructure margins.
- AMD gains ground with neocloud partners like Coreweave by positioning as chip-only vendor, while Nvidia's cloud entry risks alienating the same startups it needs to scale.
- Cloud incumbents defend turf through custom silicon while Nvidia faces pressure from multiple directions, fragmenting a market long dominated by generalist hyperscalers.
Summary
Nvidia is launching DGX Cloud, rebranded as DGX Lepton, to compete directly with AWS, Azure, and GCP. Rather than simply selling chips to cloud providers, Nvidia is now attempting to capture cloud margin itself.
Nvidia has invested in AI-focused cloud startups like Coreweave and Lambda, but those companies have yet to make a dent in the incumbents' dominance. AWS alone generated over $107 billion in sales last year. Nvidia's real leverage comes from a different angle. As AI workloads increasingly dominate cloud demand and Nvidia remains the principal supplier of AI chips, the company has leverage to move upstream into cloud provision.
That leverage is creating strategic tension for cloud startups. AMD is positioning itself as a chip-only vendor and winning over the neocloud ecosystem (Coreweave, Lambda, and others) by not competing with them. Nvidia, by contrast, is alienating potential partners by entering their market directly. AMD's MI355 chip competes favorably on performance-per-TCO against Nvidia's H100, though it lacks the rack-scale capability of Nvidia's GB200.
The incumbents are protecting their own turf through proprietary silicon. Google has the TPU, which competes with Nvidia's H200. AWS and others have built custom inference chips. Nvidia now faces pressure from multiple directions: custom silicon from big cloud providers, a growing AMD alternative for cost-conscious buyers, and its own move to capture cloud margin, which risks alienating the neocloud partners it needs to scale beyond its own infrastructure.
The neocloud businesses remain constrained by their focus on AI computing and pale against AWS's scale. If Nvidia succeeds in moving cloud-native, the strategic map could shift. Cloud would no longer be dominated by generalist providers, but fragmented between hyperscalers and AI-specialist operators backed by Nvidia's chip advantage.