AI-driven demand is diverting manufacturing capacity away from NOR Flash and SLC NAND toward higher-margin products, creating a severe undersupply for legacy electronics.

The memory chip that enables a router to function operates in a completely different sphere from the sleek GPUs driving eye-popping investments, staggering valuations, and soaring stock markets. These smaller chips have historically been inexpensive and rely on mature, decades-old technology. Despite their distance from the GPU spotlight, prices for these frequently overlooked chips are skyrocketing amid a comprehensive memory price crisis fueled by the AI boom. While public and media attention remains fixed on premium memory products, older, lower-margin chips are experiencing equally severe impacts. High-bandwidth memory (HBM) is essential for AI accelerators, while DRAM and high-capacity NAND are being absorbed by rapidly expanding data centers. According to a June report from Morgan Stanley, memory prices have surged more than sixfold over the past year, reversing decades of steady cost declines driven by increased production.

The shortage extends beyond HBM and DRAM, reaching older memory technologies such as NOR flash and single-level cell (SLC) NAND. Morgan Stanley projects that NOR flash will remain undersupplied through 2026, while JPMorgan has cautioned that its own forecasts may not fully account for a looming SLC NAND supply crunch. Market indicators already reflect this shift: TrendForce reports that contract prices for both NOR flash and SLC NAND increased by more than 100% in the first half of 2026, with SLC NAND prices expected to climb an additional 120% to 170% in the second half relative to the first.

Rising costs will directly affect everyday technology. NOR flash is widely used to store boot and program code, serving as a critical component in automotive, industrial, and networking equipment. SLC NAND is similarly deployed across numerous applications due to its reliability and endurance in embedded hardware designed for long operational lifespans. Both technologies are vital, yet they are increasingly sidelined in favor of higher-margin products, pushing established components into unprecedented supply constraints. “The SLC NAND market is probably under a billion dollars a year,” Jim Handy, a semiconductor and SSD analyst at Objective Analysis, told Tom’s Hardware Premium. This limited market size creates a significant bottleneck, as it discourages manufacturers from funding new capacity.

“What you’ve got going on is a purely economic phenomenon,” Handy noted. Hyperscalers and cloud providers are “all trying to outspend each other,” channeling unprecedented capital into semiconductors to construct AI infrastructure. This aggressive spending ensures that the most lucrative customers secure priority access to manufacturing capacity. Firms such as Nvidia, Broadcom, and Marvell require massive semiconductor fabrication capacity for AI system chips. “They’re sucking up all of the wafers,” Handy stated. “And then the companies who make NOR flash and SLC are having a hard time getting wafers to build their product, and so they have to raise prices.”

The disparity is even more pronounced in the NAND sector. Bryan Ao, research manager at TrendForce, told Tom’s Hardware Premium that major manufacturers—including Micron, Kioxia, and SK Hynix—have reduced wafer allocation for SLC production in favor of newer, more profitable NAND technologies. Ao estimates that a 12-inch wafer dedicated to mainstream NAND can yield nearly $20,000 in revenue, whereas dedicating the same wafer to SLC generates only $6,000 to $8,000. Even if smaller SLC suppliers in China and Taiwan wished to expand, they cannot rapidly ramp up production. Lead times for certain semiconductor manufacturing equipment have extended to 12 to 15 months, according to Ao. The resulting bottleneck has created what he describes as a “severe undersupply.”

Financial institutions project sustained price increases. BNP Paribas forecasts that the average NAND price will reach $279.50 per terabyte in 2026, a sharp rise from $73.10 in 2025. JPMorgan anticipates the shortage will endure for at least two more years, leaving customers able to fulfill only 70% to 80% of their orders. Meanwhile, TrendForce notes that manufacturers are redirecting capacity toward advanced, higher-value memory products, further constraining mature process nodes.

Some technical alternatives exist. Kioxia, for instance, positions its serial SLC NAND as a substitute for NOR flash. However, migrating existing industrial or networking designs to a different chip requires substantial engineering and qualification efforts. Furthermore, memory manufacturers lack strong incentives to resolve the shortage by building new SLC capacity; investing billions in facilities that could face obsolescence is financially unappealing. This dynamic creates a structural paradox: insufficient demand to justify new fabrication plants, yet persistent demand that outpaces a contracting supply. Hardware manufacturers face a difficult choice: absorb the rising component costs and compress margins, or transfer the expense to end users. “We’ll just have to either have lower margins, or we’ll have to raise the prices to the consumer,” Handy concluded.

The shortage appears structurally entrenched, with no near-term resolution. Ao projects that memory prices will remain elevated for the next five years, with no expectation of returning to 2023 or 2024 baselines. This aligns with TrendForce’s assessment of a structural deficit, noting that manufacturers have announced no significant capacity expansion plans for NOR flash or SLC NAND. Handy identifies a potential catalyst for change, though it offers little comfort. “As long as the race between the hyperscalers keeps up to spend, then it will continue to be an issue,” he warned.

Handy draws parallels between the current AI infrastructure buildout and the telecommunications boom of the late 1990s. Rather than eventual capacity additions restoring market equilibrium, he cautions that excessive capital expenditure may simply exceed what the market can economically sustain. “I’m expecting the same kind of a thing to happen here that we’ve got too many people spending too much money on AI, and not really making any return on it yet,” he added. Until market dynamics shift, the most utilitarian memory chips in computing systems risk becoming the most difficult to source. As Ao summarized: “Pretty much we have to get used to this high price, no matter which segment of memory.”