RAM Shortage to Last Through 2028, Micron CEO Says

Micron CEO Sanjay Mehrotra told investors on September 30 that demand for his company's memory products will outstrip available supply for at least the next two years — a forecast that extends the current crunch into 2028 and reshapes expectations for everyone from data center operators to laptop buyers. Samsung executives echoed the same view this week. Together, the two companies control a substantial share of global DRAM and high-bandwidth memory (HBM) output, which makes their public guidance unusually consequential for the wider electronics supply chain.

RAM Shortage Expected to Last Through 2028, Executives Warn

Mehrotra's remarks came during a quarterly investor call, where he said Micron expects demand for its memory to exceed what the company can produce over the next couple of years. That timeline places the shortage's resolution no earlier than 2028, and it aligns with separate commentary from Samsung executives this week. The two firms rarely speak with one voice on supply outlooks, so the convergence is notable.

The current cycle already ranks among the more severe memory contractions in recent history. During the 2017–2018 DRAM upcycle, contract prices for 8GB DDR4 modules roughly doubled over seven quarters before capacity additions finally caught up, according to TrendForce tracking data. That episode offers a rough template: shortages rarely break on their own. They break when new fab capacity — which takes 18 to 36 months to bring online — finally reaches volume production.

Micron's forecast implies the industry is not close to that inflection point. The company has already exited the consumer RAM business, a strategic shift that removes it from the retail channel entirely. What remains is business-to-business: HBM for AI accelerators and DRAM for servers. Those are the segments where Mehrotra says supply will remain tight through the end of the decade's first act.

Anyone searching for clarity on the RAM shortage 2028 timeline will find the most authoritative signal in these earnings calls, not in retail price trackers. Executive guidance reflects booked capacity, customer commitments, and wafer start schedules — data that retail buyers see only after it has already been priced in.

Why AI Is Driving the Memory Supply Crunch

A single AI accelerator module can carry six or eight stacks of HBM, and each stack contains multiple DRAM dies stacked vertically and connected through silicon vias. Producing one HBM stack consumes roughly three times the wafer area of an equivalent density of conventional DRAM, according to semiconductor manufacturing analyses. That multiplier is the core mechanic behind the shortage.

When NVIDIA, AMD, Google, Amazon, and Microsoft place orders for AI training and inference hardware, they are effectively bidding for DRAM wafer capacity that would otherwise flow into servers, PCs, and phones. HBM commands far higher margins per bit, so foundries and memory makers allocate accordingly. The result is a structural reallocation of capacity, not a temporary spike in demand.

Mehrotra's visibility into this pipeline is direct. Micron sells HBM to the accelerator vendors building the current generation of AI infrastructure, and it sells server DRAM to the hyperscalers operating them. When he says supply will lag demand for two more years, he is describing orders already in the queue.

The pattern resembles the automotive chip shortage of 2021, when automakers lost production because foundry capacity had been committed to consumer electronics. In that case, the misallocation took roughly 18 months to unwind. Memory is harder to redirect because HBM and conventional DRAM use different packaging and testing lines, making supply switching slow and expensive.

How the B2B Memory Crisis Spills Into Consumer Devices

Micron no longer sells consumer RAM, but its decisions still reach the retail shelf. Every wafer allocated to HBM or server DRAM is a wafer unavailable for the DDR5 and LPDDR5 modules that go into laptops, desktops, and smartphones. That reallocation is why consumer memory pricing tracks enterprise contract pricing with a lag of one to two quarters.

The consumer impact shows up in three places. First, DIY builders and system integrators face higher module prices and thinner inventory. Second, OEMs like Dell, HP, Lenovo, Apple, and Samsung must decide whether to absorb memory cost increases or pass them through — and in a competitive PC market, absorption usually means cutting configurations elsewhere. Third, smartphone makers, who consume LPDDR on very thin margins, are especially exposed.

A 2024 precedent is instructive. When HBM demand first surged, conventional DDR5 contract prices rose sharply even though consumer demand was flat. The mechanism was capacity displacement, not consumer appetite. The same dynamic is now extended through 2028 under the executives' guidance.

For buyers, the practical translation is simple: memory-dependent devices are unlikely to get cheaper on a per-gigabyte basis for the foreseeable future. Sales may still appear, but they will be margin-funded promotions, not evidence of easing supply.

What This Means for PC and Smartphone Buyers

If you are planning a PC build, a laptop purchase, or a phone upgrade between now and 2028, the calculus changes. Waiting for prices to fall has historically worked after memory upcycles peak. Mehrotra's guidance says that peak has not arrived.

Buyers should expect three things. Configurations at a given price point may carry less RAM than they did a year earlier, as OEMs protect margins on higher-spec models. Upgrade paths — the ability to add memory later — become more valuable, which favors desktop platforms and modular laptops over soldered-down designs. And promotional pricing will be concentrated on models with memory configurations the manufacturer has already committed to in high volume.

There is no evidence in the executives' statements that consumer devices will face outright unavailability. The issue is cost and configuration, not absence. That distinction matters for anyone deciding whether to buy now or wait. Waiting has an opportunity cost, and in a shortage that runs through 2028, that cost compounds.

For businesses planning refresh cycles, the implication is to lock in memory-heavy configurations earlier rather than later. Enterprise procurement teams that delayed DDR5 transitions during the 2022–2023 downturn now face repricing at the worst possible moment in the cycle.

Industry Context: Manufacturing Capacity and Market Dynamics

Memory manufacturing is a capital-intensive, slow-cycling business. A new DRAM fab costs well over $15 billion and takes two to three years from groundbreaking to volume output, according to industry cost estimates. That lag is why supply cannot respond quickly to price signals, no matter how strong the demand.

Capacity utilization across the leading memory makers has been running at or near full load, according to semiconductor industry reporting. When utilization is maxed and new capacity is years away, the only variables left are allocation and pricing. Both currently favor AI and server customers.

Market concentration amplifies the effect. Three firms — Samsung, SK Hynix, and Micron — produce the overwhelming majority of global DRAM. When two of them publicly signal multiyear tightness, the market prices that signal immediately. Historical memory cycles, including the 2017–2018 upcycle and the 2021 automotive shortage, show that once executives guide for extended tightness, prices tend to stay elevated until new capacity physically arrives.

The 2017–2018 episode is the closest analogue. DRAM contract prices rose for seven consecutive quarters. They fell only after Samsung, SK Hynix, and Micron brought new wafer capacity online in 2019. If the current cycle follows that pattern, and Mehrotra's two-year guidance holds, the next meaningful relief would arrive around 2028 or later.

Key Takeaways and Outlook for Memory Supply Through 2028

The headline is straightforward. Micron's CEO expects demand to exceed supply for at least two more years, and Samsung executives agree. The cause is HBM and server DRAM demand from AI infrastructure, which is displacing capacity that would otherwise serve consumer devices.

For readers tracking the RAM shortage 2028 outlook, the actionable signals are these:

• Executive guidance beats retail pricing as a leading indicator. Mehrotra sees booked capacity; retail trackers see the aftermath.
• HBM consumes disproportionate wafer area. Every AI accelerator order tightens conventional DRAM supply.
• New fabs take years. Capacity cannot respond within the shortage window, which is why the timeline extends to 2028.
• Consumers face configuration pressure, not disappearance. Expect less RAM at a given price point, not empty shelves.
• Enterprise buyers should commit early. Refresh cycles that wait for lower prices may wait past the shortage itself.

The memory market has broken shortages before, and it will break this one. But the mechanism is physical capacity, and the executives with the clearest view of that capacity say it will not arrive in time to change the picture before 2028. Plan accordingly.

Source: Ars Technica - All content