Methodology

What enters the ledger

Each row is one dated numerical observation with its own source, product scope, metric and reading type. Different forecast vintages are separate readings, not independent measurements of the same outcome. The ledger combines historical press releases, weekly price updates and public TrendForce / DRAMeXchange DRAM tables, NAND tables, TrendForce’s original pricing forecasts, Counterpoint public contract-price changes, original company disclosures and Micron’s fiscal 2026 Form 10-K, Item 7, page 56. Every non-empty field maps to a source URL and accessed date.

Four reading types

Spot: the publisher’s absolute spot average and reported percentage change, preserving its exact product description. Session low/high appear only when the source publishes them; weekly articles leave those fields empty. Module and graphics tables label their percentage as average change; other included spot tables label it session change.

Contract: the displayed contract survey or a labelled quarterly percentage change, with its own product and period. Counterpoint’s public changes are average contract-price estimates subject to revision; its gated dollar values are omitted. A contract survey is a market reading rather than a supplier’s offer to fill an order.

Forecast: the publisher’s estimate as of publication. Ranges stay ranges. January’s aggregate outlook was explicitly revised in the February release; those earlier rows are marked superseded. Other dated vintages remain forecasts even after their target quarter ends.

Reported ASP: a supplier-wide realized average selling price change. Product mix can move that average. Micron fiscal-quarter and fiscal-year results remain separate from SK hynix calendar-quarter results, market forecasts and component quotes. Verbal ranges such as “high-teens” retain the source wording and have no invented numerical midpoint. TrendForce estimates of Samsung and other supplier ASPs are attributed to TrendForce, not recast as issuer disclosures.

The date belongs to the table

Source date is the price-table update, article/forecast publication, source-page update or filing signature date. Counterpoint’s current public summary uses its displayed last-updated date; its period is the measured quarter. SK hynix’s financial release uses the explicit results date in the release, which can differ from the newsroom navigation date. Period holds the quoted session or target period. Verified date and the source’s accessed date record our check. All copied price-table timestamps retain GMT+8.

The public contract tables display September observations alongside notices for October reports. The public client SSD contract table remains 3Q26 alongside a 4Q26 report link. The newer notices do not change the dates of the displayed data. Dynamic source pages can advance after our check.

Currency, precision and comparability

Dollar columns use USD per listed item; TrendForce’s DRAMeXchange guide labels its price-table currency USD. Decimal precision follows the source. A Gb chip and a GB module are different items. A TLC wafer-market density quote is not the price of a whole physical wafer. The ledger does not normalize those products into a common price-per-GB series.

Change includes its sign and comparison basis. HBM’s annual blended ASP outlook and its per-Gb configuration premium have different denominators. Empty dollar cells on forecasts and supplier ASP rows mean that a dollar price was not published for that observation.

Market coverage

The tracker covers product-market readings rather than ranking manufacturers. TrendForce’s DRAM research identifies Samsung, SK hynix and Micron as the leading suppliers; its NAND report also covers Kioxia and SanDisk. Supplier-specific prices enter the ledger only when a numerical ASP change or quote has a verifiable public source.

Charts and coverage gaps

Build-time history charts keep absolute USD spot quotes in separate panels from QoQ contract-price forecast ranges. Publication date drives the horizontal axis; the exact-value table retains each target quarter. Forecast endpoints are the published range, not a confidence interval. No interpolation, compounding or inferred actual outcome is added. DDR4 speed grades remain separate series.

HBM readings are percentages: preliminary annual negotiations, blended ASP growth and a per-Gb configuration premium, each on its own basis. Spot history uses dated weekly and session observations.

Samsung’s interim reports give one combined memory ASP figure, so Samsung’s DRAM and NAND rows carry TrendForce’s research estimates under TrendForce’s name.

Inclusion and corrections

A reading needs a numerical price, numerical change, source-described percentage range or bounded forecast, plus a product scope, source date and period or comparison basis. Qualitative price direction and gated figures stay outside the ledger. The table starts with the newest source dates; every column can be sorted, and filters are stored in the URL.

Send a correction with the reading URL and a public source to hello@memorypricetracker.com. A substantive update changes the checked date. Sponsorship pays for a labelled placement beside the ledger and has no role in its observations.

Attribution and source rights

We quote individual facts with attribution. We never reproduce a TrendForce, DRAMeXchange or Counterpoint table wholesale: each public price table contributes a few benchmark rows, and the full table stays on the publisher’s page, linked from every reading. Forecasts and supplier commentary are cited as short facts with the publisher named.

CSV and JSON carry the editorial compilation and its provenance. Third-party research keeps its owners’ rights, and the downloads grant no licence to it. Cite the dated reading and its original publisher. See citation details.