Verwachten
Prognoses per kwartaal met 80% en 95% betrouwbaarheidsbanden. Modelpeil 2026-09-18; 4 series.
95%-band 80%-band mediaan
conv_dram_contract_qoq_pct
Q+1..Q+3 (calendar quarters after 2026-10-01) · methode: Holt damped exponential smoothing (phi=0.8, backtest alpha=0.1, one-step rolling-origin selection; 95% band = 1.96*sigma, 80% band = 1.2816*sigma, z-fallback per forecasting-method
- Q+110,780%: -2,6 – 23,9
95%: -9,6 – 30,9 - Q+2-2,780%: -21,5 – 16,0
95%: -31,4 – 25,9 - Q+3-13,480%: -36,4 – 9,5
95%: -48,6 – 21,7
nand_contract_qoq_pct
Q+1..Q+3 (calendar quarters after 2026-07-01) · methode: Holt damped exponential smoothing (phi=0.8, backtest alpha=0.1, one-step rolling-origin selection; 95% band = 1.96*sigma, 80% band = 1.2816*sigma, z-fallback per forecasting-method
- Q+117,580%: -15,3 – 50,3
95%: -32,7 – 67,6 - Q+28,380%: -38,0 – 54,7
95%: -62,6 – 79,3 - Q+31,180%: -55,8 – 57,9
95%: -85,8 – 87,9
samsung_dram_asp_qoq_pct
Q+1..Q+3 (calendar quarters after 2026-07-01) · methode: Holt damped exponential smoothing (phi=0.8, backtest alpha=0.9, one-step rolling-origin selection; 95% band = 1.96*sigma, 80% band = 1.2816*sigma, z-fallback per forecasting-method
- Q+1-13,880%: -16,2 – -11,4
95%: -17,5 – -10,1 - Q+2-41,180%: -44,6 – -37,7
95%: -46,4 – -35,9 - Q+3-6380%: -67,2 – -58,8
95%: -69,4 – -56,6
nvme_2tb_street_price_usd
Q+1..Q+3 (calendar quarters after 2026-09-17) · methode: Holt damped exponential smoothing (phi=0.8, backtest alpha=0.1, one-step rolling-origin selection; 95% band = 1.96*sigma, 80% band = 1.2816*sigma, z-fallback per forecasting-method
- Q+1407,780%: 395,3 – 420,1
95%: 388,7 – 426,6 - Q+2471,180%: 453,6 – 488,7
95%: 444,3 – 497,9 - Q+3521,980%: 500,4 – 543,4
95%: 489,1 – 554,7
Outlook-scenario's 2027–2029
OUTLOOK, not a forecast. Capacity-anchored scenarios, 2027-2029. The statistical forecast envelope ends at Q+3 (Holt damped ETS, forecast/forecast.py); this layer is handed off at that point. planned_date tokens are coarser (2027-H2, 2028, year-ranges) where the primary source is itself coarse — no month is invented. (peil 2026-09-17).
Basisscenario
Discipline holds: suppliers keep the announced-but-staged buildouts and keep adding capex modestly while demand keeps CSP record levels — DRAM stays tight into 2028 and eases only as new fabs ramp (TrendForce: meaningful ramp only H2-2027, substantial output not before 2028).
2027 krap
- SK hynix Yongin Y1 fab initial cleanroom opens (release 2026-08-07: construction progressing toward February next year) 2027-02 · confirmed
- Micron Idaho ID1 first DRAM wafer output (company: 'second half of calendar 2027, with customer qualifications to follow') 2027-H2 · announced
- TrendForce: new 2027 capacity ramps meaningfully only in H2 2027 (construction/equipment/materials delay ramp-ups) 2027-H2 · announced
- SEMI 300mm fab equipment spending 2027 = $120B (+4%) — equipment spine (Oct-2025 edition; supersedes Mar-2024 edition's $137B 'record' for the same year) 2027 · announced
- WSTS 2027 consensus: global market +27% to ~USD 1.9T, Memory +32% (next editions May & November 2027) 2027-05 · announced
2028 versoepeling
- SK hynix Cheongju M17 NAND fab (KRW 19.1T) first cleanroom December 2028 2028-12 · announced
- Micron ID2 enters production before the first New York fab (exact quarter uncommitted) 2028-pre-NY · announced
- TrendForce: 'substantial output contributions are not expected to materialize until 2028' — the capacity lag resolves 2028 · announced
- SEMI 300mm fab equipment spending 2028 = $138B (+15%), peak of the published $374B 2026–2028 window — equipment spine 2028 · announced
- SEMI: advanced 1.4nm process enters high-volume production by 2028-2029 — node spine 2028-2029 · announced
2029 versoepeling
- SK hynix Yongin Y2 DRAM fab (KRW 35.2T, HBM/next-gen DRAM) opens cleanroom June 2029 2029-06 · announced
- SEMI 1.4nm HVM window includes 2029 (structural statement; no T1 numeric 2029 market figure exists) 2029 · announced
- Yongin cluster acceleration (completion pulled from 2045 to 2033) — 2029–2033 is where Korean greenfield supply arrives 2029-2033 · announced
Bull — versoepeling
Supply lands on time: every announced fab hits its cleanroom date AND demand normalises (CSP memory-share / negative-FCF demand-tipping trigger); NAND already follows this path — sufficiency ratio turns positive in 2027, looser environment from 2H27.
2027 versoepeling
- SK hynix Yongin Y1 fab initial cleanroom opens (release 2026-08-07: construction progressing toward February next year) 2027-02 · confirmed
- Micron Idaho ID1 first DRAM wafer output (company: 'second half of calendar 2027, with customer qualifications to follow') 2027-H2 · announced
- TrendForce: new 2027 capacity ramps meaningfully only in H2 2027 (construction/equipment/materials delay ramp-ups) 2027-H2 · announced
- SEMI 300mm fab equipment spending 2027 = $120B (+4%) — equipment spine (Oct-2025 edition; supersedes Mar-2024 edition's $137B 'record' for the same year) 2027 · announced
- WSTS 2027 consensus: global market +27% to ~USD 1.9T, Memory +32% (next editions May & November 2027) 2027-05 · announced
2028 loose
- SK hynix Cheongju M17 NAND fab (KRW 19.1T) first cleanroom December 2028 2028-12 · announced
- Micron ID2 enters production before the first New York fab (exact quarter uncommitted) 2028-pre-NY · announced
- TrendForce: 'substantial output contributions are not expected to materialize until 2028' — the capacity lag resolves 2028 · announced
- SEMI 300mm fab equipment spending 2028 = $138B (+15%), peak of the published $374B 2026–2028 window — equipment spine 2028 · announced
- SEMI: advanced 1.4nm process enters high-volume production by 2028-2029 — node spine 2028-2029 · announced
2029 loose
- SK hynix Yongin Y2 DRAM fab (KRW 35.2T, HBM/next-gen DRAM) opens cleanroom June 2029 2029-06 · announced
- SEMI 1.4nm HVM window includes 2029 (structural statement; no T1 numeric 2029 market figure exists) 2029 · announced
- Yongin cluster acceleration (completion pulled from 2045 to 2033) — 2029–2033 is where Korean greenfield supply arrives 2029-2033 · announced
Bear — verscherping
Allocation starves conventional DRAM: HBM keeps consuming wafer input faster than bits arrive (HBM bit shipments +50–60% YoY in 2027, still insufficient), buyers cut memory content, and supplier guidance (SK hynix: balance 'not before 2030') holds — tightness persists through the whole outlook.
2027 krap
- SK hynix Yongin Y1 fab initial cleanroom opens (release 2026-08-07: construction progressing toward February next year) 2027-02 · confirmed
- Micron Idaho ID1 first DRAM wafer output (company: 'second half of calendar 2027, with customer qualifications to follow') 2027-H2 · announced
- TrendForce: new 2027 capacity ramps meaningfully only in H2 2027 (construction/equipment/materials delay ramp-ups) 2027-H2 · announced
- SEMI 300mm fab equipment spending 2027 = $120B (+4%) — equipment spine (Oct-2025 edition; supersedes Mar-2024 edition's $137B 'record' for the same year) 2027 · announced
- WSTS 2027 consensus: global market +27% to ~USD 1.9T, Memory +32% (next editions May & November 2027) 2027-05 · announced
2028 krap
- SK hynix Cheongju M17 NAND fab (KRW 19.1T) first cleanroom December 2028 2028-12 · announced
- Micron ID2 enters production before the first New York fab (exact quarter uncommitted) 2028-pre-NY · announced
- TrendForce: 'substantial output contributions are not expected to materialize until 2028' — the capacity lag resolves 2028 · announced
- SEMI 300mm fab equipment spending 2028 = $138B (+15%), peak of the published $374B 2026–2028 window — equipment spine 2028 · announced
- SEMI: advanced 1.4nm process enters high-volume production by 2028-2029 — node spine 2028-2029 · announced
2029 krap
- SK hynix Yongin Y2 DRAM fab (KRW 35.2T, HBM/next-gen DRAM) opens cleanroom June 2029 2029-06 · announced
- SEMI 1.4nm HVM window includes 2029 (structural statement; no T1 numeric 2029 market figure exists) 2029 · announced
- Yongin cluster acceleration (completion pulled from 2045 to 2033) — 2029–2033 is where Korean greenfield supply arrives 2029-2033 · announced