Microsoft is preparing for one of the largest expansions of artificial-intelligence computing infrastructure in the technology industry, planning a data-center buildout that could add roughly 26 gigawatts of computing capacity as demand for AI services accelerates.
The scale of the expansion illustrates how quickly the economics of artificial intelligence are moving beyond chips and software. Microsoft now needs enormous quantities of electricity, land, cooling equipment, networking infrastructure and data-center hardware to support the AI models and applications running across Azure.
The company has already been expanding its infrastructure at an extraordinary pace. Microsoft said earlier this year that it was on track to roughly double its overall data-center footprint within two years. In its latest fiscal-year results, the company said it added another gigawatt of capacity in the fourth quarter and continued expanding across multiple continents.
Microsoft’s investment is being driven partly by the growing demand for AI workloads from customers and by its partnership with OpenAI. The company has also been developing its own AI models and Copilot products, meaning that its infrastructure requirements are no longer tied to a single AI provider.
The sheer amount of planned computing capacity highlights the changing nature of the data-center industry. Traditional cloud computing required large facilities, but AI workloads can demand dramatically higher concentrations of electricity and computing power. Training frontier models and running millions of inference requests require specialized accelerators, high-speed networking and sophisticated cooling systems.
Microsoft has been developing its own AI infrastructure alongside chips from Nvidia and AMD. Its Maia accelerators and Cobalt processors are increasingly being deployed throughout Azure, while the company continues to adopt the latest generations of Nvidia and AMD hardware. Microsoft said its Maia 200 accelerator was already supporting both OpenAI and Microsoft’s own AI models.
The company has also focused heavily on improving how efficiently existing infrastructure is used. Microsoft said it had increased throughput for Copilot workloads fourfold since the beginning of the year, demonstrating why raw gigawatts are not the only measure of AI capacity. Improvements in software, networking, chip design and cooling can allow more computing work to be performed without proportionally increasing electricity consumption.
Nevertheless, the 26-gigawatt expansion illustrates the enormous physical requirements behind the AI boom. Microsoft currently operates more than 500 data centers across more than 30 countries, according to its data-center business. The company says it has contracted more than 40 gigawatts of new renewable energy across 26 countries, reflecting the increasing importance of power procurement to its infrastructure strategy.
Electricity availability is becoming one of the biggest constraints on the industry’s growth. Data-center developers increasingly face long waits for grid connections, shortages of transmission capacity and opposition from communities concerned about electricity prices, water consumption and environmental impacts.
Those constraints are pushing technology companies to look beyond the United States. Microsoft and other hyperscalers are expanding in markets where electricity can be secured more quickly or where renewable energy is plentiful. Recent industry developments show major technology companies increasing investment in Europe and the Middle East as they search for additional capacity.
The financial burden is also increasing. Microsoft spent $41 billion on capital expenditures in its fiscal fourth quarter, with roughly two-thirds directed toward short-lived assets such as CPUs and GPUs and the remainder toward longer-lived infrastructure. The company said finance leases totaled $5.6 billion during the quarter, primarily for large data-center sites.
That spending is occurring across the technology industry. AI-related debt issuance in the US had approached $500 billion by early August, according to Reuters, as companies and infrastructure providers seek financing for a construction boom that could ultimately require trillions of dollars. Investors are increasingly examining whether the projected demand for AI computing will justify the enormous infrastructure investment.
Microsoft’s strategy therefore represents both an enormous bet on AI and a test of the industry’s ability to build the physical infrastructure required to support it. A 26-gigawatt expansion is not simply a technology investment; it represents a massive commitment to electricity generation, transmission, construction and long-term capital spending.
The company is betting that demand for AI will continue rising rapidly enough to absorb that capacity. If AI adoption expands as Microsoft expects, the additional infrastructure could become a crucial competitive advantage for Azure and its broader AI business. But if demand growth slows, the industry could face excess capacity and expensive assets built around forecasts that prove too optimistic.
For now, Microsoft is choosing to build ahead of demand rather than risk being constrained by a shortage of computing power. Its expanding data-center network shows that the next phase of the AI race will be fought not only over models and chips, but over who can secure enough electricity and infrastructure to run them.
