Memory Squeeze on AI Demand: Memory Takes 60% of Costs, Device Redesigns Spread

Memory shortages originating from expanding demand for AI data centers are directly affecting manufacturing sites for smartphones and laptop PCs. Reporting by Kunal Khullar of Tom’s Hardware says that, centered on small manufacturers, difficulty in obtaining supply is becoming more severe than price surges, driving changes to board designs and revisions to procurement methods. The share of memory in component costs is rising sharply, with some cases forcing a complete rebuilding of product planning itself.

Tight Supply Hits Small Manufacturers Hard

According to reports, the core of this memory shortage lies not only in rising prices but in constraints on supply volume. Reuters interviews report that availability itself has become the biggest challenge, with companies unable to secure the quantities they need even after placing orders. Smaller manufacturers with weaker purchasing power than large companies are hit hardest. In some cases supply cannot be secured even with additional budget, leading to delays in production plans and downward revisions to shipment volumes.

Behind this tightness is a structure in which AI data centers are absorbing global memory and storage supply. Demand for high-bandwidth memory and large-capacity DRAM for servers has surged, with the knock-on effects spilling over to general-purpose memory for consumer devices. SK Hynix CEO Kwak Noh-jung said in July 2025 that 2027 would be the most difficult year of this shortage, adding that tight conditions could continue until 2030.

Among consumer devices, affordable mass-market products are hit particularly hard. Rising memory costs account for a high share of total product cost, leaving limited room to pass costs on to prices. As a result, manufacturers are being forced to respond on both design and sales strategy.

Changing Cost Structure: Memory Accounts for 60%

The case of Fairphone, which makes repairable smartphones, symbolizes the change in component cost structure. CEO Raymond van Eck explained that in devices priced around $400, memory can account for about 60% of materials costs.

Fairphone CEO Raymond van Eck said that in devices priced around $400, memory accounts for about 60% of materials costs

Reporting by Kunal Khullar of Tom’s Hardware notes that this high ratio is squeezing profitability for small manufacturers. While displays and processors traditionally dominated costs, soaring memory prices have drastically transformed the cost structure. When a single component accounts for 60% of costs, cuts to other components cannot absorb the impact.

There are also concerns about the impact on the overall market. A Counterpoint Research report forecasts that global smartphone shipments in 2026 will fall 13.9% year on year to 1.08 billion units, the largest annual decline on record. If weakening demand coincides with higher component costs, more manufacturers — centered on affordable models — may be forced to make decisions on business continuity.

Avoiding Supply Risks Through Design Changes

In response to supply uncertainty, moves to make board design itself more flexible are emerging. Finnish device maker Jolla has designed two types of motherboards for the same product. It can support either combined packages or discrete chips, establishing a system that can switch depending on which components are available.

This approach aims to broaden procurement options by avoiding dependence on a specific memory package. When supply is unstable, anticipating substitution with alternative components from the design stage is key to maintaining production. Jolla’s example of securing design redundancy despite its small size can be seen as a useful reference for other companies.

Greater design flexibility also involves rethinking the development process. Component changes require additional verification work and firmware adjustments, necessitating a redesign of development resource allocation. An integrated system for managing procurement and design together is likely needed. When considering how to streamline development systems, a breakdown of branch strategies such as Git Flow vs. GitHub Flow vs. Trunk Based Development: 2026 Branch Strategy Comparison also offers insights from a change-management perspective.

Responses Through Modularization and

Advance Orders

Laptop maker Framework is taking a different approach that leverages modularization. Its products adopt a structure that allows users to replace memory themselves, offering the option to reuse memory removed from older machines or second-hand parts. Even amid limited supply, leaving room for users to secure components aims to extend product lifecycles.

On the procurement side, it uses non-cancellable advance orders. This is a method of confirming orders even before final prices or delivery quantities are fixed, in order to secure supply slots. It can be seen as a decision to avoid the risk of being unable to obtain supply, while accepting price fluctuation risk.

Pricing policies are also shifting. In July 2025, Framework temporarily almost doubled prices for 32GB and 64GB models of the Framework Laptop 13 Pro following cost updates from LPCAMM2 suppliers. It later announced price cuts, including retroactively for orders already shipped, after succeeding in securing inventory at lower cost. Quickly reflecting component cost fluctuations in prices has become a realistic response under unstable supply.

AI Data Centers Squeeze Supply and Demand

AI data centers are essential to a structural understanding of this memory squeeze. Servers used for training and inference of generative AI consume large volumes of high-capacity, high-bandwidth memory. As data center operators lock up supply through long-term contracts, spot supply for consumer devices is thinning.

Major memory makers including SK Hynix are shifting production toward high-value-added products for AI. Prioritizing production of high-bandwidth memory over general-purpose DRAM can be seen as a move to improve profitability. As a result, spare supply capacity for general-purpose memory for smartphones and PCs is shrinking further.

This supply-demand structure goes beyond short-term price fluctuations. Assuming AI demand continues to expand, consumer device makers will need to build product plans that permanently factor in supply constraints. Designs that do not assume guaranteed memory supply and adoption of alternative technologies are emerging as medium- to long-term challenges.

Counterfeit Risks and Stronger Inspection Systems

The shortage is also creating secondary risks. In the process of trying to secure hard-to-obtain memory, there are concerns about refurbished products being circulated as new and the infiltration of counterfeit chips that do not meet specifications. Jolla has introduced a system to sample every received lot and inspect whether memory delivered as new is actually refurbished.

Such inspections add costs for small manufacturers. They are forced to divert resources that should go to quality assurance before mass production to authenticity checks on arrival. Because incorporating counterfeit parts directly affects product reliability and safety, skipping inspections is not an option.

Transparency across distribution is also being questioned. As supply thins even through authorized distributors, greater reliance on broker-mediated procurement widens potential routes for counterfeit infiltration. Balancing diversification of procurement sources with quality assurance is becoming a new focus of supply chain management.

Ripple Effects Across the Industry and

Future Focus

The impact of the memory shortage is not limited to specific manufacturers or product categories. It could spread to a wide range of DRAM-equipped products beyond smartphones and laptop PCs, including embedded devices and networking equipment. Especially for low-priced products, higher memory costs are likely to feed directly into sales prices, and some companies are expected to consider exiting the market or shrinking product lines.

Meanwhile, improving memory efficiency on the OS and driver side is also an important issue. Kernel optimization and improved memory management to maintain performance with limited memory capacity could help ease hardware constraints. On the kernel side, memory-related improvements such as those reported in Linux 7.2-rc1 Released, Integrating AMDGPU HDMI 2.1 FRL and Cache Aware Scheduling indirectly support greater freedom in product design.

The future focus is on when supply constraints will ease. As SK Hynix’s outlook suggests, in a scenario where tightness bottoms out in 2027 and continues until 2030, manufacturers will need structural strategic shifts rather than temporary responses. Efforts such as design standardization, component reuse, and long-term procurement contracts could become embedded in standard product development processes.

Editorial Opinion

In the short term, design redundancy and front-loaded procurement are likely to spread rapidly among small manufacturers. Jolla’s dual-board design and Framework’s modularization could spread to other companies as survival strategies under supply uncertainty. Over a 3- to 6-month horizon, prioritizing component procurement is likely to increase verification workloads, causing continued product launch delays and modest price increases. Inspection costs for counterfeit countermeasures are also an element easily passed on to prices.

In the long term, the memory market could become fixed in a split between AI data centers and consumer devices. Over 1 to 3 years, production concentration on high-bandwidth memory will likely continue, keeping general-purpose DRAM supply structurally tight. On the consumer device side, software optimization that maintains performance while limiting memory capacity, and circular product design premised on memory reuse, are likely to become sources of competitiveness. Ensuring supply chain transparency will also become an element directly tied to brand value.

This situation raises the question of whether component availability should be built into design assumptions. The issue is how to evaluate not only price but also certainty of supply, and reflect it in product specifications.

References

• “ AI-induced memory shortage is changing how devices are built, Fairphone says memory now 60% of materials cost — smaller laptop and phone makers are redesigning products and have to test for fake chips “, by Kunal Khullar — Tom’s Hardware, 2026-09-16T12:15:42.000Z (ARR)
• Source URL: https://www.tomshardware.com/pc-components/ram/ai-induced-memory-shortage-is-changing-how-devices-are-built-fairphone-says-memory-now-60-percent-of-materials-cost-smaller-laptop-and-phone-makers-are-redesigning-products-and-have-to-test-for-fake-chips