Peter Lewis, executive director of Essential, argues that Australia should not squander the opportunity presented by the AI gold rush. Instead of merely taxing AI companies, he proposes a genuine co-investment model akin to Norway's sovereign wealth fund, which was built on North Sea oil. This would ensure that the benefits of datacentres and AI training are shared with the public.
The AI Gold Rush and Australia's Opportunity
Tech titans like Microsoft and Anthropic are courting Canberra, seeking secure locations to train their massive AI models. Having exhausted public tolerance in the US for energy-intensive datacentres, they need stable environments with access to renewable energy and political stability. Australia, with its space, wind, and solar resources, could become a hub for AI production, much like the Middle East for oil.
Lessons from Norway and Australia's Gas Industry
Norway's sovereign wealth fund, now worth over A$3 trillion, was designed as a co-investment with industry, capturing benefits through taxes, licences, and dividends. In contrast, Australia's gas industry operates with little public control, leading to calls for a different approach. Lewis warns that financial arrangements for new industries must be designed upfront, as seen with the ill-fated mining tax.
Copyright and Co-investment
AI models are built on scraped data, often without regard for copyright law. Lewis insists on holding the line on copyright, requiring reparations and ongoing licensing for creators. An AI sovereign wealth fund would demand a significant share of profits from models trained in Australia, supporting creators and displaced workers.
Global Precedents and a Green Future
South Korea is exploring an AI profits fund, and US thinktanks propose a 'token tax'. However, Lewis argues that co-investment, not taxation, embeds joint benefit. By partnering with AI companies, Australia can control the technology's direction, ensure compliance with laws, and build a renewable-powered infrastructure that exports green energy. This could make Australian-trained AI the gold standard in machine learning.
