Memory Buyers Secure Only 65% of Needs, VanEck Manager Reports

Memory customers can currently source only 65% of their requirements, according to VanEck Emerging Markets equity portfolio manager Ola El-Shawarby, who returned from a field visit to Korea and Taiwan with reinforced conviction in the AI memory supply chain through 2027.

The figure is the hardest number in her account. It describes an AI memory market where demand runs well ahead of supply. Customers — principally the hyperscalers building out data centers — are structurally short of product, and that shortage underpins what El-Shawarby calls strong sector fundamentals.

"It's the supply chain that basically enables the build-out of data centers and for the hyperscalers to continue to develop and advance their capabilities," she said.

What did the Korea visit confirm?

El-Shawarby met with participants across the memory supply chain in Korea. The most striking finding was comfort with the 2027 horizon. Suppliers, in her account, are very comfortable with 2027 as the period over which current demand visibility extends.

That comfort matters for a simple reason. Memory producers have historically suffered brutal cycles of oversupply and price collapse. A demand window that suppliers themselves trust out to 2027 changes how they contract, how they invest, and how investors should value them.

The two companies at the center of her account are SK Hynix and Samsung. Both Korean memory giants are, in her telling, making a deliberate commercial choice.

• They favor long-term trade agreements over maximizing short-term profits.
• They are sacrificing immediate earnings for prolonged earnings duration.
• They are buying greater visibility into future demand.

This is a strategy of trading margin today for duration tomorrow. It is not the behavior of suppliers who fear a sudden demand cliff. It is the behavior of suppliers who believe the shortage persists.

Why does the 65% figure matter?

The gap between what customers need and what they can secure is the story's core economic fact. When buyers obtain only 65% of their requirements, the remaining 35% represents unserved demand — revenue the memory makers have not yet captured and capacity the industry has not yet built.

For a portfolio manager, that gap supports the case for holding memory stocks. El-Shawarby frames it explicitly: demand significantly outpaces supply, reinforcing strong sector fundamentals and supporting conviction in positioning.

She also flags a valuation disconnect in memory stocks. The market, in her view, has not fully priced what the supply chain participants themselves are telling her about 2027 visibility. Note the caveat that follows every promise: this is one manager's read of a finite set of meetings, not a company disclosure or a reported financial figure.

What could break the thesis?

El-Shawarby names the risk factors herself. Investors should watch two things closely.

First, hyperscaler capital expenditure trends. The memory shortage only persists as long as the data center build-out continues, and that build-out depends on the spending decisions of a small number of US technology companies.

Second, the sustainability of AI investment and monetization on the client side — primarily in the United States. If end clients cannot convert AI spending into revenue, the demand behind the 65% figure erodes.

Neither risk has materialized in her assessment. Both sit outside the memory makers' control.

What is the investment conclusion?

VanEck's on-the-ground conclusion is that the AI memory theme remains intact, with supplier behavior — long-term agreements, 2027 comfort, rationed allocation — confirming what balance sheets alone cannot. Whether the 35% unmet demand gets priced into memory stocks before hyperscaler capex peaks is the question investors will now watch.

VanEck is a global asset management firm founded in 1955, overseeing $161.7 billion in assets as of September 30, 2025.