Omdia projects 62.7% semiconductor growth

Semiconductors are having one of those moments where a forecast stops looking like a forecast and starts looking like a stress test. Omdia just raised its 2026 semiconductor revenue growth view to 62.7%, and the point is not simply “AI demand is strong.” The point is that AI is now pulling on every part of the stack at once — memory, CPUs, packaging, and wafers — hard enough to create shortages in places that used to look boring. That changes how to read the whole market. (omdia.tech.informa.com) ### What did Omdia actually change? On April 23, Omdia said 2026 semiconductor revenue should grow 62.7%, a big upward revision tied to sustained AI and data-center demand plus ongoing supply shortages. Computing and data storage are expected to lead the market, with that segment rising 90% year over year to more than $700 billion. This is not a broad consumer rebound story. It is a concentrated infrastructure boom. (omdia.tech.informa.com) ### Why is memory doing so much of the work? Because AI servers eat memory in a very specific way. Omdia says DRAM could nearly double in value in 2026, while NAND — smaller today, but tighter — could quadruple versus 2025. The catch is that manufacturers are st(omdia.tech.informa.com)y and equipment away from more ordinary memory chips. That is how an AI boom turns into a broader memory crunch. (omdia.tech.informa.com) ### Why does HBM squeeze everything else? HBM is not just “more memory.” It is stacked, packaged, and integrated in a much more demanding way. So when suppliers chase HBM, they do not just reallocate bits — they reallocate process steps, packaging capacity, and (omdia.tech.informa.com)ts and the conventional memory left behind. Meaningful supply relief, in Omdia’s view, may not show up until well into 2027. (omdia.tech.informa.com) ### Is this only about GPUs? No — and that is the more interesting part. AMD has been arguing that agentic AI and inference create more work for CPUs because CPUs handle scheduling, data prep, memory movement, I/O, and control flow around the accelerators. Lisa Su put it more bluntly in February: CPU demand was “going gangbusters.” So the AI buildout is widening from the flashy chip to the supporting compute around it. (amd.com) ### What about Intel? Intel is saying something similar from the other side. In its April 23 earnings call, the company said its outlook for server CPU demand improved over the prior 90 days and that it expects strong double-digit unit growth for both the industry and Intel, with momentum extending into 2027. That line(amd.com) were assuming a year ago. (fool.com) ### So where does the bottleneck move next? Into packaging, memory, and upstream fab inputs. If memory values are exploding and CPU demand is reaccelerating, foundries and OSATs do not get a clean breather just because GPU headlines dominate. Advanced packaging becomes the tollbooth. AI-grade wafers and memory com(fool.com)n if they are not the headline AI winner. That is the market implication hiding inside Omdia’s number. (omdia.tech.informa.com) ### Is 62.7% believable? It is aggressive, but it is internally coherent. Omdia is not calling for every chip market to boom at once. It is saying one part of the market — AI infrastructure — has become so large, and so memory-intensive, that it can drag the who(omdia.tech.informa.com)at is why this forecast feels extreme but not random. (omdia.tech.informa.com) ### Bottom line? The semiconductor story in 2026 is no longer “who has the best AI accelerator.” It is “who can supply the whole machine around it” — especially memory, CPUs, and packaging. Omdia’s 62.7% call is really a warning that the AI boom has spread beyond one chip category and into the physical limits of the supply chain. (omdia.tech.informa.com)