Gartner expects worldwide semiconductor revenue to almost double to roughly $1.56 trillion this year as artificial-intelligence infrastructure combines with an exceptional memory cycle to concentrate a growing share of chip-industry economics around data centres.

The research company forecasts revenue growth of 92% from about $809 billion in 2025, followed by a further increase during 2027. Memory is doing much of the immediate work, with Gartner expecting the category to generate more than $837 billion during 2026 and account for more than half of total semiconductor revenue.

Within that market, DRAM and NAND revenue are expected to rise particularly sharply as AI systems consume large quantities of high-performance and conventional memory. Additional production capacity is due to arrive, although Gartner still expects supply and demand to remain tight as infrastructure providers continue expanding compute fleets.

The longer-term change extends beyond an unusually favourable pricing cycle. Gartner estimates that the AI data-centre ecosystem will account for more than a third of semiconductor revenue this year and over half by 2030, pulling processors, memory, networking equipment, and supporting components towards the same infrastructure investment cycle.

AI spending spreads through the hardware stack

The forecast puts scale around an investment boom often described through the capital expenditure of individual cloud providers or the sales of accelerator manufacturers. Training and running large AI systems requires much more than GPUs, because every new compute cluster also consumes high-bandwidth memory, conventional memory, networking silicon, storage, power electronics, and a range of supporting hardware.

That widening demand explains why the financial effects of AI are reaching companies several stages removed from the best-known model developers. Semiconductor-material suppliers, networking vendors, infrastructure operators, specialist chip designers, and data-centre developers are exposed to the same build-out, although their margins and investment risks differ substantially.

Memory is particularly important because the headline revenue growth combines underlying demand with sharply higher prices. A 92% rise in semiconductor revenue does not mean that physical chip volumes have doubled, and that distinction leaves the current expansion sensitive to future capacity additions and changes in purchasing behaviour.

Organisations planning AI infrastructure over several years therefore face costs beyond the accelerator itself. Higher component prices flow into server economics and cloud capacity, while buyers that secure GPUs without accounting for memory, networking, electricity, cooling, and financing can underestimate what it takes to place those accelerators into productive operation.

Europe faces the economics of dependence

European ambitions around AI adoption and digital sovereignty sit alongside heavy dependence on semiconductor supply chains and cloud infrastructure extending well beyond the region. Access to advanced accelerators is only part of that exposure; organisations also have to finance and source the memory, networking, power systems, and facilities required to turn chips into usable compute.

Those pressures are already changing how infrastructure is financed. Nvidia’s willingness to use its own balance sheet to support major projects, examined in Techopia’s recent coverage of AI compute financing, reflects a market in which demand can move beyond customers’ conventional ability to fund increasingly large deployments.

Gartner’s forecast implies that the semiconductor industry itself will become more dependent on those projects. If AI data centres account for more than half of chip-industry revenue by 2030, changes in model economics, construction activity, power availability, or enterprise AI demand will carry greater consequences for semiconductor suppliers than they did when demand was distributed more evenly across consumer electronics, conventional servers, industrial equipment, and other markets.

The immediate revenue figures are striking, although the structural shift is more durable than a single year’s growth rate. AI is turning semiconductor capacity into a strategic infrastructure input, while memory pricing shows that the cost of expansion reaches far beyond accelerators. Europe’s push for additional compute therefore enters a market in which access to components, electricity, financing, and physical infrastructure is becoming part of the same industrial problem.