Memory chips are now more expensive than compute chips on a per-area basis
- ID
- 27197
- Status
- summarized
- Published
- 22 Sep 2026, 7:12 PM
- Fetched
- 22 Sep 2026, 10:32 PM
- Provider
- Tom's Hardware
- Category
- technology
- Original URL
- https://www.tomshardware.com/pc-components/dram/dram-is-now-more-expensive-than-compute-chips-on-per-area-basis-ai-demand-drives-memory-die-value-past-leading-edge-silicon
- Source URL
- https://www.tomshardware.com/feeds/all
Summary
- Score
- 6.5
- Created
- 22 Sep 2026, 10:36 PM
- Tags
- Audience
- ai_ml_learnerssaas_startup_foundersdevelopers
What happened
DRAM die value has surpassed leading-edge compute silicon on a per-area basis, driven by AI demand for memory. The article reports that memory chips are now more expensive to manufacture than compute chips when normalized by die area.
Why it matters
If memory is now the cost bottleneck rather than compute, builders running inference or fine-tuning should expect GPU/accelerator pricing and cloud instance costs to stay elevated or rise further, especially for memory-heavy workloads like long-context LLMs. This shifts the economics toward memory-efficient model choices (smaller context windows, quantization, KV-cache optimization) over raw compute optimization.
Discussion angle
How should a Malaysian startup or indie builder factor memory-cost inflation into model selection — is it worth switching to smaller-context or quantized models now to hedge against rising per-token inference costs?