Memory chip shortage deepens as Sandisk CTO warns of a “memory‑centric” AI race
Global demand for memory chips surges as AI growth drives a tight market; customers are actively seeking long‑term supply deals amid constrained NAND flash capacity.
KUALA LUMPUR — The executive technology team at Sandisk says the global artificial intelligence race is increasingly “memory‑centric,” a shift that is intensifying a memory chip shortage and prompting customers to lock in supply agreements. Sandisk chief technology officer Alper Ilkbahar told reporters that demand for NAND flash and other memory components is rising rapidly as AI workloads scale, putting pressure on production capacity. The comment comes as device makers, cloud providers and data centers compete for limited inventories of high‑performance memory.
Sandisk CTO Says Memory Is Central to the AI Computing Race
Alper Ilkbahar described the AI era as one where memory architecture and capacity are becoming as important as raw processing power. He argued that neural networks and large language models require vast volumes of high‑speed NAND and DRAM to feed accelerators efficiently. That shift, he warned, is changing procurement dynamics across the technology industry.
Ilkbahar’s comments highlight why memory chip shortage concerns have moved beyond short‑term inventory swings to structural market pressures. Companies that once prioritized CPUs and GPUs are increasingly focused on memory supply to sustain model training and inference at scale.
Customers Are Actively Seeking Long‑Term Memory Supply Deals
Industry sources say major cloud operators and device manufacturers are pursuing multi‑year contracts and advance purchase agreements to secure memory chips. These deals aim to guarantee deliveries of specific NAND configurations and low‑latency modules that are now in short supply. The scramble for supply reflects the central role memory plays in system performance for AI workloads.
Smaller customers are being squeezed as large buyers gain leverage to secure priority allocations from suppliers. That dynamic is prompting some firms to redesign systems to use different memory tiers or to negotiate co‑investment arrangements with vendors.
AI Workloads and Data Growth Driving Sustained Demand for NAND
Analysts point to a confluence of factors behind the memory chip shortage, chief among them the exploding size of AI models and the datasets they require. Training state‑of‑the‑art models consumes large volumes of NAND for scratch storage, checkpoints and caching, while inference at scale multiplies memory use across distributed systems. At the same time, consumer devices and edge applications continue to demand high‑capacity flash for features like on‑device AI.
This persistent demand pattern means memory consumption is not merely cyclical but shows structural growth, complicating capacity planning. Manufacturers face long lead times to retool fabs and expand production, making shortfalls more likely to persist even as prices fluctuate.
Manufacturers Respond with Capacity Expansion and Product Focus
Memory suppliers have signaled plans to boost output of advanced NAND and DRAM, prioritizing higher‑density parts used in data centers and enterprise storage. Investments include additional production lines and process upgrades aimed at increasing yields for multi‑level cell and next‑generation flash technologies. Some vendors are also reallocating wafer starts away from lower‑margin consumer products toward enterprise‑grade memory.
Despite announced expansions, industry executives caution that ramping new capacity takes months to years, and that production cycles for fabs and packaging remain constrained. The timing gap between rising AI demand and meaningful increases in supply is a central reason customers are pushing for long‑term contracts.
Pricing, Product Mix and the Broader Market Impact
The tightening market has begun to influence pricing for key memory categories, with premium NAND and performance DRAM commanding higher premiums in secondary and spot markets. This pricing pressure can squeeze OEM margins and influence product road maps, encouraging device makers to bundle storage differently or delay upgrades. It also affects the competitive landscape for cloud services, where memory costs feed into total infrastructure expenses.
Market observers note that stronger prices can accelerate investment in capacity, but may also prompt customers to pursue design changes that reduce reliance on the most constrained parts. Those changes include shifting workloads across memory tiers, optimizing model architectures for lower memory footprints, and increasing use of compression and offloading strategies.
Supply Chain Risks and Regional Implications for Asia
The memory chip shortage carries particular implications for countries with large semiconductor supply chains, including Japan and other parts of Asia. Manufacturers, contract assemblers and materials suppliers in the region may face order volatility as customers secure long‑term supplies and reconfigure sourcing. For Japanese technology firms, the landscape creates both risk and opportunity depending on their roles in the value chain.
Policy responses and corporate strategies will matter for regional resilience, with potential for governments to incentivize local capacity or support ecosystem upgrades. Companies that can offer flexible manufacturing, packaging or materials services may find new business from buyers seeking to diversify supply.
Industry watchers say the memory chip shortage driven by the memory‑centric AI race will influence technology investment and procurement strategies for the foreseeable future. Companies across the hardware and cloud ecosystem are adjusting to a market where memory availability and configuration choices are critical factors in competitive performance and product planning.
As demand for NAND flash and high‑performance memory continues to escalate, the industry faces a period of rebalancing in which production, pricing and procurement practices will determine how quickly shortages ease and which players gain advantage.
