Since the start of the generative AI boom, demand for Nvidia GPUs has been insatiable. Nvidia is still raking in cash from its AI infrastructure, but there are few recent developments indicating that the industry is maturing a bit.
In an announcement largely overlooked amid Nvidia CEO Jensen Huang's keynote at Computex, the company launched Nvidia DGX Cloud Lepton, an AI platform and marketplace that will connect developers to available GPUs.
Nvidia isn't providing the service, but has lined up a set of cloud partners including CoreWeave, Crusoe, Firmus, Foxconn, GMI Cloud, Lambda, Nebius, Nscale, Softbank Corp. and Yotta Data Services. These providers are offering Nvidia Blackwell and the rest of Nvidia's stack. GPU marketplaces including Fluidstack, Foundry and Hydra are also participating.
The marketplace "connects our network of global GPU cloud providers with AI developers," said Huang, who added the Nvidia and partners are "building a planetary-scale AI factory." Huang said leading cloud providers and other GPU marketplaces are expected to participate in DGX Cloud Lepton.
Developers would purchase GPU capacity directly from cloud providers or bring their own compute clusters. GPU capacity can also be deployed across multi-cloud and hybrid environments.
With the marketplace, Nvidia is trying to democratize GPU access a bit. After all, Microsoft, Meta, OpenAI, AWS and Google Cloud are likely gobbling up the capacity.
The other Nvidia move that made me go hmm was NVLink Fusion, which enables other CPUs to plug into the Nvidia stack. Speaking at his Computex 2025 keynote, Nvidia CEO Jensen Huang unveiled NVLink Fusion. NVLink Fusion allows cloud providers and presumably sovereign AI efforts and ultimately private infrastructure to use any ASIC or CPU to scale out Nvidia GPUs.
MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys and Cadence are the early adopters of NVLink Fusion for its custom silicon. Qualcomm also announced its data center efforts and moves to integrate its CPUs into Nvidia infrastructure. Fujitsu and Qualcomm CPUs can also be integrated into Nvidia GPUs via NVLink Fusion.
Constellation Research analyst Holger Mueller said: "Nvidia once again acknowledges the importance of the network for AI. The speed and efficiency how data is served to the precious and inexpensive CPUs is what matters."
The other development in AI infrastructure this week was revealed at Dell Technologies World. Dell outlined a plan to disaggregate data centers and use AI and automation to create a stack that can better mix and match layers, vendors and AI workload architectures.
Dell's disaggregated approach to the data center will apply to private cloud and on-premises deployments.
And finally, Dell is diversifying its AI factory approach. Yes, Nvidia's full stack is the headliner for Dell AI factories, but the company added an AMD version and early-stage Intel offerings.
Add it up and there seems to be some acknowledgement that there will be Nvidia diversification whether it's with custom processors from AWS, Microsoft and Google Cloud or custom ASICs and rival efforts. Even Nvidia's NVLink Fusion acknowledges the broader ecosystem at least for CPUs. Nvidia is playing the long-game, which revolves around the ecosystem and connecting AI infrastructure with its platform.
