Demand for high-bandwidth memory (HBM) from AI workloads has grown to the point where ordinary RAM now costs more than the most advanced chips produced by TSMC, according to a report by Gagadget.com.

The finding inverts a long-standing assumption in semiconductor economics: that logic chips manufactured on the most advanced process nodes carry the highest price tags in a system. According to the report, the scramble among memory makers to allocate production capacity to HBM — the stacked DRAM used in AI accelerators — is tightening supply of conventional DRAM and pushing its prices to levels that exceed those of TSMC's most advanced chips.

Why is AI making ordinary RAM more expensive?

The mechanism, as the report describes it, is capacity reallocation. HBM production consumes substantially more wafer capacity per gigabit than standard DRAM, because HBM stacks multiple DRAM dies and requires more complex packaging. When manufacturers shift their output mix toward HBM to serve AI customers, the supply of commodity memory modules available for PCs, servers and consumer devices shrinks.

The report frames the consequence in stark terms: the same AI boom that is filling order books for HBM suppliers is making ordinary RAM — the commodity memory that goes into everyday hardware — more expensive than the most advanced logic silicon TSMC produces.

What does this mean for the market?

For system builders, PC manufacturers and data center operators outside the AI segment, the report signals that memory has become a cost pressure point rather than a declining-cost component. Conventional DRAM has historically followed a price-decline curve as process technology matured; the current AI-driven distortion breaks that pattern.

For memory manufacturers, the incentive structure now favors HBM production disproportionately, since AI customers are willing to pay premium prices for the bandwidth that large-model training and inference require. That premium pulls fab capacity away from standard DRAM lines.

For TSMC, the comparison is less about competition than about context: the report uses the company's most advanced chips as a price benchmark to illustrate how far conventional memory pricing has climbed.

Who feels the impact?

The ripple effects, as the report outlines them, extend across the hardware industry:

• PC and laptop makers face higher BOM costs for standard memory configurations.
• Server builders outside AI workloads compete for a tighter DRAM supply pool.
• Consumers face higher prices for RAM upgrades and memory-equipped devices.
• Memory suppliers weigh the trade-off between premium HBM contracts and conventional DRAM volume.

What happens next?

The report does not forecast when the imbalance will correct. The outcome depends on how quickly memory makers expand capacity, how much additional HBM supply reaches the market, and whether AI-driven demand continues to absorb the premium segment of the memory business.

What the report does establish is the current state of the market: the AI buildout has repriced memory across the board, and ordinary RAM now sits above TSMC's most advanced chips on the cost scale — a reversal with direct consequences for anyone building or buying hardware that depends on commodity DRAM.