AI Data Centers Are Eating Away the World's Memory Chips

The same AI infrastructure investment that retailers are racing to deploy is now raising the cost of the hardware their businesses and customers depend on.

A coalition of nine US trade associations made that argument explicitly in a letter to Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick on June 3, 2026, warning that AI data center expansion is creating a memory chip shortage with direct consequences for retail supply chains, consumer electronics pricing, and the IT infrastructure retailers rely on to operate.

Signatories include the National Retail Federation, the Retail Industry Leaders Association, the Alliance for Automotive Innovation, AdvaMed, the Medical Device Manufacturers Association, The Internet and Television Association, the Telecommunications Industry Association, ACA Connects, and The Rural Broadband Association.

"Expanding artificial intelligence data centers consumes an enormous share of available memory chip capacity," the coalition wrote. "The result has been an unprecedented surge in the price of memory chips and reduced supply of these chips for manufacturing and consumer-facing industries."

How the Shortage Is Being Created

Samsung, Micron, and SK Hynix, which together control over 95% of global DRAM production, have increasingly redirected manufacturing capacity toward high-bandwidth memory and high-performance DRAM, the specialized chips that power AI data centers.

That redirection has been deliberate and accelerating. HBM commands significantly higher margins than standard DRAM, and with AI infrastructure spending at $650 billion in 2026 alone, the economic incentive to prioritize data center customers over consumer electronics manufacturers is straightforward.

AI data centers are projected to consume roughly 70% of total global memory chip output by the end of 2026.

TrendForce reported that conventional DRAM contract prices rose roughly 93% to 98% quarter over quarter in Q1 2026 alone, a price movement that far outpaces inflation in any other input category retailers are managing.

DRAM prices surged over 60% in 2025 on top of that baseline. Both Samsung and SK Hynix have warned publicly that significant constraints will continue through at least 2027.

New Street Research analyst Pierre Ferragu noted in a note to investors this week that agentic AI is expanding memory demand beyond high-bandwidth memory into conventional DRAM and NAND, the commodity memory that consumer electronics, automotive, and retail IT infrastructure depend on.

"Prices for those have already jumped 70 to 80% quarter over quarter," Ferragu wrote. The implication is that the shortage is no longer isolated to a specialized chip category. It is moving into the same supply pool that retailers and manufacturers draw from.

IDC has already revised its 2026 PC market forecast downward by up to 9% as a direct consequence of memory scarcity. Synopsys CEO Sassine Ghazi has predicted the shortage will persist into 2027.

What It Means for Retailers Specifically

Memory chips are embedded in the devices consumers buy and in the infrastructure retailers operate. Laptops, tablets, smartphones, smart home devices, and connected appliances all depend on the same DRAM supply that AI data centers are now consuming at an accelerating rate.

Cost increases at the chip level flow to manufacturers and eventually to merchants, who face a hard choice. Absorb the hit to margin or pass it to consumers at a time when price sensitivity is already elevated across most retail categories.

Beyond consumer-facing product costs, chip shortages can reduce inventory availability, disrupt sourcing timelines, and inflate the cost of the IT infrastructure retailers rely on for point-of-sale systems, inventory management, and ecommerce fulfillment.

Retailers have spent the last few years investing heavily in AI-powered demand forecasting, personalization, and supply chain optimization. The coalition is acutely aware of the irony: AI investments driving efficiency are being partially offset by AI‑driven input cost increases.

The retail industry's inclusion in the letter is significant precisely because it places retailers on both sides of the AI infrastructure equation simultaneously, investing in AI to drive efficiency and growth while being squeezed by the same infrastructure buildout through higher hardware costs and potential supply chain disruption.

Stephen Sopko, semiconductor and deep tech analyst at HyperFrame Research, said the effects are no longer confined to AI infrastructure. "The risk to watch is demand destruction outside of AI," New Street Research's Ferragu added. "Samsung and Hynix are already flagging softening mobile and PC demand."

The Policy Response and What Comes Next

The coalition urged the Trump administration to work with memory chipmakers and chip buyers to address the market imbalance, calling for faster and larger memory chipmaking capacity expansions in the US.

The specific asks include addressing what the coalition described as "an urgent imbalance in the market for memory chips" that could hinder procurement obligations across critical US industries.

Commerce Secretary Lutnick has previously signaled a confrontational posture toward foreign memory makers. "Everyone who wants to build memory has two choices: they can pay a 100% tariff, or they can build in America," he said at the groundbreaking of a Micron facility outside Syracuse, New York — where Micron, the sole US-based member of the three dominant memory producers, is receiving CHIPS Act support for domestic capacity expansion.

The challenge is that meaningful domestic capacity increases are still years away. CHIPS Act investments in memory manufacturing were designed with a multi-year buildout horizon.

The shortage the coalition is warning about is happening now, in 2026 and 2027. The policy tools available to address it in the near term including production incentives, allocation frameworks, or tariff pressure on foreign chipmakers, are either slow-moving or carry their own downstream cost risks.

AI infrastructure is not a parallel economy that operates independently of the rest of the supply chain. It competes for the same components, the same manufacturing capacity, and the same physical resources as the industries it is supposed to transform.

The memory chip shortage is the first moment that tension has been documented publicly, at scale, with named industry groups and a direct appeal to federal intervention.