The Organisation for Economic Co-operation and Development (OECD) says that large expenditures on AI-related infrastructure — data centres, chips and other equipment — have materially dampened the economic impact of the Middle East–related energy-price shock. According to the OECD’s September forecast, without the recent AI boom the current shock would likely have had a much larger effect.
The OECD projects global GDP growth of 2.9 percent this year following last year’s 3.4 percent expansion. While this is a slowdown, it is slightly above the 2.8 percent forecast published in June, and the organisation specifically highlights that AI-related investment helped support growth in the United States and boosted exports from Japan and South Korea through higher demand for semiconductors and other tech products.
How AI spending is translating into GDP support
The OECD notes the impact goes beyond developing new AI models. Building and operating data centres requires steel, concrete, electrical equipment and substantial power, while operation demands chips, memory, networking equipment and ongoing services. The effect of data-centre and technology-equipment construction was already visible in second-quarter GDP data in the United States, Canada and Australia.
These investments have partly offset the negative consequences of higher energy prices caused by the Middle East crisis, which reduced household disposable income and increased corporate costs. The OECD also points out that some mitigating factors — releasing strategic petroleum reserves, increased non-Gulf production and weaker Chinese oil demand — helped, but AI investments remain an important pillar of economic resilience.
Why protection may weaken by 2027
The OECD warns that the cushioning effect may diminish by 2027 not because AI spending will stop, but because the economy will need to offset an increasingly large energy-related burden. Other analysts expect AI spending to keep growing: Gartner estimates global AI-related expenditures could rise by about 36 percent between 2026 and 2027, with spending on AI infrastructure increasing roughly 30 percent year-on-year. Goldman Sachs likewise expects the largest tech firms’ investments in 2027 to be about 20 percent higher than this year’s levels.
At the same time, the OECD cites a sharp rise in power demand for US data centres — increasing from 31 gigawatts in 2025 to 66 gigawatts in 2027 — implying that by 2027 the AI wave will be bigger but also have to offset a larger energy and financing burden.
The OECD has already revised its outlook: it now expects global growth of 3.0 percent for 2027 (down from a June estimate of 3.1 percent) and raised the G20 inflation forecast for 2027 from 3.1 to 3.6 percent. Energy-price increases tend to filter through over time into transport, manufacturing, food and, ultimately, broad consumer prices.
Risks and a downside scenario
The OECD warns that continued investment will depend on whether AI-related capacity generates sufficient returns. If corporate revenues fall short of expectations, investment could retract quickly. High government bond yields, energy-supply uncertainty and El Niño effects would exacerbate the situation. In a negative scenario, these factors combined could subtract 0.7 percentage points from global growth in 2027 and add 1.1 percentage points to inflation.
Regional outlook (OECD forecast)
- United States: 2.2 percent growth this year and 2.1 percent next year, partly supported by AI investments.
- Eurozone: weaker outlook, around 1.0 percent growth in both years.
- China: growth slowing from 4.5 to 4.2 percent.
- Japan: 0.8 percent growth in 2026 and 0.7 percent in 2027.
Conclusion
The OECD concludes that AI-related investment has played a meaningful role in shoring up the global economy against an energy-price shock, helping explain why 2026 may perform slightly better than earlier forecasts. However, because energy-cost effects are increasingly embedded in the economy and AI infrastructure’s energy and financing needs are rising, the same level of protection may not be sufficient by 2027, especially under adverse revenue, financing or supply scenarios.
