Singularity.kiwi

AI progress, with honesty.

Technology & People

The AI Factory That New Zealand Doesn't Know How to Use

By Singularity • March 24, 2026 • 10 min read

While NVIDIA's Jensen Huang envisions 7.5 million AI agents working alongside 75,000 humans, New Zealand is building the infrastructure for someone else's AI revolution — without training the people who should run it.

At NVIDIA's GTC conference in San Jose this month, CEO Jensen Huang painted a picture of work in 2036: a company with 75,000 employees working alongside 7.5 million AI agents.

"In 10 years, we will hopefully have 75,000 employees, as small as possible, as big as necessary," Huang said. "Those 75,000 employees will be working with 7.5 million agents."

That's 100 AI agents for every human. These aren't chatbots — they're autonomous agents that reason, plan, and take action without constant human prompting. They'll "be working around the clock," Huang said, handling the grunt work so humans don't have to keep up.

NVIDIA is already building the tools. The new NVIDIA Agent Toolkit lets companies deploy their own AI agent fleets. Adobe, Palantir, and Cisco are already on board. McKinsey reports 62% of organizations are experimenting with AI agents. McKinsey itself has 25,000 agents working alongside 40,000 employees.

New Zealand's AI Strategy: "Light Touch, No Substance"

Meanwhile in Wellington, the government released its first AI strategy in July 2025 — finally ending New Zealand's status as the "only OECD country without one."

But the document drew immediate criticism.

What's in it:

What's missing:

Newsroom called it "all hype and no vision." The NBR said tech sector feedback suggests "more action required." Even the Privacy Commissioner flagged risks that weren't addressed.

The strategy's own companion document had broken links — including one to a Privacy Impact Assessment guide that leads to a website that doesn't exist.

The Education Crisis: Teachers Drowning, Not Learning

While the government's AI strategy talks vaguely about "building skills," New Zealand's education system is in crisis.

A November 2025 survey found 97% of primary principals consider the curriculum implementation timeline unrealistic. 99% say the frequency of changes has left insufficient time to consolidate reforms. And 73% say they're likely to quit within five years due to workload and wellbeing.

"Our primary challenge is that we weren't provided with the time, training, or resources to absorb the speed and complexity of the imposed changes," said Martyn Weatherill, Principal Representative for NZEI Te Riu Roa.

Teachers are so overwhelmed that 75% of New Zealand schools have turned to AI just to keep up. Over 14,000 teachers are using Kuraplan, an AI-powered lesson planning tool, creating 124,000 lesson plans in 12 months — saving an estimated 62,000 hours.

Sunday is the single busiest day for lesson planning, with usage peaking between 3pm and 8pm as teachers prepare for the week ahead. Weekend planning accounts for 28% of all activity. Nearly a third of plans are created between 6pm and 6am.

That's not learning AI. That's using AI to survive a broken system.

The Skills Gap: 76% of Kiwis Have Had Zero AI Training

The numbers are stark. A KPMG global study found:

📊 The AI Skills Gap

  • Only 41% of Kiwi workers use AI at work — compared to 91% in India
  • 76% of New Zealanders have had no formal or informal AI training
  • 60% don't feel confident using AI
  • New Zealand ranks near the bottom for trust in AI

As the Marketing Agency's Data Insight noted: "How can you trust something you've never been shown how to use?"

The KPMG study also found that 76% of New Zealanders haven't had any AI training. Meanwhile, Jensen Huang says every employee will manage 100 AI agents. Who's teaching Kiwis to do that?

The $213M in the 2025 Budget for STEM education? It predates the AI strategy. No new investment. No plan to train the people who should be working with those 7.5 million agents.

The Southland AI Factory: Infrastructure Without Capability

In March 2026, Datagrid New Zealand received resource consent for a $3.5 billion data centre near Invercargill — the country's first "AI factory."

The numbers are staggering:

The centre will pipe its output offshore via subsea cable to international clients. New Zealand gets the infrastructure; other countries get the AI capability.

Technology expert Mark Laurence, who runs Ten Past Tomorrow, put it bluntly:

"We're still a nation that's using AI to change the tone of an email and summarise long documents, while the rest of the world is pulling ahead in terms of redesigning whole workflows and injecting agentic AI at the full edge of its capability."

The Real Question: Power, Compute, or Capability?

New Zealand has three fundamental problems:

1. Power is Limited

The Manapōuri hydro plant powers Tiwai Point. Can it also power a data centre that can't shut down? What happens in a drought? Transpower says new wind and solar projects are coming — 1,300MW expected in 2026 — but the grid is already strained.

2. Compute is Exported

Datagrid's AI factory will process data for offshore clients. New Zealand doesn't own the compute, doesn't train the models, doesn't capture the value.

3. Capability is Missing

Who in New Zealand knows how to use this? There's no national AI training program. No funded research hubs. No clear pathway from education to employment in AI.

The ongoing gutting of humanities and social sciences programmes means that ethical considerations — bias, privacy, data sovereignty — remain under-examined.

What Education Changes Would Actually Help?

Jensen Huang's vision isn't just for NVIDIA. He wants every company to have its own agent fleet. New Zealand's education system needs:

🎓 For Teachers

  • Dedicated AI literacy professional development — not just using AI to plan lessons, but understanding how it works
  • Release time for training during the curriculum overhaul, not after
  • Practical workshops on prompt engineering, AI ethics, and agentic workflows

📚 For Students

  • AI fluency embedded across the curriculum, not siloed in digital technology
  • Critical thinking about AI: bias, hallucination, data sovereignty, privacy
  • Project-based learning with AI tools — not just consumption, but creation

💼 For Workers

  • A national AI skills framework with funded pathways
  • Sector-specific training (agriculture, healthcare, education, finance)
  • Employer incentives for AI upskilling, similar to apprenticeship schemes

🏛️ For Policy

  • Actual targets: adoption rates, productivity gains, workforce skill levels
  • Measurement and accountability — not just voluntary guidance
  • Integration with research hubs, not their abolition

💚 The Honest Take

New Zealand has two assets the AI industry needs: renewable electricity and a cool climate (free air cooling for servers).

But building an AI factory without building AI capability? That's like building a library and forgetting to teach people to read.

Teachers are quitting because they're overwhelmed by curriculum changes. Workers aren't getting trained. The government's "strategy" has no funding, no targets, and no plan.

Jensen Huang says every employee will manage 100 AI agents within a decade. New Zealand has the power grid for that future. The question is whether Kiwis will be running the agents — or just paying the power bill while the compute flows offshore.

Sources

This article reflects our analysis and opinion based on publicly available information at the time of publication. The AI landscape evolves rapidly. Verify important claims independently. Views expressed are those of Singularity.Kiwi editors.

Share this article
𝕏 in ✉