MT

INDICATOR AUDIT · MEMORY CYCLE

Memory cycle score

A 0–100 research indicator of memory-fundamental heat and market confirmation. It is neither a probability of stock gains nor a forecast date for a market top.

MODEL · V1.2.0All inputs within freshness limitsConfidence: medium

Current output · 2026-09-23

81/100

Cycle heat is above the 80-point overheating threshold. Prices, supply-demand, AI demand and equities jointly confirm strength, but this does not imply further stock gains.

Unrounded result81.5
How the current score is calculated
Four components are normalized to 0–100 and multiplied by fixed weights. The page, homepage and public API now consume the same program output.
0.25 × price + 0.30 × supply + 0.25 × AI + 0.20 × equity confirmation
Price momentum62.8/100
Weight 25%Contribution 15.7
Supply tightness81.3/100
Weight 30%Contribution 24.4
AI demand intensity85.6/100
Weight 25%Contribution 21.4
Equity confirmation100/100
Weight 20%Contribution 20
Weighted total15.7 + 24.4 + 21.4 + 20.0 = 81.5 → 81
Input freshness
Daily market data has a four-day weekend buffer; research forecasts use 120–180 days or event-driven updates. A stale input prevents the page from calling the score live.

TrendForce DRAM spot

Current

2026-09-23 · 0d / 4d

TrendForce DRAM spot

Current

2026-09-23 · 0d / 4d

TrendForce DRAM spot

Current

2026-09-23 · 0d / 4d

吴梓豪先生DRAM bit出货预测

Current

2026-06-30 · 85d / 180d

MemoryTicker demand model

Current

2026-08-08 · 46d / 120d

MemoryTicker AI rack model

Current

2026-08-08 · 46d / 120d

Yahoo Finance adjusted closes

Current

2026-09-21 · 2d / 4d

Calculation trace

Raw values, normalization, data type and source are shown together. clamp bounds a result to 0–100 so an outlier cannot grow without limit.

InputRaw valueNormalizationScoreEvidence
DDR5 monthly change6.15 %clamp(6.15 / 8.00 × 100)76.9Observed
DDR4 8Gb monthly change6.02 %clamp(6.02 / 15.00 × 100)40.1Observed
Weekly DRAM breadth66.7 %66.7% of tracked DRAM series positive66.7Observed
2027 supply-gap ratio4.83 %(47.41 - 45.12) / 47.41 × 100; 6% = 10080.5Forecast
Demand growth over supply growth5.75 pptdemand growth 18.97% - supply growth 13.22%; 7ppt = 10082.1Forecast
AI-implied data-center growth34.08 %5.839 / (39.85 × 0.43) × 100; 40% = 10085.2Model estimate
Data-center DRAM share43 %43.0 / 50 × 10086Model estimate
Average 30-day memory-equity return10.18 % / 30dclamp((10.18 + 25) / 25 × 100); -25% = 0, 0% = 100100Observed
Is it scientific? The precise answer: reproducible, but not fully validated
No cross-cycle validation yet

What is rigorous today

A standalone calculation function produces the same result from the same inputs
Weights, thresholds, rounding and the full trace are disclosed
A missing configured input fails the calculation instead of silently reweighting
Automated tests verify the score, total weights and 0–100 bounds

What is still missing

No out-of-sample test across several complete DRAM cycles
Full-score thresholds such as 8% and 15% are expert-set, not statistically estimated
Contract prices, channel inventory and consensus EPS revisions lack connected continuous series
The current score cannot be read as an equal probability of gains or a certain equity top

How to use the score

It answers how hot the cycle is and whether the market still confirms it—not how much a stock can rise. 60–79 is hot and 80+ is overheated. A more credible top signal requires a persistent score decline alongside weaker contract prices, inventory and EPS revisions.

View machine-readable resultFull methodology