Inside GTC 2026: How NVIDIA’s Six Chip Megastack Signals a New Era — and What It Means for RS

NVIDIA’s GTC has always been one of the must-see events if you’re an engineer, but 2026 felt like a milestone, a point where AI’s future finally snapped into focus. NVIDIA CEO Jenson Huang’s keynote wasn’t just a product reveal… it was a manifesto for the next decade of computing. And as the dust settled in San Jose, I sat down with Richard Curtin, SVP, Product & Supplier Solutions at RS Group, fresh off the conference floor, to unpack what mattered most… especially for organisations like RS.

The official highlight reel below reinforced the tone of the week: Jensen Huang’s keynote “lit the fuse on a new era of token‑powered, full‑stack AI,” spanning everything from CUDA to the new Vera Rubin platform, Nemotron, and OpenClaw.

Play video

A New Scale for AI: Six Chips, One Vision

“The scale of NVIDIA’s ambition was on full display,” Richard told me.

The centrepiece was Vera Rubin, a full-stack architecture rather than a single chip family. NVIDIA unveiled six chip systems, all co-designed to operate as a unified AI supercomputer, and NVIDIA put it like this: “these chips form a synchronized architecture in which GPUs execute transformer-era workloads, CPUs orchestrate data and control flow, scale-up and scale-out fabrics move tokens and state efficiently, and dedicated infrastructure processors operate and secure the AI factory itself.”

Vera Rubin (Image Credit: NVIDIA)

The keynote underscored how Rubin fits into a 20‑year CUDA legacy, a continuation of NVIDIA’s multi-decade push toward vertically integrated AI computing.

This wasn’t incremental progress. Jensen reminded the audience that modern AI compute has grown substantially in a decade, and he now expects at least $1 trillion in AI infrastructure demand through 2027.

As Richard put it:

“Rubin is NVIDIA showing the industry what an AI datacentre should look like… and frankly, it’s hard to argue with them.”

Agentic AI Arrives: OpenClaw’s ‘Operating System’ Moment

But hardware wasn’t the biggest philosophical shift.

“This was the year NVIDIA declared that agentic AI is the new computing model,” Richard said.

The keynote hammered home the same message, emphasising NVIDIA’s pivot toward systems powered by tokens as the unit of compute, enabling long‑running, autonomous agents rather than prompt‑response chatbots.

The star here was OpenClaw, described by Jensen as “the operating system for personal AI” and the fastest-growing open-source project ever! NVIDIA unveiled NemoClaw, the enterprise-grade layer providing security boundaries, so agents can’t leak data or run unauthorised code.

For Richard, this was the cultural turning point:

“OpenClaw isn’t just a tool. It’s NVIDIA signalling that agents are the successor to apps.”

NemoClaw (Image Credit: NVIDIA)

Infrastructure to Match: Dynamo 1.0 and the AI Factory Era

To support these agents at industrial scale, NVIDIA introduced Dynamo 1.0, which is now in production. It boosts Blackwell inference by up to 7×, with industry giants like AWS, Azure, Google Cloud and Oracle already committing to it.

This aligns seamlessly with the keynote framing of “AI factories as the industrial infrastructure of the AI era,” a theme Jensen emphasised repeatedly.

Dynamo 1.0 (Image Credit: NVIDIA)

In parallel, NVIDIA launched the Nemotron Coalition, a group including Mistral, Cursor, Perplexity and others, to build powerful open frontier models. Nemotron 3 is already top-three on the OpenClaw leaderboard.

Nemotron is being deployed globally. (Image Credit: NVIDIA)

Robotics, Automation, and Even Orbit

Of course, no GTC would be complete without robots (remember Grek from the 2025 GTC?), and 2026 was overflowing with them. NVIDIA revealed robotaxi deployments with Uber across the globe starting 2027, backed by partners like BYD, Hyundai, Nissan, Mercedes, Toyota, and GM.

Jensen described autonomous vehicles as “the first multitrillion-dollar robotics industry.”

The keynote drove this home by highlighting NVIDIA’s expansion of physical AI, including robotics demos and industrial automation partners.

And then came the moment that felt like a sci‑fi reveal: NVIDIA is sending a Vera Rubin module into orbit as part of Space‑1. Yes… space-based AI compute is happening.

What This Means for RS: Richard’s Perspective

After we’d dissected the announcements, I asked Richard how he expects RS to intersect with the future NVIDIA is building. His answers were optimistic.

1. Infrastructure Supply Chain Leadership

Rubin’s multi-chip, rack‑scale design requires industrial-grade supply chains, power systems, networking, compliance, and delivery at scale.

“RS already excels here,” Richard said. “As AI factories become mainstream, RS will be a vital partner in making them physically real.”

2. Supporting AI Factory Deployments

Dynamo 1.0 and DSX platforms turn datacentres into automated AI factories, and those factories need the components RS is known for.

“That’s the invisible layer GTC depends on… and RS provides it.”

3. Enabling Robotics Rollout at Scale

NVIDIA’s partnerships with FANUC, ABB, Agility, and others align directly with RS customers in manufacturing, automation, and industrial robotics.

“Robotics is about to explode,” Richard said. “RS is positioned to become a primary ecosystem supplier.”

4. Agentic AI for Engineering and Procurement

OpenClaw and NemoClaw open the door for persistent engineering agents that:

• streamline design workflows
• automate bill-of-materials generation
• run simulations
• optimise procurement and stock management

“Imagine an RS Design Solutions agent that learns your engineering habits,” Richard said.

“That is completely feasible within NVIDIA’s new agentic stack.”

5. Developer Ecosystem & Educational Kits

Rubin isn’t just for hyperscalers, developers will need kits, training systems, and accessible hardware.

“That’s where RS shines… making advanced platforms usable for engineers everywhere.”

Richard Curtin at the GTC.

Conclusion: A Future Built on Full‑Stack AI

GTC 2026 marked a true industry pivot. NVIDIA made its intentions clear: to own every layer of the AI stack, from chips to racks to agents to robots, even extending into space. The highlight video hammered home that this isn’t hype, it’s a coordinated, full‑stack execution plan.

For RS, this isn’t just a technology trend.

It’s the beginning of an era where AI becomes industrial infrastructure, a space we already understand deeply.

As Richard said while we wrapped up:

“AI is becoming physical. It needs power and components, everything RS delivers. NVIDIA is building the future, but RS is going to help build the world around it.”