OpenAI, a leading artificial intelligence firm, has announced a staggering increase in its planned infrastructure investment, now projecting to spend $750 billion by the year 2030. This figure represents a 25% hike from earlier estimates made this year, according to reports. The colossal sum, which is comparable to the current Gross Domestic Product of Sweden, underscores the immense capital and resources required to develop and operate cutting-edge AI technologies, potentially reshaping global technology infrastructure and energy markets for years to come.
The initial phase of this unprecedented spending spree is a $20 billion data centre campus in Georgia, USA, dubbed 'Project Camellia'. This expansive development will cover 1,400 acres northwest of Savannah and is set to draw a minimum of 3.2 gigawatts of power from Georgia Power, the regional utility. While electricity is expected to begin flowing to the site from 2028, the company has not yet provided a timeline for when the first Graphics Processing Units (GPUs) will become operational within the facility.
OpenAI has committed to covering the full cost of the infrastructure and electric-service expenses for the new data centre. This commitment aligns with a rule adopted by the Georgia Public Service Commission last year, which prevents utilities from passing on costs associated with new users drawing more than 100 megawatts. Furthermore, Georgia Power has indicated that OpenAI will have the capability to reduce its power consumption by up to 1 gigawatt during periods of high demand on the grid, aiming to mitigate strain on the local power supply.
The energy source for Project Camellia raises environmental questions. Regulatory filings suggest that the bulk of the new capacity secured by Georgia Power to meet this demand will come from natural gas. The utility received approval in December to generate an additional 9,885 megawatts, expecting to contract all this capacity by the end of 2026, with the OpenAI deal accounting for approximately a third. Plans indicate that about 5.8 gigawatts of new natural gas generating capacity will be built or bought, with a quarter of this from more polluting simple-cycle turbines, effectively doubling Georgia Power’s natural gas fleet. The remaining power will be supplied by grid-scale batteries and solar.
The sheer scale of OpenAI's investment highlights the intense competition and escalating costs within the AI sector. Such vast expenditures on infrastructure, particularly data centres, reflect the insatiable demand for computational power needed to train and run increasingly complex AI models. This trend could lead to further consolidation in the AI industry, favouring companies with deep pockets, and potentially driving up the cost of components and energy globally.