Tasmania has struck a controversial power deal with an AI company that raises urgent questions about how governments allocate scarce energy resources in the race to build AI infrastructure. Firmus, a Singapore-based AI company, has secured 104 megawatts of hydroelectric power from Tasmania's state-owned generator Hydro through energy retailer Aurora, enough to launch its "AI tokens factory" in the state's north. The arrangement has drawn scrutiny from opposition lawmakers and environmental groups, who point out that other major manufacturers were recently denied similar power allocations. The timing of the deal is particularly striking. Just months earlier, in June 2025, Hydro informed the Boyer Paper Mill that it could not supply enough power for the mill to transition from coal-fired burners to electric heating. The Boyer mill, one of Tasmania's largest power consumers, uses around 100 megawatts and would have needed an additional 45 megawatts to complete the conversion. Similarly, Andrew "Twiggy" Forrest's plan to establish a 250-megawatt green hydrogen plant capable of producing 250,000 tonnes of green ammonia per year was abandoned after Hydro said there was insufficient power available. Why Is Tasmania Prioritizing AI Over Traditional Manufacturing? Government officials argue that AI represents the future of Tasmania's economy and that the state must embrace the opportunity. Premier Jeremy Rockliff defended the Firmus arrangement by emphasizing that AI is a clean industry aligned with Tasmania's renewable energy advantages. Madeleine Ogilvie, the Minister for Innovation, Science and the Digital Economy, said Tasmania wanted to "grasp the economic opportunity" offered by the AI industry. However, opposition lawmakers have raised concerns about the lack of transparency surrounding the deal. Greens MP Tabatha Badger told state parliament that "such arrangements will have far-reaching consequences for Tasmania's energy grid" and called for clarity on how many data center projects are in the pipeline and what their total energy and water needs would be. Labor shadow treasurer Dean Winter also questioned why Hydro could suddenly supply 104 megawatts to Firmus when it had just told other major customers there was no power available. Aurora Energy, the retailer brokering the deal, declined to disclose the unit cost of power to Firmus, citing "commercial in confidence" considerations. The company stated that the agreement reflects "strong commercial value for Tasmania" and positions the state as a destination for "next-generation digital infrastructure". How Does the Global AI Token Economy Work? The Firmus deal is part of a much larger global shift in how AI infrastructure is being built and monetized. Across the world, companies are converting electricity into AI computing power and selling it globally in the form of tokens, tiny data units used by AI models. China's National Data Administration reported that daily token consumption had surpassed 140 trillion as of March 2026, a more than 1,000-fold increase from the 100 billion recorded at the beginning of 2024. The economic logic is compelling: raw electricity exported from China fetches about 0.5 yuan (approximately $0.07) per kilowatt-hour, but industry estimates suggest that converting that same energy into AI processing services can increase its value by up to 22 times. This value multiplication is driving governments and companies worldwide to secure power supplies for AI data centers, often at the expense of other industrial priorities. "The surge signals that China's AI industry is expanding quickly and evolving from basic chat functions to more sophisticated systems capable of decision-making and task execution," said Liu Liehong, head of the National Data Administration. Liu Liehong, Head of the National Data Administration China's western regions provide abundant low-cost renewable energy, and the country has built a vertically integrated supply chain spanning ultra-high-voltage transmission equipment, liquid-cooled data centers, and server assembly. This combination gives China a significant competitive advantage in the global AI token market. Steps to Understanding AI Data Center Power Allocation Decisions - Assess the economic multiplier effect: AI computing services can increase the value of electricity by up to 22 times compared to raw power exports, making AI data centers economically attractive to governments seeking to maximize energy value. - Compare competing industrial demands: Evaluate how power allocation decisions affect other major manufacturers and industries, such as traditional manufacturing, green hydrogen production, and paper mills that are transitioning away from fossil fuels. - Review transparency and commercial terms: Demand clarity on power pricing, contract duration, and expansion plans for AI facilities, since "commercial in confidence" arrangements can obscure the true costs and benefits to the public. - Consider grid stability and long-term capacity: Understand whether power allocations to AI facilities are sustainable given existing demand from other sectors and whether renewable energy generation can reliably support rapid scaling of AI infrastructure. Firmus has stated that it views its AI operations as a "grid asset, not grid strain," and Oliver Curtis, one of the company's co-CEOs, said the firm is "committed to proving" that AI factories can create new baseload power capacity. The company plans to progressively ramp up operations from August 2026, reaching the maximum 104 megawatts by November 2026. The Tasmania case illustrates a broader tension emerging globally: as AI infrastructure becomes increasingly power-hungry, governments face difficult choices about how to allocate scarce renewable energy resources. The decision to prioritize Firmus over traditional manufacturers suggests that policymakers view AI as a higher-value use of electricity, but the lack of public debate and transparency around these decisions raises questions about democratic accountability and long-term economic planning. The global AI token economy is still in its early stages, but the rapid growth in token consumption and the willingness of governments to restructure energy policy around AI infrastructure suggest that this trend will only accelerate. How Tasmania and other regions manage these competing demands will likely set precedents for AI infrastructure development worldwide. " }
