Morgan Stanley: Memory Prices Surge 6x in One Year, Beyond Monetary Policy's Reach
chuong wang
Memory prices have surged more than sixfold in a year, and Morgan Stanley says this 'Chipflation' largely bypasses CPI — instead feeding through PPI and corporate cost chains, beyond the reach of conventional monetary policy.
Memory prices jumped sixfold — where is the money going?
Memory prices rose more than sixfold over the past year, yet bit shipments grew only about 30% annually. This means → nearly all of the spending surge comes from higher prices, not higher volumes.
In plain terms = buyers are paying far more money but getting barely more memory — the gap is pure price inflation.
TrendForce projects the global memory market will leap from roughly $220 billion in 2025 to about $890 billion in 2026 — an increment of roughly $600 billion, larger than the entire smartphone, PC, or server market individually.
Why can't supply catch up?
Three DRAM makers control about 90% of DRAM capacity and virtually all HBM — high-bandwidth memory, an ultra-fast memory designed specifically for AI chips — supply. This oligopoly locks the pace of expansion.
HBM consumes 3–4× as much wafer capacity per usable bit as conventional DRAM. Its share of advanced memory wafers will rise from about 6% in 2023 to roughly 34% by 2028. This means → the same wafer pool is increasingly consumed by HBM, squeezing the share left for standard memory.
Morgan Stanley estimates that even with total DRAM wafer capacity expanding about 30% by 2027, PC DRAM will face a shortfall of roughly 15% (about 58 million units) and smartphone DRAM about 12% (about 134 million units).
Why doesn't this show up in CPI?
Households rarely buy memory directly. It enters products as an intermediate input — and computing devices carry a low weight in the consumer basket, so CPI barely registers this price wave.
The inflation shows up in electronics-component PPI (up about 30% year-on-year), corporate cost of goods sold, cloud-service bills, and capital-expenditure budgets.
In plain terms = you won't feel chip inflation at the supermarket shelf, but corporate procurement bills and cloud invoices have already taken the hit.
If manufacturers passed on the full cost, how much would devices rise?
Morgan Stanley estimates: smartphone average selling prices would need to rise about 34%, PCs about 67%, servers about 83%, and storage devices about 14%.
A 67% PC price increase would be the largest in the market's history — roughly 8–9× the inflation seen during the Covid pandemic.
In reality, manufacturers will not pass on the full amount. This means → the pressure instead surfaces as margin compression, spec reductions, and launch delays — consumers may pay more, or get less, or both.
Why can't central banks fix this kind of inflation?
Morgan Stanley draws a clear line: conventional inflation is a demand-management problem that central banks can cool with rate hikes. Chipflation is fundamentally a supply problem — three companies, two countries, multi-year build cycles. No interest-rate decision can add a single fab.
New capacity takes roughly two years from tool installation to yield ramp. This reflects a physical manufacturing cycle far slower than the transmission speed of monetary policy.
AI buyers are price-insensitive, and hyperscale cloud operators lock up supply with long-term contracts and prepayments. Everyone else competes for a smaller, more volatile residual pool. In plain terms = the deep-pocketed buyers book the inventory first; everyone else scrambles for leftovers, keeping prices stubbornly high.
Content is for reference only, not financial advice.