Australia’s AI Data Centre Own Goal

Australia’s outdated copyright rules are blocking the very AI infrastructure that could supercharge the economy, as an estimated $900 billion in AI spending by a handful of US technology giants this year looks set to bypass the country and undermine jobs, energy investment and its ambitions to lead in the region. Right now, the world’s biggest technology platforms are racing to build the data centres and computing power needed for advanced AI, with just four US firms expected to spend the equivalent of around 40% of Australia’s yearly economic output on AI infrastructure alone. This is money on a scale that could buy several of the country’s largest mining groups and still have cash spare, and it is driven by the belief that AI will reshape work, productivity and entire industries over the next decade. Yet while Australia is poised to feel the effects of this change, it is not positioned to capture much of the value. The core issue is legal, not technical. Under current Australian copyright settings, training large-scale AI models locally is effectively off limits, even though these models underpin leading global systems and the countless applications now being built on top of them. That makes it uneconomic to construct the kind of hyperscale AI data centres seen in other markets, and the facilities being built domestically tend to be smaller and focused on serving local users of AI tools rather than hosting the training of foundational models. Industry analysts argue that this legal barrier is already costing Australia tens of billions in potential capital investment along with thousands of high-skilled construction, engineering and operational roles that would have accompanied large-scale builds. The missed upside does not stop at jobs and buildings. Australia has one of the best combinations of land, sun and wind needed to generate cheap, large-scale renewable power, which is exactly the kind of green energy that energy-hungry AI operators increasingly demand. In theory, that gives the country a natural edge as a location for AI clusters powered by solar and wind, feeding clean electricity into fleets of advanced data centres. In practice, restrictive copyright rules mean these operators look elsewhere and Australia forgoes both a major boost to its clean energy rollout and the chance to anchor itself as a stable, democratic AI hub for Asia. The broader risk is that the country drifts into the role of AI consumer rather than AI builder and ends up importing tools and expertise instead of exporting them. Regulators have already recognised the problem. The national productivity watchdog has highlighted AI and copyright as a textbook example of regulation throttling innovation and competitiveness and suggested a shift to a “fair use” approach similar to the US, UK and Singapore. Under that kind of regime, AI models could be trained in Australia while still giving content owners a way to opt out of having their material used and striking a balance between innovation and rights. The blueprint exists overseas and does not require Australia to reinvent the wheel, and it mainly requires political will to align with modern standards and unlock a new wave of digital and energy investment. However, the federal government has chosen to delay. Rather than moving quickly on the commission’s recommendations, it has pushed any decision on copyright reform for AI out by roughly three years. In public administration terms that may feel brisk, but in AI terms it is an eternity, and the period since late 2022 has already taken generative AI from emerging curiosity to mainstream tool. On current settings, the next three years of breakthroughs, capital flows and ecosystem building will probably consolidate in countries that have already updated their laws while Australia watches from the sidelines and continues to debate rules that competitors resolved years earlier.

Sources

• Australian Financial Review