GTC 2026: Jensen Huang Confirms What I Predicted Three Weeks Ago — SaaS Is On Its Way Out

Three weeks ago I published an article with a thesis that raised a few eyebrows:

"The SaaS model is on its way out. AI agents will replace traditional software — and it's happening faster than anyone expects."

Tonight, March 16, 2026, Jensen Huang opened his GTC keynote with a slide titled "Enterprise IT Renaissance: From SaaS to Agent-as-a-Service".

That's not a coincidence. It's a pattern.

What I wrote three weeks ago

In the article «The End of SaaS? How I built a modern AI layer on top of a 30-year-old ERP system» I described something very concrete: I had built 13 specialized AI agents running in production — not as a demo, but as actual infrastructure in daily operations. These agents integrated directly with legacy systems, made decisions autonomously, and communicated with each other.

The thesis was simple: When you can build an AI layer that does what a SaaS application does — but tailored, without subscription fees, and integrated into your existing data — the argument for SaaS weakens dramatically.

What Jensen Huang said tonight

From the GTC 2026 stage in San Jose:

• "Enterprise IT Renaissance" — not evolution, not adjustment. Renaissance. A new era.
• "From SaaS to Agent-as-a-Service" — direct phrasing. SaaS is the starting point; Agent-as-a-Service is the direction.
• Data Centers → AI Factories — the infrastructure is being rebuilt. Not for storage or computation, but to produce intelligence at scale.
• Tokens replace files — the fundamental unit of IT is shifting. Not documents, not databases. Tokens. Inference.
• $1 trillion in computing demand — this is no longer niche economics.

And then: NemoClaw.

NemoClaw — the enterprise version of what I'm already running

NemoClaw is NVIDIA's enterprise-ready agent architecture, built on OpenClaw. From the keynote slide: sub-agents, skills, MCP integrations, memory and orchestration — exactly the architecture I've been running with JP Claw since February.

For those who wondered whether what I built was a hobby project: NVIDIA used the same thing as reference architecture at the world's largest AI conference.

I'm not making that point to brag. I'm making it because it means the technology is mature. What three weeks ago seemed experimental is tonight enterprise standard.

What this means for business leaders

Let me be direct, without tech jargon:

The SaaS tools your company pays for today — CRM, HR systems, project management, reporting — are about to be outcompeted by AI agents that do the same thing, better, for a fraction of the cost.

It won't happen overnight. But it will happen faster than the last technology shift.

Companies like SAP, Salesforce and ServiceNow see this. They're rebuilding. But they struggle with the same thing all large companies struggle with: legacy, complexity, and investors who expect returns from existing products.

Startups and internal teams without that baggage will deliver agent-based solutions that eat the market from below.

The question for business leaders isn't whether this happens. It's whether you're positioned to benefit from it — or whether you'll still be locking yourself into new SaaS agreements three years from now.

My take

This isn't prophecy. It's pattern recognition.

Everyone working with AI in production — not in demos, not in PoCs, but actually in operation — sees the same pattern. AI agents aren't a layer on top of existing software. They are the software. And when an agent can learn, adapt and integrate directly with your data, the argument for buying pre-packaged functionality from a SaaS vendor disappears.

Jensen Huang said it from the biggest stage in the AI world tonight. That's not a signal to ignore.

Joachim Høgby is CTO and runs AI infrastructure for Norwegian companies through hogby.ai. He runs 13 specialized AI agents in production based on the OpenClaw architecture.