Jensen Huang and Michael Dell on the era of useful AI

Jensen Huang, NVIDIA founder and CEO made an appearance at Dell Technologies World 2026 in Las Vegas during Michael Dell, Chairman and CEO of Dell Technologies’ keynote session. Both leaders shared an interesting moment on the stage as they discussed the growing impact of AI to businesses around the world.

With agentic AI now driving the demand for compute, Huang shared that the amount of computation has grown 100X to 1,000X, depending on the work being done and the query as well. But the reality for Huang is that that it’s all about speed and capabilities today and companies are now seeing a gigantic leap in computation requirements.

“Every enterprise, our company, your company, everybody is using agents all over the place to do software development, DevOps, SRE (site reliability engineering), all of our CI/CD (continuous integration/continuous delivery) work, QA (quality assurance) testing. The amount of software work that we do in our company now supported by agents is incredible,” Huang said.

For example, Huang stated that seeing an engineer work with an agent is great. At the same time, a really great engineer in the future is going to be orchestrating a whole bunch of agents.

“We’re going to be orchestrating a whole bunch of sub-agents to do work. We’ve now arrived at the era of useful AI, which is just really exciting for all of us because until now it’s been novel,” he said.

The need for more compute

Huang, who shared how his friendship with Dell remains strong over the years, also mentions that both companies have been working on a whole new line of computers to run agents.

“The way to think about agents is this. There's a large language model, it's gigantic, it's the most computationally intensive piece of software the world's ever known. And now we're running that system towards you. That MPLink 72, the world's largest scale-up single domain computer, is just one giant system operating as one computer. And that system has a large language model in it, which can be anything between one terabyte to ten terabyte parameters. And so that's one giant system, that's the brain. However, an agent starts with the harness,” Huang pointed out.

For Huang, the harness has to sit in a secure and governed container.

“We call that a sandbox. So the NVIDIA OpenShell, open source sandbox, is the security system that just about everybody in the industry is using. Inside, we have a reference harness we call NemoClaw. NemoClaw runs on a CPU. It can also run NVIDIA Nemotron or any of the models, open source models that you would create your own specialized agents for your own companies, trained for your own special domain of data or skills. And that would run locally if you like. And then the large language models could run in the cloud,” he explained.

What makes this hybrid AI unique is that NVIDIA's architecture is the only architecture in the world that runs every frontier AI model.

“So now we support every single frontier model, we support every open source model, and we support it in the cloud as well. As you can see, our computer is the first part in the world that runs in every cloud, but it also runs local. And if you have models that you're quite sensitive about, our systems are built with confidential computing. That way, you don't have to trust the operator that's operating the data center with your secure data. So all of these architectures now run in every cloud, runs every model, runs hybrid AI, and runs agentic systems,” he added.

The economics of the AI world

At DTW 2026, Dell unveiled the Dell PowerEdge XE9812, built on NVIDIA Vera Rubin NVL72, which is capable of delivering up to 10x lower cost-per-token than NVIDIA Blackwell for massive-scale agentic AI inferencing. Dell also introduced the PowerEdge XE9880L, XE9885L and XE9882L servers which are the first Dell systems built on NVIDIA HGX Rubin NVL8, supporting up to 144 GPUs per rack with 100% direct liquid-cooled compute nodes and up to 5.5x the performance of HGX B200.

Dell is also introducing Dell PowerRack, a fully integrated system – compute, networking and storage engineered as one – with thermal design, power management and software optimization built to work together from the ground up. The result is accelerated AI and high-performance computing workloads at enterprise scale, without the integration overhead of component assembly.

On the CPU side, Dell PowerEdge M9822 and R9822 servers bring NVIDIA Vera CPUs to the enterprise AI factory. Purpose-built for agentic AI, Vera runs data pipelines, analytics, sandboxed tools and code workloads where each step waits on the last.

With 1.2 TB/s memory bandwidth and predictable performance under load, Vera completes agentic workloads 50% faster than x86 processors, helping PowerEdge systems increase AI factory output with faster agent responses and shorter feedback loops.

For Huang, while the CPUs of the past were built for hyperscale clouds, companies today are optimizing as many CPU cores as possible.

“You're not renting CPU cores anymore. You're generating tokens and marketing them. That's the economics of this AI curve. And so, the AI wants to generate, run its work, and generate the right tokens as quickly as possible, because that's the element of the intelligence. NVIDIA Vera CPU has the highest single-threaded performance of any CPU in the world. It has three times the memory bandwidth of the fastest CPU in the world. And as a result, databases like Starburst and DuckDB can run incredibly fast because the agents are pounding on the databases. So, the CPU is going to be super fast as the agents want to get through its work,” Huang said.

This is why Huang believes the relationship with Dell is key as they have the infrastructure needed to support the demand.