AI Infrastructure Demands Resource-Style Taxation and Accountability

Another AI giant is going public, and an exciting day for Squiz Kids

The impending public debut of Open AI--and the trillion-dollar valuations surrounding AI firms--reveals a deeper systemic imbalance: the infrastructure required to sustain these technologies is being treated as an afterthought, not a core cost. While investors bet on AI as "the new electricity," the physical and environmental toll of the data centres powering it remains largely unpriced and unregulated, especially in countries like Australia where foreign-owned tech giants may extract value without proportional tax contributions. Senator David Pocock’s call to tax data centre developments like resource extraction underscores a critical shift in thinking--AI isn’t just software; it’s an industrial-scale operation with real-world consequences. This conversation matters most to policymakers, civic leaders, and long-term investors who understand that unchecked technological optimism without structural accountability leads not to progress, but to concentrated risk and public burden. Recognising this misalignment early offers a strategic advantage: the ability to shape rules before power becomes entrenched.

Why the “New Electricity” Needs Its Own Resource Tax

AI’s promise is framed in transformational terms--replacing human labour, accelerating discovery, reshaping entire industries. But the metaphor of AI as “the new electricity” only holds if we also accept that generating it requires something like coal, gas, or uranium: raw inputs, physical plants, and environmental impact. The transcript makes clear that Open AI, Microsoft, and Amazon are not just launching software--they’re building massive data centres in Australia, deals worth billions, consuming vast amounts of energy and land. This isn’t digital vapour; it’s industrial infrastructure. And like any extractive industry, it risks creating private profit while socialising the cost.

"We've watched foreign-owned companies extract enormous value from Australian resources before while finding ways to minimise the tax they pay here."

-- Andrew Williams, summarising Senator David Pocock

Pocock’s warning is a systems-level insight: history repeats when incentives aren’t corrected. Past resource booms saw multinational corporations reap profits while local communities bore the environmental and economic costs, often with minimal tax return. Now, the same pattern threatens to unfold with AI infrastructure. The data centre isn’t just a server room--it’s a power-hungry facility that will draw from national grids, potentially displacing capacity from households or other industries. And because these companies are often structured across jurisdictions, they can optimise for low tax exposure even as they consume public resources.

This creates a feedback loop: high valuations attract investment, which funds more infrastructure, which increases energy demand and political leverage. Over time, these firms don’t just participate in the economy--they begin to shape it. If untaxed or lightly regulated, they gain competitive advantage over local businesses that do pay full infrastructure costs. The system responds by favouring scale over sustainability, speed over equity.

The delayed payoff of acting now--taxing data centres like resource projects--is a fairer, more resilient digital economy. But most governments won’t act because the immediate pressure is to attract investment, not impose conditions. That’s where the advantage lies: jurisdictions that impose smart, proportional taxes early won’t just capture revenue--they’ll set the terms of engagement. They’ll force AI firms to internalise their real costs, leading to more sustainable innovation. Others will inherit congestion, blackouts, and public backlash.

The Hidden Cost of Believing the Hype

Not everyone agrees on AI’s valuation. The transcript notes that “some experts have questioned whether Open AI makes enough revenue from its products to warrant the big numbers.” That doubt is the first crack in the narrative. Behind the trillion-dollar claims lies a business model still searching for profitability. ChatGPT is popular, but popularity doesn’t pay for data centres. The immediate benefit of going public is access to capital. The downstream effect? Pressure to monetise at scale, fast.

This creates a dangerous incentive cascade. To justify valuation, AI companies must grow revenue, which means embedding their tools deeper into workplaces, education, and government--places where errors or biases can do real harm. The system adapts: organisations adopt AI to stay competitive, not because it’s ready. Then, when failures occur--misinformation, job displacement, security breaches--the backlash falls not on the vendor, but on the adopter. The AI firm, insulated by distance and complexity, survives. The school, hospital, or agency takes the hit.

And the environmental cost compounds. Every query to a large language model burns energy. Multiply that by millions of users, billions of requests, and the carbon footprint becomes undeniable. Yet this cost is invisible at the point of use. Like plastic straws or fast fashion, the convenience masks the damage. The real kicker? Much of this infrastructure is being built in regions like Australia, which are already vulnerable to climate change. The very places asked to bear the physical cost of AI may suffer most from its indirect consequences: drought, fire, and extreme weather.

"AI is a very complex and wide-ranging technology... those who believe in its value think it's going to be a transformative economic game changer."

-- Andrew Williams

The belief is real. But belief doesn’t generate electricity. And when belief outpaces reality, bubbles form. The 18-month payoff for sceptics isn’t short-selling stock--it’s credibility. When the gap between promise and performance widens, the organisations that held back, that demanded proof, that built slowly and sustainably, will find themselves in a stronger position. Everyone else will be scrambling to unwind bad integrations, fix biased systems, or explain why their AI project failed.

Where Immediate Pain Creates Lasting Moats

There’s a quiet contrast in the episode: the collapse of Barbeques Galore versus the expansion of News Hounds. One is a retail chain that couldn’t adapt, closing 62 stores and making 500 staff redundant. The other is a media literacy program being rolled out across Tasmania as a “core life skill.” The difference? Time horizon.

Barbeques Galore likely chased quarterly sales, promotions, and foot traffic--the visible metrics. News Hounds, backed by government funding, is playing a longer game: teaching children to think critically about information. That doesn’t generate revenue. It doesn’t go viral. But it builds resilience. And in a world where AI can generate convincing fake news in seconds, that resilience becomes a moat.

The system responds to misinformation by demanding literacy. But most organisations won’t invest in prevention--they’ll wait for a crisis, then spend more on damage control. That’s the path of higher cost and lower impact. The ones who act now, who accept the discomfort of teaching slow thinking in a fast world, will be the ones who maintain trust when the next wave of synthetic media hits.

This isn’t just about schools. It’s about culture. If we train people to question sources, verify claims, and understand how algorithms shape their views, we create a society that’s harder to manipulate--not just by foreign actors, but by the very AI systems we’re building. That’s a durable advantage. And it starts with uncomfortable choices: funding education over entertainment, depth over speed, long-term safety over short-term gains.

Key Action Items

• Advocate for data centre taxation frameworks -- Over the next quarter, engage with local policymakers to model resource-style levies on AI infrastructure. This pays off in 12--18 months when new projects seek approval.
• Audit AI dependencies in your organisation -- Within 60 days, map where AI tools are used without costed energy or risk assessments. Flag high-exposure areas for review.
• Invest in media literacy training -- Start now with teams that handle public communication or content. The payoff emerges over 1--2 years as judgment improves and misinformation incidents drop.
• Delay AI adoption in high-risk domains -- Where errors could harm people (e.g., hiring, healthcare triage), impose a six-month moratorium. Use that time to build oversight protocols. Discomfort now prevents crisis later.
• Support local digital sovereignty initiatives -- Over the next year, partner with regional tech or education programs that keep data and decision-making local. This builds resilience against foreign-controlled platforms.
• Track the gap between AI valuation and revenue -- Monitor Open AI, Anthropic, and similar firms for signs of overreach. When hype exceeds earnings, prepare for market correction--opportunities will emerge.
• Teach critical thinking as a core skill -- Whether in schools, workplaces, or families, prioritise questioning over consuming. This doesn’t scale fast, but it lasts.