New Zealand's AI Divide: Building Infrastructure While Skills Lag Behind
New Zealand is about to spend billions on AI infrastructure. But a growing chorus of experts warns we're building the hardware without training the people who should use it — and the gap is becoming a national problem.
Two Stories, One Problem
This week, two seemingly unrelated AI stories emerged from New Zealand. Together, they reveal a troubling disconnect.
Story one: Southland will host a $3.5 billion AI data centre — New Zealand's first purpose-built facility for AI training. It will consume 280 megawatts, roughly 6% of New Zealand's entire electricity generation. A new subsea cable will export data offshore.
Story two: A leading AI expert warns New Zealand is "AI illiterate" — using AI to "change the tone of emails" while the rest of the world uses agentic AI to redesign workflows and transform industries.
One story is about building the machine. The other is about not knowing how to run it.
The Numbers
- $3.5 billion — Investment in Southland AI factory
- 280 MW — Power consumption (6% of NZ's generation)
- 1,200 jobs — Construction employment
- $4 billion — Projected economic injection
- $60 million/year — GDP contribution
- 400,000 teachers — US National Academy's AI training target (NZ equivalent unclear)
The Southland AI Factory
Datagrid's Southland facility represents a significant infrastructure investment. The numbers are substantial:
- Power: 280MW — enough for a small city, about 6% of NZ's current generation
- Jobs: 1,200 during construction, ongoing operations staff
- Economic impact: $4 billion injection, $60M/year GDP
- Connectivity: First South Island international cable landing
The facility is designed for AI training — the intensive computation that teaches models how to think. Companies like OpenAI, Anthropic, and Google train their models on similar infrastructure.
But here's the question: who in New Zealand will use this capability? And do they know how?
The AI Literacy Warning
While the infrastructure is being built, New Zealand's AI adoption tells a different story.
An AI expert (whose name has circulated in local tech discussions) recently warned that New Zealand is "AI illiterate" — not in the sense of having no AI, but in how we use it.
What we do: Use ChatGPT to rewrite emails, generate social media posts, and summarize documents.
What the world does: Deploy agentic AI that can research, reason, and execute multi-step workflows without human intervention.
The gap isn't about access — New Zealanders can use the same AI tools as anyone else. It's about knowing what's possible and having the skills to implement it.
When NVIDIA's Jensen Huang says AI agents will soon number in the millions, he's not talking about chatbots that answer questions. He's talking about agents that can:
- Research topics overnight without human feedback
- Test hypotheses and evaluate their own results
- Decide what to work on next based on outcomes
- Coordinate with other agents to solve complex problems
This isn't science fiction. Karpathy demonstrated this recently: an AI that researches itself all night, generates hypotheses, tests them, and decides what to explore next — all without human intervention.
The Disconnect
Here's the tension: New Zealand is building a massive AI training facility, but the people who could benefit from it — local developers, researchers, businesses — may not have the AI literacy to use it effectively.
It's like building a Formula 1 track in a country where most drivers have only ever used rental cars.
The infrastructure is necessary. But without parallel investment in training, the facility becomes an export — New Zealand hosts the hardware, but the value flows overseas.
What Skills Are We Missing?
AI literacy isn't about learning prompt engineering. It's about understanding:
- Agentic workflows: How to design systems where AI agents work autonomously
- Model evaluation: How to assess whether a model is fit for your use case
- Integration: How to connect AI to your existing tools and data
- Safety and security: How to use AI without exposing your organisation to risk (see yesterday's supply chain attack)
- Cost optimisation: How to get results without burning budget on unnecessary computation
These aren't skills you pick up from a weekend course. They require hands-on experience, experimentation, and — crucially — understanding what's now possible.
What New Zealand Could Do
There are practical steps that could bridge the gap:
1. University programmes. AI courses exist, but they're often theoretical. We need programmes focused on agentic AI, model training, and production deployment.
2. Industry partnerships. The Southland facility could partner with local universities for research projects, giving students hands-on experience with real infrastructure.
3. Government procurement. When government agencies adopt AI, they should be required to upskill their teams, not just outsource to consultants.
4. Startup support. New Zealand AI startups need more than funding — they need access to expertise and infrastructure that doesn't exist locally.
5. Public awareness. Most New Zealanders have no idea what agentic AI is or why it matters. A public information campaign could help.
The Opportunity
Despite the warning, there's reason for optimism. New Zealand has advantages:
- Renewable energy: Our hydro and wind power makes us attractive for energy-intensive AI training
- Stable government: Investors like predictable regulatory environments
- English-speaking workforce: We can adopt global AI tools without translation
- Small, adaptable economy: We can move faster than larger nations when we choose to
The Southland facility itself is an opportunity. If New Zealand can train a generation of AI engineers who understand both how to build models and how to use them, the infrastructure investment multiplies in value.
The Risk
But without that training, the risk is real: New Zealand becomes an AI colony — hosting infrastructure owned by overseas companies, exporting data, importing finished AI products, and missing the value creation entirely.
Yesterday's supply chain attack was a reminder that security literacy matters too. As we build AI capability, we need people who understand the risks as well as the opportunities.
What This Means For You
If you're a New Zealand developer, researcher, or business leader:
- Learn agentic AI. Move beyond chatbots. Understand how agents can work autonomously.
- Experiment locally. You don't need Southland's facility to start — smaller models run on consumer hardware.
- Connect globally. Join communities where people are building with the latest tools. What you learn will be years ahead of local courses.
- Build security awareness. AI development has supply chain risks. Learn to protect yourself.
- Advocate. Push for education, training, and infrastructure access.
The infrastructure is coming. The question is whether New Zealand will have the people ready to use it.
Sources
- Datagrid/Transpower announcements on Southland data centre
- New Zealand AI industry commentary
- NVIDIA Jensen Huang on AI agents (7.5 million figure)
- Andrej Karpathy on self-researching AI
- US National Academy of Sciences AI teacher training ($23M, 400K teachers)
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.