Advanced Demand Forecast Simulator

이 simulator는 단순 계산기가 아니라 고차원 forecast cockpit입니다. 기존 domain별 계산기는 특정 시점의 units × mix × content를 투명하게 보여주는 데 강하지만, 실제 메모리 수요 예측에는 adoption transition, content CAGR, 가격 탄력성, 채널 재고, bullwhip, supply allocation이 동시에 작동합니다.

모델 계층

text

End demand path
= base demand
× unit growth index
× content growth index
× S-curve adoption/content uplift index

Cycle-adjusted demand
= end demand path × price-inventory-cycle factor

Supply-realized demand
= min(cycle-adjusted demand, available supply × yield × allocation)

Unmet demand
= max(0, cycle-adjusted demand - supply-realized demand)

확장 분석 계층

Stress scenarios: demand shock, supply shock, price shock를 별도 입력으로 두어 base case가 어느 방향으로 깨지는지 확인합니다.
Uncertainty envelope: downside/base/upside/supply-squeeze 결과를 연도별 low/base/high band로 접어 봅니다.
Hierarchy reconciliation: HBM/NAND/DRAM/DRAM+NAND preset을 family별로 합산해 segment forecast와 전체 forecast의 정합성을 점검합니다.
Source-calibrated nowcast overlay: SIA 월간 매출, 기관별 2026 memory revenue band, HBM 가격/BOM CSV를 wiki/data/advanced-forecast-source-signals.json으로 변환해 bounded adjustment factor와 institution forecast band KPI로 표시합니다.
CSV-derived preset candidates: institution low/mid/high band와 SIA momentum을 보수/기준/상방 candidate preset으로 자동 생성해 버튼 클릭만으로 base demand, unit growth, allocation 가정의 민감도를 비교합니다.
SIA indexed time-series: SIA sales row를 2025-FY = 100 index와 bar chart로 바꿔, 월간/분기별 nowcast 추세를 표와 별개로 시각적으로 확인합니다.
Institution-weighted ensemble tuning: WSTS/TrendForce/IDC/Gartner raw CSV revenue view를 slider weight로 조정해 weighted revenue, demand multiplier, peak effective demand를 즉시 갱신합니다.
HBM/BOM stress controls: HBM price decline, target BOM share, demand elasticity를 affordability stress scenario로 변환해 adjusted HBM cost와 unmet demand를 확인합니다.

사용 규칙

Unconstrained demand는 최종 소비/설치 수요가 만드는 이론적 경로입니다.
Supply-realized demand는 실제 공급·수율·allocation gate를 통과한 실행 가능 수요입니다.
Unmet demand는 수요가 사라졌다는 뜻이 아니라 backorder, price rationing, spec downgrade, delayed refresh로 전이될 수 있는 압력입니다.
price-inventory-cycle이 dominant constraint이면 주문량을 end demand로 그대로 읽으면 안 됩니다.
supply-realization이 dominant constraint이면 시장 TAM보다 CoWoS/HBM/NAND wafer/DDR allocation이 near-term cap입니다.

구현 파일

wiki/lib/advancedDemandForecast.ts
wiki/data/advanced-forecast-presets.json
wiki/data/advanced-forecast-source-signals.json
wiki/components/AdvancedDemandForecastSimulator.vue
tests/advancedDemandForecast.test.ts
tests/advancedForecastUi.test.ts
API extensions: stress scenario table, uncertainty envelope, hierarchical forecast rows, source-calibrated nowcast adjustment, CSV-derived preset candidates, SIA indexed time-series, institution-weighted ensemble, HBM/BOM affordability stress

Source traceability

Forecast preset	Source-backed axes	Raw source anchors
HBM AI infrastructure constraint	GPU content uplift, HBM growth, package/supply realization, datacenter constraints, long-term TAM band	`raw/articles/aisilicon-nvidia-amd-hbm-capacity-roadmap-2026.md`, `raw/articles/nvidia-gpu-hbm-capacity-roadmap-2026.md`, `raw/reports/trendforce-hbm-bit-demand-supply-2026.md`, `raw/reports/pkg-cowos-allocation-nvidia-amd-2026.md`, `raw/reports/hbm-longterm-tam-by-application-2030.md`
PC DRAM shortage elasticity	AI PC content, PC shipment drag, memory shortage price pass-through, refresh cycle	`raw/reports/idc-2026-memory-shortage-supply-growth.md`, `raw/articles/oem-pc-dell-hp-lenovo-price-hike-memory-2026.md`, `raw/articles/oem-windows10-eol-pc-refresh-memory-demand-2026.md`, `raw/articles/microsoft-ai-pc-16gb-ram-baseline.md`
Smartphone on-device AI premiumization	on-device AI DRAM uplift, premium mix, storage capacity, spec downgrade risk	`raw/articles/on-device-ai-smartphone-dram-content-2026.md`, `raw/articles/trendforce-smartphone-storage-capacity-2026.md`, `raw/articles/oem-smartphone-dram-content-spec-downgrade-2026.md`, `raw/articles/on-device-slm-smartphone-ram-2026.md`
NAND AI storage QLC breakout	QLC enterprise SSD, AI/data-center storage mix, capacity-per-drive uplift	`raw/reports/trendforce-qlc-enterprise-ssd-nearline-hdd-2026.md`, `raw/articles/storage-enterprise-ssd-qlc-nand-demand-2026.md`, `raw/datasets/idc-global-datasphere-zettabytes.md`
Ensemble / envelope / nowcast	Monte Carlo-style stress, MinT-style hierarchy reconciliation, high-frequency nowcast, institution forecast band	`raw/papers/monte-carlo-simulation-demand-forecasting-uncertainty.md`, `raw/papers/hierarchical-forecast-reconciliation-mint.md`, `raw/papers/nowcasting-google-trends-bayesian-structural-time-series.md`, `raw/datasets/sia-global-semiconductor-sales-monthly-2026.md`, `raw/datasets/institutional-memory-forecast-comparison-2026.md`

CSV signal interface

시뮬레이터의 “Raw CSV signal interface” 섹션은 raw CSV를 직접 mutable app data로 바꾸지 않고, 다음 경로를 유지하는 파생 JSON을 사용합니다.

UI signal	Derived from	Display role
Institution forecast band	`raw/datasets/institutional-memory-forecast-comparison-2026.csv`	2026 memory revenue low/mid/high band ($294.82B–$633.3B)
CSV-derived preset strip	`raw/datasets/institutional-memory-forecast-comparison-2026.csv`, `raw/datasets/sia-global-semiconductor-sales-monthly-2026.csv`	low/mid/high institution band와 latest SIA momentum을 보수/기준/상방 preset candidate로 변환
SIA high-frequency nowcast	`raw/datasets/sia-global-semiconductor-sales-monthly-2026.csv`	monthly sales YoY/MoM 기반 nowcast adjustment factor와 indexed time-series bar chart
HBM price/cost pressure	`raw/datasets/hbm-price-per-gb-per-stack-by-generation.csv`, `raw/datasets/ai-accelerator-bom-hbm-cost-share.csv`	HBM3/3E/4 price ladder와 accelerator BOM cost share context
Institution-weighted ensemble	`raw/datasets/institutional-memory-forecast-comparison-2026.csv`	기관별 revenue row 가중치를 slider로 조정해 weighted forecast를 산출
HBM/BOM affordability stress	`raw/datasets/hbm-price-per-gb-per-stack-by-generation.csv`, `raw/datasets/ai-accelerator-bom-hbm-cost-share.csv`	HBM 가격 하락률과 BOM share 민감도를 stress scenario shock로 변환

각 derived row는 rawPath와 sourceNote를 포함하므로, 화면의 KPI가 어떤 CSV와 paired markdown source note에서 왔는지 추적할 수 있습니다.

Advanced Demand Forecast Simulator ​

모델 계층 ​

확장 분석 계층 ​

사용 규칙 ​

구현 파일 ​

Source traceability ​

CSV signal interface ​

관련 문서 ​