Breaking
Tech NewsDeveloping Story

Memory shortage warning grows for 2027

A KB Securities report says Samsung and SK Hynix inventories have fallen below 10 days as AI demand reshapes the memory market.

··2 hours ago·6 min read
SODIMM RAM stick
Photo by Franck V. on Unsplash

AI's appetite for memory is straining the companies that make it, and a new analyst report warns the squeeze may not ease anytime soon. According to a report from KB Securities, the memory chip market is heading toward a severe supply shortage in 2027.

That projection, reported by Business Korea and flagged by Wccftech, points to shrinking inventories at two of the world's largest memory producers at the same time that AI infrastructure spending continues to climb. Because the Business Korea article was translated from Korean, wording may differ slightly from the original, but the substance is consistent: supply is tight, and demand is not slowing.

Inventories fall below ten days

The KB Securities report suggests Samsung and SK Hynix are now down to fewer than 10 days of memory inventory. That figure is the kind of buffer that gives manufacturers little room to absorb demand swings or production hiccups.

Inventory levels function as a rough gauge of how much slack exists in the supply chain. When they get that low, any unexpected spike in orders — or any disruption at a fabrication plant — can translate quickly into shortages and price movement.

The report frames the situation as a supply crunch that could persist into 2027, with analysts quoted in the coverage saying the market is heading toward a severe shortage.

AI infrastructure spending climbs

The demand side of the equation is being driven by AI. KB Securities notes that memory chip manufacturers are struggling to keep up with demand for HBM4 memory used in AI data centers.

Infrastructure investments for AI development have been revised upward to $1.3 trillion for 2027, according to the report. That represents a 60% year-on-year increase.

HBM, or high-bandwidth memory, sits close to AI accelerators and is a critical component in training and running large models. The report's central claim is that this demand is not isolated to HBM alone.

"AI servers will absorb demand for not only HBM but also server DDR5 and enterprise SSDs, potentially leading to a historic shortage."

— Kim Dong-won, head of research at KB Securities, as told to Business Korea.

What the report actually claims

It is worth being precise about what the KB Securities report does and does not say. The claims come from a single research note, relayed through a translated Korean-language article and then picked up by Wccftech.

The report does not offer a detailed supply-demand model in the coverage available, and it does not quantify how much prices might rise or which specific products would be affected first. What it does state is that inventories at Samsung and SK Hynix have fallen below 10 days and that AI server demand is pulling on multiple memory categories at once.

Because this is a translated report citing an analyst firm, the exact phrasing of the original Korean text may differ from the English coverage. The core numbers — the sub-10-day inventory figure, the $1.3 trillion infrastructure investment, and the 60% year-on-year increase — are as reported in that coverage.

Framework cuts prices on some orders

Against that backdrop, laptop maker Framework has taken a step in the opposite direction. As VideoCardz reported, the company has secured fresh memory modules at a lower price and is reducing the price of some of its products as a result.

Framework had previously raised laptop prices because of the increased cost of RAM, a move other notebook manufacturers have also made. In a post on X, the company said it is "retroactively reducing prices for orders that shipped at higher price" as well.

That means some current Framework Laptop orders will be adjusted downward in price. The company framed the move as a response to sourcing memory at a better rate than it had been paying.

The consumer spillover question

The KB Securities report focuses on server DDR5 and HBM memory, categories used in data centers rather than in consumer devices. Whether and how quickly that pressure reaches consumer products is not something the report quantifies.

What the coverage does note is that less profitable memory types could be affected as manufacturers direct capacity toward higher-value products. That is a general observation about how memory production works, not a specific forecast tied to a number in the report.

GPU price hikes have also been discussed again, according to the source coverage. That adds to a broader sense of unease around component pricing, though the report does not connect the two directly.

Reading the numbers carefully

There are a few figures worth keeping straight as this story develops. Each comes from the KB Securities report as relayed through Business Korea and Wccftech:

  • Fewer than 10 days of memory inventory at Samsung and SK Hynix, per the report.
  • $1.3 trillion in revised AI infrastructure investments for 2027.
  • 60% year-on-year increase in those infrastructure investments.
  • The shortage projection is tied to 2027.

No dollar figures for memory prices, no percentage forecasts for price increases, and no unit shipment estimates appear in the source coverage. Anyone citing specific price targets should be clear about where those numbers come from.

Why the timing matters

The report arrives during a period when memory supply has already been a subject of concern. The mention of a "historic shortage" is a notable escalation in tone from an analyst firm, even if the underlying claims are still early and single-sourced.

Inventory below 10 days is the kind of data point that tends to draw attention because it is easy to compare against normal operating levels. Whether that figure reflects a temporary dip or a sustained trend will depend on production decisions and demand through the coming quarters.

The AI infrastructure number — $1.3 trillion for 2027 — is the demand-side anchor for the report's argument. If that spending materializes as projected, the memory requirements it implies would be substantial.

What to watch next

Several things could clarify or complicate this picture. Memory makers' own earnings calls and production guidance would provide a more direct read on inventory and capacity plans than a single analyst note.

Framework's price adjustment is a small counterexample, but it reflects a company that happened to source modules at a better rate, not a signal that the broader market is loosening. The company has been transparent about its memory costs driving prior price increases.

For now, the KB Securities report stands as a projection, not a confirmed outcome. The numbers it cites are specific and worth tracking; the interpretation of what they mean for 2027 is an analyst's call.

Why it matters

If the KB Securities projection is even directionally right, the consequences could reach well beyond data center operators. Memory is a component in nearly every computing device, and sustained tightness in the supply chain tends to show up in component pricing and availability over time. That could mean higher costs for businesses refreshing server infrastructure and for consumers buying laptops, prebuilt desktops, or upgrades.

It could also affect the pace at which AI features are deployed if memory becomes a bottleneck rather than an abundant resource. None of that is guaranteed by a single report, and the supply chain has surprised analysts in both directions before. But the combination of sub-10-day inventories and a projected 60% jump in AI infrastructure spending is the kind of setup that suggests the memory market may stay tight for a while. Readers planning hardware purchases or infrastructure budgets may want to treat this as a data point worth monitoring rather than a settled forecast.

#ram#memory chips#ai infrastructure#samsung#sk hynix#supply chain

Sources

Iliyas

Founder & Editor, Xploitwire

This article was written and reviewed against the sources listed above before publication, under editorial policies set by Iliyas. Read our Editorial Policy →

← Back to all stories