AI infrastructure spending set to cross USD 769 billion in 2026, McKinsey finds

Global spending on artificial intelligence infrastructure and model architectures is projected to reach roughly USD 769 billion in 2026, a more than fivefold jump from the USD 145 billion recorded in 2025, according to McKinsey's Technology Trends Outlook 2026.

The forecast rests on nearly USD 384 billion already invested in the first half of 2026, with the full-year figure assuming that pace holds. That makes AI infrastructure and model architectures the most heavily funded technology trend tracked in the report.

The capital surge is being driven by a fresh round of commitments from major AI developers and their financiers. OpenAI is in talks with investors over funding that could value it at about USD 1.2 trillion. SoftBank has launched an USD 11 billion bond offering to help finance a further USD 10 billion investment in OpenAI. Anthropic, meanwhile, is discussing a potential investment of up to USD 10 billion from Nvidia as part of a proposed initial public offering that could raise as much as USD 100 billion.

Yet the investment boom is unfolding against mounting unease within the industry about how quickly AI capabilities are advancing and whether existing safeguards can keep pace. Anthropic CEO Dario Amodei has urged the global AI community to slow the release of new capabilities to allow more time to address safety risks. OpenAI has called for mandatory national safety requirements, including independent assessments, cybersecurity measures and incident reporting for advanced AI systems, and has separately pressed the United States to lead international efforts on technical standards for frontier AI and systems capable of recursive self-improvement.

McKinsey identified five technology trends — agentic software development, AI infrastructure and model architectures, AI for scientific discovery and engineering, the future of space technologies, and the future of robotics — that are on course to draw more than double the investment in 2026 compared with 2025. Every trend except advanced connectivity is expected to see higher year-on-year funding.

The shift reflects the growing weight of the physical infrastructure needed to scale AI, including semiconductors, data centres and power systems. Energy technologies alone attracted nearly USD 200 billion in 2025, while AI infrastructure spending doubled during the year.

McKinsey cautioned that the expansion could run into constraints from shortages of energy, talent and capital, as well as legacy technology systems and cybersecurity risks. Data centres running AI workloads in the United States are projected to consume as much electricity by 2030 as California does today, while more than 2,500 gigawatts of energy projects worldwide are waiting for grid connections.

At the enterprise level, adoption is broadening but returns remain elusive. The report found that 89 per cent of organisations regularly use AI in at least one business function, yet only 37 per cent report any positive EBIT impact at the enterprise level. The next phase of adoption will require companies to move beyond experimentation and redesign workflows around humans and AI agents.

"The defining question of the agentic era is not how autonomous agents can become but how much autonomy the enterprise can safely absorb," said McKinsey partner Oana Cheta, pointing to the operational and governance challenges that could shape the coming phase of AI adoption.

Together, the figures point to a widening gap between the speed at which capital is being deployed into AI and the growing emphasis among leading developers on safety, oversight and governance as ever more capable systems are built.