Table of Contents
Memory shortage to continue due to massive demand, according to Synopsys CEO
In sum – what we know:
- A lengthening timelime – The global memory chip shortage is now expected to persist through at least 2027, significantly longer than previous forecasts.
- Root cause – Aggressive spending on AI infrastructure and High-Bandwidth Memory (HBM) is consuming virtually all available supply from major manufacturers.
- Impact – Consumer electronics markets (PCs, smartphones) face “starved” inventory and rising prices as costs are passed to end-users.
The memory chip shortage isn’t going anywhere soon. Synopsys CEO Sassine Ghazi now expects supply constraints to persist into 2027 at least — a timeline that extends well past what the industry had been forecasting. The reason, of course, isn’t all that surprising — AI infrastructure buildouts are using up virtually everything Samsung, SK Hynix, and Micron can produce.
Ghazi calls this a “super cycle,” distinguishing it from the boom-bust patterns the semiconductor industry knows well. Those traditional shortages tend to resolve in months as supply catches up. This one looks different though, and with no clear end point even as new fabs come online.
What’s driving all this? AI data centers are absorbing the lion’s share of memory output, especially the high-bandwidth memory that large language models require. Hyperscalers are spending tens of billions on infrastructure, and production simply can’t keep pace. With AI claiming most of the supply, everything else, including smartphones, PCs, laptops, and consumer electronics broadly, get what’s left over.
New capacity takes a minimum of two years to bring online, so near-term relief was never realistic. Samsung and SK Hynix are ramping up around 175,000 additional wafer starts per month, roughly an 8% increase, and they’ve pulled those timelines forward by about three months. Even so, meaningful impact won’t materialize until early 2027 at the earliest, and that assumes AI demand doesn’t simply absorb the new supply the moment it arrives — a bet few would take given current trajectories for HBM and agentic AI workloads.
Consumer electronics are already feeling the squeeze. PC volumes are projected to drop 9% this year, with smartphones expected to fall at least 5%. High-end products will get more expensive, while low-end products may simply become unavailable rather than just pricier.
Rising prices
Prices are climbing throughout the supply chain, and those increases are making their way to consumers, though thankfully not all electronics companies will pass on those costs. Lenovo has said that despite the increases, it aims to avoid passing on costs to consumers.
For memory manufacturers themselves, the shortage has been extraordinarily lucrative. Micron’s stock has surged in the past year. Samsung and SK Hynix have posted strong gains as both raise prices sharply. Ghazi called it “a golden time for the memory companies,” with margins expanding as demand overwhelms supply. Micron’s HBM inventory is reportedly sold out through 2026.
The competitive dynamics among suppliers are sharpening too. Samsung is accelerating HBM4 production, targeting Nvidia as a key customer and trying to narrow the gap with SK Hynix. Nvidia’s forthcoming Vera Rubin platform, which pairs with HBM4, is intensifying the race among suppliers angling for position in the next wave of AI infrastructure.
Conditions that benefit memory makers create real headwinds for consumer electronics companies squeezed on margins and inventory. Whether this imbalance eventually prompts coordination between tech companies and chip manufacturers, or even government intervention to ensure more balanced allocation, remains uncertain.
