PDF 原檔:報告_UBS_記憶體月報_20260807_original.pdf
圖片清單(已驗證 2026-08-10)
ingest 時建立的「眼見為憑」圖片索引,是 lib/ 嵌圖的唯一真相來源;嵌入時只從這裡挑分類為「真資料圖」的。<40KB 未 Read 者未列(預設 logo/裝飾)。
| 檔名 | size | 分類 | 親眼所見內容 |
|---|---|---|---|
報告_UBS_記憶體月報_20260807_001.png |
42KB | 真資料圖 | Figure 4:柱狀圖,HBM bit shipment(供給)vs bit end-consumption(需求,紅框標示 inventories 差額),分 2025/2026E/2027E 三組,需求柱逐年顯著高於供給柱 |
報告_UBS_記憶體月報_20260807_011.png |
55KB | 真資料圖 | Figure 15:DDR 價格循環走勢線圖(1Q16-4Q28E),標註各段漲跌百分比(-59%/+89%/-12%/+11%/+36%/-54%/+45%/-6%/+621%),末端標示未來上行段長度「13Qs」箭頭 |
報告_UBS_記憶體月報_20260807_014.png |
62KB | 真資料圖 | Figure 16:NAND 價格循環走勢線圖,同樣式(1Q16-4Q28E),標註各段漲跌百分比與「10Qs」箭頭 |
報告_UBS_記憶體月報_20260807_017.png |
41KB | 真資料圖 | Figure 21:柱狀+折線圖,Memory Cost($)與 Memory as % of BOM(智慧型手機 flagship/premium),1Q22-4Q27E,折線由約 20% 升至約 35% |
報告_UBS_記憶體月報_20260807_018.png |
46KB | 真資料圖 | Figure 23:三星/SK 海力士記憶體成品晶圓與晶粒庫存週數線圖(4Q17-2Q26),紅色虛線標示「Normal level」約 4 週,右側標註「1Q26: 3.5」「2Q26: 2.5」 |
報告_UBS_記憶體月報_20260807_021.png |
50KB | 真資料圖 | Figure 24:三星/SK 海力士記憶體成品晶圓與晶粒庫存週數線圖,樣式與內容標註(含 1Q26: 3.5、2Q26: 2.5)與 018 相同,無法從圖面本身確認與 018 的 DRAM/NAND 分項差異 |
報告_UBS_記憶體月報_20260807_022.png |
45KB | 真資料圖 | Figure 25:DRAM 客戶端庫存週數折線圖,分 Server(Hyperscalers)/Smartphone OEMs/PCs 三條線(1Q22-2Q26),近期落在約 7-10 週區間 |
報告_UBS_記憶體月報_20260807_023.png |
48KB | 真資料圖 | Figure 27:DRAM 產業晶圓投片量(k wpm,柱狀)與 YoY 成長率(折線),1Q14-3Q27E,柱狀由約 1,000 升至逾 2,000 |
報告_UBS_記憶體月報_20260807_026.png |
52KB | 真資料圖 | Figure 28:NAND 產業晶圓投片量(k wpm,柱狀)與 YoY 成長率(折線),1Q14-3Q27E,柱狀約 1,000-1,700 區間震盪 |
報告_UBS_記憶體月報_20260807_030.png |
88KB | 真資料圖 | 個股目標價/股價/評等歷史圖,貨幣為 US$,2023-06 至 2026-08,Price Target 由約 80 逐步階梯上調至約 1,600,Stock Price 同步走高至約 800-1,200,下方評等軸全程 Buy;圖面未標公司名,依貨幣(US$)與本報告涵蓋券商評等清單(Samsung/CXMT/Kioxia/MU/Nanya Tech/SK Hynix)推斷極可能為 MU.US(micron)(美光為報告內唯一美元計價個股) |
報告_UBS_記憶體月報_20260807_031.png |
100KB | 真資料圖 | 個股目標價/股價/評等歷史圖,貨幣為 NT$,2023-06 至 2026-08,Price Target 由約 90 階梯上調至約 540-580,Stock Price 同步走高至約 400-480;下方評等軸顯示 No Rating(2023 上半)→ Buy(2023 底起)→ 短暫 Neutral(2025 底)→ 恢復 Buy;圖面未標公司名,依貨幣(NT$)推斷極可能為 2408_南亞科技(市)(報告內唯一新台幣計價個股,價格區間與 vault 既有 TP/股價紀錄一致) |
報告_UBS_記憶體月報_20260807_032.png |
98KB | 真資料圖 | 個股目標價/股價/評等歷史圖,貨幣為 Won,2023-06 至 2026-08,Price Target 階梯上調至約 540,000,Stock Price 同步走高至約 240,000-350,000;評等軸顯示 Buy 為主、中間有一段 Neutral(約 2025Q3-Q4);圖面未標公司名,為報告內兩張韓元計價個股圖之一,無法從圖面本身確認是三星或 SK 海力士 |
報告_UBS_記憶體月報_20260807_033.png |
92KB | 真資料圖 | 個股目標價/股價/評等歷史圖,貨幣為 Won,2023-06 至 2026-08,Price Target 階梯上調至約 3,200,000,Stock Price 同步走高至約 1,300,000-2,800,000;評等軸全程 Buy 無中斷;圖面未標公司名,為報告內兩張韓元計價個股圖之一,無法從圖面本身確認是三星或 SK 海力士(連續無中斷的 Buy 軸與文字段落「Samsung Key Call Buy」語氣較吻合,但不足以視為確認) |
報告_UBS_記憶體月報_20260807_034.png |
84KB | 真資料圖 | 個股目標價/股價/評等歷史圖,貨幣為 ¥(日圓),股價線自 2024-12 附近才開始出現(此前為空白),階梯上升至約 140,000;評等軸顯示 No Rating(大部分期間)→ 僅圖表末段(約 2026Q3)轉為 Buy;圖面未標公司名,依貨幣(¥)與「股價線 2024 年底才出現+近期才轉 Buy」的模式,推斷極可能為 285A.JP(kioxia)(鎧俠 2024-12 於東京證交所掛牌,報告內唯一日圓計價個股) |
原始內容
Memory Semis Monthly
August '26 Edition: Addressing investors' three key concerns
HBM de-specing for Rubin Ultra but overall demand holding up
Our industry checks suggest Nvidia will likely initially use HBM4 8-Hi for Rubin Ultra followed by a switch to HBM4E 8-Hi. This is compared to our prior expectation of Rubin Ultra using HBM4E 12-Hi. We believe this is in response to persistent tight supply. This implies that the initial version of the VR300 will use 384GB of HBM4 per GPU (8-Hi x 24Gb x 8 x2), then moving to 512GB HBM4E (8-Hi x 32Gb x8 x2) vs. our 768GB prior estimate. Despite this, our forecast for '27E Nvidia HBM end-consumption increases slightly to 25.1bn Gb (was 24.0bn) as GPU units implied go up to 12.6mil from 11.0mil. We also increase our HBM forecast for Google TPUs to 19.2bn Gb in '27 (was 17.4bn) on unit upside. All in all, our HBM industry end-consumption estimate moves to 61.5bn Gb in 2027 (was 58.7bn) or up 91% YoY. On pricing, we now estimate HBM4E could reach US$3.9 per Gb in '27 (+75% YoY vs HBM4 in '26); HBM4 US$3.5 (+60% YoY). We forecast industry HBM blended ASPs up 79% YoY (was 66%) in '27. We see further upside to server DDR demand and hence forecast DRAM industry bit end-demand up 40% YoY in '27 (was +36%) after +21% in '26.
NAND flash contract pricing more dispersed but positive trend remains
Our checks on NAND flash (ex LTA) contract pricing negotiations suggest a wider range of QoQ NAND flash contract price increases for 3Q26E from +20% to >30% QoQ, still ahead of some of the recent market expectations of high teens up. On average, we now forecast ex LTA NAND ASPs up +23% QoQ (was +30%), +12% in 4Q (u/c), and with LTAs, closer to +20%/+8% QoQ respectively. Demand weakness for smartphones (22% of total '26E industry bit demand) and client/PC SSD (10% of total) continues. On the other hand, we see further strength in server/storage SSD demand (43% of total) and now forecast 62% YoY bit demand growth combined in '26 (was 54%) and 51% in '27 (was 45%). All in all, we forecast total NAND industry bit end demand up 23% YoY in '26 and 26% in '27 (was 20/26%) and total NAND industry revs US$301bn/492bn, close enough to SNDK's forecast (US$300bn/500bn).
CXMT not a threat to overall DRAM supply / demand
We initiate coverage of CXMT with a Buy rating (link). We forecast CXMT to reach 9% bit supply share in '27 vs. 7% in '25, with wafer capacity moving to 382k wpm by 4Q27 from 292k in 4Q26 and 240k wpm in 4Q25. We note that some of the market expectations are probably more aligned with equipment delivery vs actual installation/ validation and actual production capacity - and hence too high. CXMT is moving to G5 process tech late '26 (sub 15 nm) vs. Samsung / SK Hynix / MU moving to 1d nm (sub 11 nm) in 2H27. Furthermore, in part due to manufacturing equipment exports control, production yields for CXMT are not at par with the top 3 industry leaders.
Buy the pullback on a more sustainable cycle: Prefer DRAM, Samsung top pick
On average memory stocks have pulled back 36% from late June peak and now discount pre-AI through-cycle ROEs (eg. Samsung: 13.9% vs '12-22 average 15.2%). The main fundamental risk is memory spending sustainability for customers, notably hyperscalers. LTAs are reigning in some of the pricing upside into the upcycle and hence, we bring down our '27E memory industry revs forecasts to US$1.61tn ('26E: US $0.94tn), or 9% below prior. LTAs are OPM/ROE accretive through the cycle (link) and in essence, for memory FCF and ultimately, shareholder returns. We continue to forecast the DRAM upcycle to continue for longer than NAND (2Q28E vs. 4Q27E). We rate Samsung Key Call Buy (link); CXMT, Kioxia, MU, Nanya Tech and SK Hynix Buy.
Equities
Global
Semiconductors
Nicolas Gaudois
Analyst nicolas.gaudois@ubs.com +65-6495 5148
Timothy Arcuri
Analyst timothy.arcuri@ubs.com +1-415-352 5676
Kenji Yasui
Analyst kenji.yasui@ubs.com +81-3-5208 6211
Jimmy Yu
Analyst
S1460517080002 jimmy.yu@ubs.com +86-21-3866 8880
Jimmy Yoon
Analyst jimmy.yoon@ubs.com +65-6495 4617
Luke Yoo
Analyst luke.yoo@ubs.com +82-2-3702 8132
Gianmarco Vella
Associate Analyst gianmarco.vella@ubs.com +1-415-352 4555
Atsuhiro Kinoshita
Analyst atsuhiro.kinoshita@ubs.com +81-3-5208 6768
Sunny Lin
Analyst sunny.lin@ubs.com +886-2-8722 7346
Francois-Xavier Bouvignies
Analyst francois.bouvignies@ubs.com +44-20-7568 7105
Randy Abrams
Analyst randy.abrams@ubs.com +886-2-8722 7338
KEY Charts
Figure 1: AI server GPU/accelerator unit forecasts ('000 units)
| HBM Gen. | HBM | Content | (GB) | 2025 | 1Q26 | 2Q26 | 3Q26E | 4Q26E | 2026E | 1Q27E | 2Q27E | 3Q27E | 4Q27E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NVIDIA L40S | GDDR6 | 2025E 48 | 2026E | 2027E | 38 | ||||||||||||||||
| H200 H20 / H20E | HBM3E HBM3 / 3E | 141 96 | 593 441 | ||||||||||||||||||
| B200 (8Hi) GB200 (2x GPU - 8Hi) | HBM3E | 192 | 384 | 1,441 | 103 | 103 | |||||||||||||||
| B300 (12Hi) | HBM3E HBM3E | 384 288 | 288 | 288 | 1,080 587 | 267 | 300 | 200 | 100 | 867 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | |||
| GB300 (2x GPU - 12Hi) R200 (12Hi) | HBM3E HBM4 | 576 | 576 | 576 | 778 | 658 | 670 20 | 580 | 450 320 | 2,358 540 | 360 360 | 270 400 | 180 320 | 90 240 | 900 1,320 | 900 1,320 | 900 1,320 | 900 1,320 | 900 1,320 | 900 1,320 | 900 1,320 |
| RV200 (2x GPU - 12Hi) | HBM4 | 288 576 | 288 576 | 40 | 200 250 | 450 | 740 | 600 | 720 | 690 | 630 | 2,640 | 2,640 | 2,640 | 2,640 | 2,640 | 2,640 | 2,640 | |||
| R300 (8Hi) R300 (8Hi) | HBM4 HBM4E | 192 | 250 400 | 400 | 400 | 400 | 400 | 400 | 400 | 400 | |||||||||||
| RV300 (2x GPU - 8Hi) | HBM4 HBM4E | 256 384 512 | 250 | 250 | 250 | 250 | 250 | 550 | 550 | 250 | |||||||||||
| RV300 (2x GPU - 8Hi) *Other GPU/Accel | HBM3 / 3E | 288 | 384 | 576 | 107 | 60 | 80 | 80 | 80 1,400 | 300 4,908 | 70 1,440 | 70 1,460 | 80 1,770 | 80 1,990 | 300 6,660 | 300 6,660 | 300 6,660 | 300 6,660 | 300 6,660 | 300 6,660 | 300 6,660 |
| Total Units %YoY | 5,066 3% | 1,088 -17% | 1,110 0% | 1,310 3% | 3% | -3% | 32% | 32% | 35% | 42% | 36% | 36% | 36% | 36% | 36% | 36% | 36% | ||||
| %QoQ Total SiPs | 6,924 | -20% 1,849 | 2% 1,820 | 18% 2,140 | 7% 2,300 | 8,109 | 3% 2,400 | 1% 2,450 | 21% 2,890 | 12% 3,260 | 11,000 | 11,000 | 11,000 | 11,000 | 11,000 | 11,000 | 11,000 | ||||
| %YoY | 38% | 27% | 23% | 15% | 8% | 17% | 30% | 35% | 35% | 42% | 36% | 36% | 36% | 36% | 36% | 36% | 36% | ||||
| %QoQ Total GPUs | -13% | -2% | 18% | 7% 2,300 | 8,109 | 4% | 2% | 18% | 13% 4,360 | ||||||||||||
| %YoY | 6,924 38% | 1,849 | 1,820 | 2,140 | 8% | 17% | 2,400 | 2,450 | 12,600 55% | 12,600 55% | 12,600 55% | 12,600 55% | 12,600 55% | 12,600 55% | 12,600 55% | ||||||
| 15% | |||||||||||||||||||||
| %QoQ | 38% | ||||||||||||||||||||
| AMD | 4% | 2% | |||||||||||||||||||
| 27% | 58% | 90% | |||||||||||||||||||
| 23% | 30% | 35% | 3,390 | ||||||||||||||||||
| MI300 MI325X (12Hi) MI355X (12Hi) MI308X MI450 (12Hi) | HBM3 HBM3E HBM3E HBM3 HBM4 | 192 256 288 192 | 256 288 | 432 | 71 224 221 96 | -13% 18 170 29 | -2% 150 28 | 18% 80 5 45 | 7% 70 230 | 18 470 62 275 | 350 2 | 450 | 29% 470 | 1,650 | 1,650 | 1,650 | 1,650 | 1,650 | 1,650 | 1,650 | |
| MI5XX (16Hi) *Other GPU/Accel Total Units | HBM4E HBM3 | 192 | 192 432 192 | 1024 288 | 4 616 | 1 218 | 1 179 | 1 131 | 1 301 | 6 | 352 | 2 452 | 2 532 | 60 2 542 | 8 | 8 | 8 | 8 | 8 | 8 | 8 |
| %YoY | 831 | 1,878 | 1,878 | 1,878 | 1,878 | 1,878 | 1,878 | 1,878 | |||||||||||||
| %QoQ | 86% | -26% | 37% | 35% | 61% | 80% | 126% | 126% | 126% | 126% | 126% | 126% | 126% | ||||||||
| 41% | -18% | 129% | 152% | 305% | |||||||||||||||||
| Google TPU v5e | 81% -1% | 224 720 | -27% | 17% | 28% | 18% | 2% | ||||||||||||||
| TPU v5p TPU v6e TPU v7 TPU v8i (Dual die AX) | HBM2E HBM2E HBM3E | 32 96 32 | 32 192 288 | 192 288 | 134 1,916 426 90 | 205 520 | 141 800 25 | 70 750 225 | 640 2,790 250 | 600 700 | 450 | 1,500 | 300 1,000 1,000 | 4,400 | 4,400 | 4,400 | 4,400 | 4,400 | 4,400 | 4,400 | |
| TPU v8t (Single *Other Total Units | HBM3E HBM3E | 192 | 216 144 | 216 144 | 1,046 | 37 | 25 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | |||||||
| die X) TPU/Accel %YoY | HBM3E | 96 | 77 | 7 | 9 | 70 10 | 380 10 | 450 | 700 | 1,200 | 25 | 25 | |||||||||
| HBM3 | 2,643 | 1,435 | 25 | 3,125 | |||||||||||||||||
| 19% | 732 | 953 66% | 82% | 68% | 4,167 58% | 2,025 177% | 2,675 | 199% | 2,175 52% | 10,000 140% | 10,000 140% | 10,000 140% | 17% | -30% | 10,000 140% | 10,000 140% | |||||
| %QoQ Amazon Inferentia 2 | 14% -14% | 30% | 10% | 37% | 590 96 1,780 | 41% 50 | 181% 32% | 10 | 7 | ||||||||||||
| Trainium 2 / 2.5 Inferentia 3 | HBM2E HBM3E | 32 96 | 144 | 144 96 | 88 496 | 130 30 | 220 28 320 | 140 20 | 100 18 | 28 | 15 | 10 7 | 280 30 | 280 30 | 280 30 | 280 30 | 280 30 | 280 30 | 280 30 | ||
| Trainium 3 Trainium 4 | HBM3 HBM3E | 96 | 144 256 | 162 | 620 | 840 | 840 | 780 | 640 | 540 240 | 2,800 | 2,800 | 2,800 | 2,800 | 2,800 | 2,800 | 2,800 | ||||
| *Other Accel | HBM4 | 96 144 | 40 | ||||||||||||||||||
| HBM3 | 144 | 144 | 144 | 37 783 | 5 | 6 | 8 | 10 | 8 | 8 | |||||||||||
| Total Units %YoY | 165 | 574 | 788 | 968 | 2,494 218% | 913 454% | 808 41% | 687 | 787 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | |||||
| %QoQ Intel / Habana | -8% | -17% -11% | 180% 248% | 302% 37% | 424% 23% | -6% | -12% | -13% -15% | -19% 15% | 28% | 28% | 28% | 28% | 28% | 28% | 28% | |||||
| Gaudi 3 Gaudi 4 | HBM2 / 3E HBM3E | 288 | 288 288 | 288 | 188 | 30 | 18 | 8 10 | 25 | 56 35 | 25 | 20 | 15 | 10 | 70 | 70 | 70 | 70 | 70 | 70 | 70 |
| *Others Total Units | HBM2 / 3E | 192 | 192 | 192 | 25 213 | 6 36 | 4 22 | 4 | 5 | 18 | 4 | 4 | 4 | 16 86 | 16 86 | 16 86 | 16 86 | 16 86 | 16 86 | 16 86 | |
| %YoY %QoQ | -32% | 22 | 30 -46% | 109 -49% | 29 -19% | 24 11% | 19 -12% -21% | 4 14 -53% -26% | -21% | -21% | -21% | -21% | -21% | -21% | -21% | ||||||
| Microsoft Maia 100 | HBM3 | -20% -36% 3 | -60% -40% 1 | -63% 0% | 39% | 4 | -3% | -17% | 35 | 35 | 35 | 35 | 35 | 35 | 35 | ||||||
| Maia 200 Maia 300 | HBM3E HBM4 | 120 | 120 288 | 288 | 34 | 2 | 6 2 | 10 | 10 | 7 8 2 | 4 14 | 4 14 | 4 14 | 4 14 | 4 14 | 4 14 | 4 14 | ||||
| *Others | HBM3 | 120 | 192 | 288 192 | 8 42 | 9 | 14 | 33 | 14 | ||||||||||||
| Total Units %YoY | 2 7 | 2 12 | 2 16 | 8 45 7% | 2 16 | 2 | 2 20 | 22 65 | 22 65 | 22 65 | 22 65 | 22 65 | 22 65 | 22 65 | |||||||
| %QoQ Meta | 22% | 22% | 4% 40% | 84% 29% | 119% 0% | 12 | 17 36% 42% | 24% 18% | |||||||||||||
| -27% -17% | -19% | 35% -25% | 45% | 45% | 45% | 45% | |||||||||||||||
| MTIA v1 | LPDDR5 | 64 | 20 | 5 | 2 | 25 | |||||||||||||||
| MTIA v2 MTIA v3 MTIA v4 | LPDDR5 HBM3E | 128 | 64 128 | 421 30 | 20 15 | 8 20 | 35 | 30 110 | 40 | 30 | 20 | 15 120 | 40 | 105 340 | 105 340 | ||||||
| HBM4 | 192 64 | 192 | 192 432 | 5 | 6 | 40 5 | 80 5 | 100 | 20 | 20 | 20 | 20 | 20 | 20 | |||||||
| *Others Total Units | LPDDR5 | 128 | 128 | 28 484 | 3 58 | 1 34 | 1 38 | 1 41 | 171 -65% | 85 47% | 115 235% | 5 125 228% 9% | 5 140 | 35% | 240% 12% | 465 171% | |||||
| %YoY %QoQ Tesla | 19% | -43% | -71% | -71% 11% | -68% 8% | 106% | |||||||||||||||
| Dojo | 64 | 62 | -56% 17 | -41% 18 | 15 | 14 | 15 | 15 | 18 | 15 5 | 20 | 63 22 | 63 22 | ||||||||
| *Others Total Units %YoY | HBM3 HBM3 | 64 | 64 64 | 64 64 | 44 106 | 7 24 | 7 25 -9% 6% | 6 21 -24% | 6 20 | 64 26 90 | 6 21 -12% | -17% | 10% | 5 23 | 2% | 6 | 85 | 85 | |||
| %QoQ Others | -39% | -12% 2% | -17% | -16% -7% | -15% | 7% | 21 0% | 10% | -5% | ||||||||||||
| 170 | -13% | 725 254 | 725 254 | 725 254 | 725 254 | 725 254 | |||||||||||||||
| Startups Others Total Units | HBM3 HBM2E /3 | 192 192 | 288 288 | 288 288 | 657 361 | 150 30 | 150 30 180 | 155 39 194 | 155 39 194 | 610 180 138 63 | 60 | 243 | 185 65 250 | 67 | 190 257 | ||||||
| %YoY %QoQ | 1,019 59% | 180 -28% -26% | -32% 0% | -26% 8% | -20% 0% | 748 -27% | 230 28% 18% | 35% 6% | 29% 3% | 32% 3% | 979 31% | 979 31% | 979 31% | 979 31% | 979 31% | ||||||
| Total GPU / Accelerator | Units | ('000) | ('000) | ('000) | 12,829 | 3,269 | 3,797 | 4,412 | 5,455 | 16,933 | 6,231 | ('000) | ('000) | ('000) | ('000) | 6,970 | ('000) | 7,848 | 7,445 | 28,493 | ('000) |
Source: Company data, UBS estimates
Figure 2: HBM demand forecasts by GPU/accelerator
| HBM Gen. | HBM | Content | (GB) | 2025 | 1Q26 | 2Q26 | 3Q26E | 4Q26E | 2026E | 1Q27E | 2Q27E | 3Q27E | 4Q27E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NVIDIA H200 | 2025E | 2026E | 2027E | ||||||||||||||||||||
| H20 / H20E | HBM3E HBM3 / 3E | 141 96 | 83,571 42,336 | ||||||||||||||||||||
| B200 (8Hi) GB200 (2x GPU - | HBM3E | 192 | 384 | 276,649 | |||||||||||||||||||
| 8Hi) | HBM3E | 414,872 | |||||||||||||||||||||
| 384 | 39,504 | 39,504 | |||||||||||||||||||||
| B300 (12Hi) | HBM3E | 288 | 288 | 288 | 169,176 | 76,896 | 86,400 | 57,600 | 28,800 | 249,696 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | 14,400 | |||
| GB300 (2x GPU - 12Hi) | HBM3E | 576 | 576 | 576 | 448,259 | 379,008 | 385,920 | 334,080 | 259,200 | 1,358,208 | 207,360 | 155,520 | 103,680 | 51,840 | 518,400 | 518,400 | 518,400 | 518,400 | 518,400 | 518,400 | 518,400 | 518,400 | 518,400 |
| R200 (12Hi) | HBM4 | 288 | 288 | 5,760 | 57,600 | 92,160 | 155,520 | 103,680 | 115,200 | 92,160 | 69,120 | 380,160 | 380,160 | 380,160 | 380,160 | 380,160 | 380,160 | 380,160 | 380,160 | 380,160 | |||
| RV200 (2x GPU - 12Hi) | HBM4 | 576 | 576 | 23,040 | 144,000 | 259,200 | 426,240 | 345,600 | 414,720 | 397,440 | 362,880 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | 1,520,640 | |||
| R300 (12Hi) | HBM4 | 192 | 48,000 | 102,400 | 48,000 | 48,000 | 48,000 | 48,000 | 48,000 | 48,000 | 48,000 | 48,000 | 48,000 | ||||||||||
| R300 (12Hi) RV300 (4x GPU - 12Hi) | HBM4E | 256 | 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | 102,400 96,000 | |||||||||||
| RV300 (4x GPU - 12Hi) | HBM4 HBM4E | 384 512 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | 281,600 | |||||||||||
| *Other GPU/Accel | HBM3 / 3E | 288 | 384 | 576 | 30,787 | 23,040 | 30,720 | 30,720 | 30,720 | 115,200 | 40,320 | 40,320 | 46,080 | 46,080 | 172,800 | 172,800 | 172,800 | 172,800 | 172,800 | 172,800 | 172,800 | 172,800 | 172,800 |
| Total ('000 GB) | 1,465,650 | 531,840 | 624,000 | 2,344,368 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | 3,134,400 | ||||||||||
| Total (bn Gb) | 11.73 | 518,448 4.15 | 4.25 | 4.99 | 670,080 5.36 | 18.75 | 711,360 5.69 | 725,760 5.81 | 783,360 | 6.27 | 913,920 7.31 | 913,920 7.31 | 913,920 7.31 | 913,920 7.31 | 913,920 7.31 | 913,920 7.31 | 913,920 7.31 | 25.08 | 913,920 7.31 | ||||
| AMD | |||||||||||||||||||||||
| MI300 | HBM3 | 192 | 13,685 | ||||||||||||||||||||
| MI325X (12Hi) | HBM3E | 256 | 256 | 57,294 | 4,608 | 4,608 | |||||||||||||||||
| MI355X (12Hi) | 135,360 | ||||||||||||||||||||||
| HBM3E | 288 | 288 | 63,569 | 48,960 | 43,200 | 23,040 | 20,160 | 11,904 | |||||||||||||||
| MI308X | HBM3 | 192 | 192 432 | 18,394 | 5,568 | 5,376 | 960 | 712,800 | |||||||||||||||
| MI450 (12Hi) MI5XX (16Hi) | HBM4 | 432 | 19,440 | 99,360 | 118,800 | 151,200 | 194,400 | 203,040 61,440 | 164,160 | 164,160 | 164,160 | 164,160 | 164,160 | 164,160 | 164,160 | 164,160 | 164,160 | ||||||
| *Other GPU/Accel | HBM4E | 812 | 274 | 285 | 163,840 | 225,280 | 225,280 | 225,280 | 225,280 | 225,280 | 225,280 | 225,280 | 225,280 | 225,280 | |||||||||
| Total ('000 GB) | HBM3 | 192 | 192 | 1,024 288 | 282 | 281 | 1,123 | 576 | 576 | 576 | 576 | 2,304 | 2,304 | 2,304 | 2,304 | 2,304 | 2,304 | 2,304 | 2,304 | 2,304 | |||
| 153,754 | 59,410 | 48,861 | 43,722 | 119,801 | 271,795 | 151,776 | 194,976 | 265,056 | 328,576 | 940,384 | 940,384 | 940,384 | 940,384 | 940,384 | 940,384 | 940,384 | 940,384 | 940,384 | |||||
| Total (bn Gb) | 7.52 | 7.52 | 7.52 | 7.52 | 7.52 | 7.52 | 7.52 | 7.52 | 7.52 | ||||||||||||||
| 1.23 | 0.48 | 0.39 | 0.35 | 0.96 | 2.17 | 1.21 | 1.56 | 2.12 | 2.63 | 2.63 | 2.63 | 2.63 | 2.63 | 2.63 | 2.63 | 2.63 | 2.63 | ||||||
| TPU v5e | HBM2E | 4,282 | |||||||||||||||||||||
| TPU v5p TPU v6e | HBM2E HBM3E | 32 96 32 | 32 | 183,913 | 2,253 | 20,480 | |||||||||||||||||
| TPU v7 | HBM3E | 192 | 192 | 192 | 13,632 17,280 | 6,554 99,840 | 7,168 138,240 | 4,506 153,600 | 144,000 | 535,680 | 115,200 | 86,400 | 57,600 | 28,800 | 288,000 | 288,000 | 288,000 | 288,000 | 288,000 | 288,000 | 288,000 | 288,000 | 288,000 |
| TPU v8i (Dual die AX) | HBM3E | 7,200 | 64,800 | 288,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | 1,152,000 | ||||||||||
| TPU v8t (Single die X) | HBM3E | 288 | 288 | 15,120 | 82,080 | 72,000 97,200 | 201,600 | 288,000 | 374,400 | ||||||||||||||
| *Other TPU/Accel | HBM3 | 216 144 | 216 144 | 1,359 | 1,492 | 1,440 | 5,335 | 151,200 3,600 | 259,200 3,600 | 324,000 | 3,600 | 216,000 | 950,400 14,400 | 950,400 14,400 | 950,400 14,400 | 950,400 14,400 | 950,400 14,400 | 950,400 14,400 | 2,404,800 | 950,400 14,400 | |||
| Total ('000 GB) Total (bn Gb) Amazon Inferentia 2 | HBM2E | 96 32 | 7,389 226,495 | 1,044 107,437 0.86 | 146,767 1.17 | 181,917 1.46 | 294,573 2.36 | 730,695 5.85 3.77 | 471,600 | 637,200 5.10 | 3,600 759,600 6.08 | 536,400 4.29 | 77,760 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | 403,200 71,680 | ||
| Trainium 2 / 2.5 Inferentia 3 Trainium 3 Trainium 4 | HBM3E HBM3 HBM3E HBM4 | 96 96 | 144 96 144 | 144 96 144 | 1.81 2,823 47,574 15,563 | 18,720 2,880 | 31,680 2,688 46,080 | 20,160 1,920 89,280 1,123 | 14,400 1,728 120,960 1,380 | 84,960 7,200 9,216 1,440 256,320 120,960 | 112,320 | 1,440 10,240 | 92,160 | 61,440 | 960 4,320 | 960 4,320 | 960 4,320 | 960 4,320 | 960 4,320 | 960 4,320 | 960 4,320 | ||
| *Other Accel | HBM3 | 144 | 691 | 818 | 1,152 | 1,152 | 1,008 | ||||||||||||||||
| 144 | 256 144 | 5,375 | 4,012 | 1,008 | |||||||||||||||||||
| Total ('000 GB) | 81,266 | 103,408 | |||||||||||||||||||||
| Total (bn Gb) | 71,336 | 22,291 | 112,483 | 138,468 | 354,508 | 130,752 | 140,208 | 490,240 | 490,240 | 490,240 | 490,240 | 490,240 | 490,240 | 490,240 | |||||||||
| Intel | 0.57 | 0.18 | 0.90 | 1.11 | 2.84 | 1.05 | 115,872 0.93 | 0.83 | 1.12 | 1.12 | 1.12 | 1.12 | 1.12 | 1.12 | 1.12 | 3.92 | 1.12 | ||||||
| Gaudi 3 Gaudi 4 *Others | HBM2 / 3E HBM3E HBM2 / 3E | 288 | 288 288 | 288 | 54,212 4,789 | 8,640 | 0.65 5,184 | 2,304 2,880 691 | 7,200 960 8,160 | 16,128 10,080 7,200 3,494 | 768 7,968 | 5,760 | 2,880 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | 20,160 3,072 | ||
| Total ('000 GB) Total (bn Gb) | 192 | 192 | 192 | 59,001 | 1,152 | 691 | 5,875 | 768 | 4,320 | 768 0.03 | |||||||||||||
| 9,792 | 5,875 | 0.05 | 0.07 | 29,702 0.24 | 0.06 | 768 | 3,648 | 23,232 | 23,232 | 23,232 | 23,232 | 23,232 | 23,232 | 0.19 | 23,232 | ||||||||
| Microsoft Maia 100 Maia 200 | HBM3 HBM3E | 120 | 120 288 | 288 | 0.47 4,050 | 0.08 360 660 | 0.05 120 1,696 | 3,016 | 4,052 | 480 9,424 | 4,032 | 2,880 | 6,528 0.05 | 5,088 | 0.04 | 1,152 4,032 | 384 | 0.14 | 0.14 | 0.14 | 0.14 | 0.14 | 0.14 |
| *Others Total ('000 GB) | HBM4 HBM3 | 120 | 5,010 | 384 | 384 | 384 | 1,536 11,440 | 384 | 384 3,264 | 2,016 384 4,704 0.04 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | 6,336 1,536 17,952 | ||||||
| 192 | 0.04 | 384 | 2,304 | ||||||||||||||||||||
| Maia 300 | 192 | 288 192 | 960 | 3,400 | 5,568 | ||||||||||||||||||
| 0.03 | 4,436 | 4,416 | |||||||||||||||||||||
| Total (bn Gb) | 1,404 0.01 | 2,200 0.02 | 0.04 | 0.09 | 0.04 | 0.03 | 0.04 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | 20,160 146,880 167,040 | ||||||||
| Meta MTIA v3 MTIA v4 | HBM3E HBM4 | 192 | 960 | 2,880 2,880 | 3,840 3,840 | 6,720 | 7,680 | 21,120 | 5,760 34,560 | 3,840 43,200 | 7,680 17,280 | 2,880 51,840 54,720 | |||||||||||
| Total ('000 GB) Total (bn Gb) | 192 432 | 960 0.01 | 0.02 1,523 | 0.03 1,613 | 6,720 0.05 1,344 | 7,680 0.06 1,255 | 21,120 0.17 | 24,960 0.20 | 0.32 1,344 | 47,040 0.38 | 0.44 | 40,320 | 1,472 | 1.34 | 1.34 | 1.34 | |||||||
| Others Tesla Startups / Etc. Total ('000 GB) | HBM3 HBM2E / 3 | 64 192 | 288 | 51,840 53,453 | 55,800 57,144 | 5,735 215,280 | 71,928 73,400 | 1,344 66,096 67,440 | 1,280 | 0.60 | 73,872 75,152 | ||||||||||||
| 64 | 6,780 | 5,440 281,880 | 5,440 281,880 | 5,440 281,880 | 5,440 281,880 | 5,440 281,880 | |||||||||||||||||
| Total (bn Gb) | 195,559 202,339 | 51,840 | 55,800 | 69,984 71,328 | 287,320 | 287,320 | 287,320 | 287,320 | 287,320 | 287,320 | 287,320 | 287,320 | |||||||||||
| 0.43 | 0.46 | 57,055 0.46 | 221,015 1.77 | 0.54 | 0.57 | 0.59 | 2.30 | 2.30 | 2.30 | 2.30 | 2.30 | 2.30 | 2.30 | 2.30 | 2.30 | ||||||||
| 1.62 | 53,363 0.43 | 874,103 6.99 | 1,332,654 10.66 77% | 4,021,364 32.17 | 1,604,832 12.84 | 1,831,968 14.66 | 2,098,536 16.79 102% | 2,154,672 17.24 | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | 7,690,008 61.52 91% | |||||
| Total HBM Demand Total HBM Demand %YoY | ('000 GB) (bn Gb) | 2,184,545 17.48 103% | 775,025 6.20 97% | 103% | 8.32 72% | 84% 107% | 1,039,582 | 110% | 62% |
Source: Company data, UBS estimates
Figure 3: HBM vendors' bit shipments
| (M 1Gb) | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SK Hynix | 12,628 | 3,803 | 3,804 | 4,559 | 5,072 | 17,237 | 4,915 | 5,118 | 5,466 | 5,796 | 21,295 | |
| %YoY | 86% | 31% | 20% | 41% | 53% | 36% | 29% | 35% | 20% | 14% | 24% | |
| %QoQ | 15% | 0% | 20% | 11% | -3% | 4% | 7% | 6% | ||||
| Samsung | 4,327 | 1,519 | 2,061 | 3,384 | 4,436 | 11,400 | 4,879 | 5,266 | 5,801 | 6,242 | 22,189 | |
| %YoY | -15% | 161% | 170% | 140% | 182% | 163% | 221% | 156% | 71% | 41% | 95% | |
| %QoQ | -3% | 36% | 64% | 31% | 10% | 8% | 10% | 8% | ||||
| Micron | 4,584 | 1,612 | 1,821 | 2,040 | 2,305 | 7,778 | 2,536 | 2,840 | 3,152 | 3,436 | 11,964 | |
| %YoY | 417% | 106% | 60% | 62% | 64% | 70% | 57% | 56% | 55% | 49% | 54% | |
| %QoQ | 15% | 13% | 12% | 13% | 10% | 12% | 11% | 9% | ||||
| Total Supply | 21,539 | 6,934 | 7,686 | 9,983 | 11,812 | 36,415 | 12,330 | 13,223 | 14,419 | 15,475 | 55,447 | |
| %YoY | 69% | 62% | 52% | 69% | 88% | 69% | 78% | 72% | 44% | 31% | 52% | |
| %QoQ | 10% | 11% | 30% | 18% | 4% | 7% | 9% | 7% |
Source: Company data, UBS estimates
Figure 4: Adjusted HBM supply vs demand

Source: Company data, UBS estimates
Figure 5: HBM as % of total DRAM revenues

Source: UBS estimates
Figure 6: HBM as % of total DRAM bits

Source: UBS estimates
Figure 7: HBM ASP forecasts (US$/1Gb)

Source: UBS estimates
Figure 9: DRAM industry capacity (k wpm) breakdown between DDR and HBM

Source: Company data, UBS estimates
Figure 11: HBM demand by type (2026E)

Source: UBS estimates
Figure 8: HBM bit share by vendor

Source: Company data, UBS estimates
Figure 10: DRAM capacity (k wpm) breakdown between DDR and HBM by supplier

Source: Company data, UBS estimates
Figure 12: HBM demand by type (2027E)

Source: UBS estimates
052$
00'2$
OS I$
00'1$
OS'O$
00'0$
%68+
%LL+
SOEL
Figure 13: HBM demand by customer (2026E)
%65-

Source: UBS estimates
Figure 15: DDR cycles in context

Source: Company data, UBS estimates
Figure 17: DRAM bit demand per application (2026E)

Source: Gartner, UBS estimates
Ot'O$
SEO$
OE'O$
SZO$
SL'O$
OL'O$
SO'O$
00'0$
1Q16
SOOL
Figure 14: HBM demand by customer (2027E)
% 09-
%EL+
4Q16

Source: UBS estimates
Figure 16: NAND cycles in context

Source: Company data, UBS estimates
Figure 18: DRAM bit demand per application (2027E)

Source: Gartner, UBS estimates
4028E
Figure 19: NAND bit demand per application (2026E)

5,500
Source: Gartner, UBS estimates
Figure 21: Memory as % of flagship/premium smartphone BOM costs, 1Q22-4Q27E

Source: Counterpoint, Fomalhaut, UBS estimates
Figure 23: Samsung and SK Hynix - DRAM finished wafers and dies inventories

Source: Company data, UBS estimates
Figure 20: NAND bit demand per application (2027E)

Source: Gartner, UBS estimates
Figure 22: Memory as % of mid-range/lower-end smartphones, 1Q22-4Q27E

Source: UBS estimates
Figure 24: Samsung and SK Hynix - NAND finished wafers and dies inventories

Source: Company data, UBS estimates
Figure 25: DRAM customer inventories (weeks)

Source: Company data, UBS estimates
Figure 27: DRAM industry wafer starts and YoY growth

Source: Company data, UBS estimates
Figure 29: DRAM industry bit supply and YoY growth

Source: Company data, UBS estimates
Figure 26: NAND customer inventories (weeks)

Source: Company data, UBS estimates
Figure 28: NAND industry wafer starts and YoY growth

Source: Company data, UBS estimates
Figure 30: NAND industry bit supply and YoY growth

Source: Company data, UBS estimates
Figure 31: CXMT wafer capacity and as % of DRAM industry

Source: UBS estimates
Figure 32: CXMT bit shipments and as % of DRAM industry

Source: UBS estimates
Figure 33: UBS DRAM supply/demand summary
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Revenues ($ mil) | 26,415 | 31,776 | 39,234 | 52,958 | 150,384 | 98,109 | 146,320 | 185,464 | 212,130 | 642,023 | 241,172 | 268,773 | 294,065 | 313,825 | 1,117,835 | |
| %QoQ | -7.3% | 20.3% | 23.5% | 35.0% | 85.3% | 49.1% | 26.8% | 14.4% | 13.7% | 11.4% | 9.4% | 6.7% | ||||
| %YoY | 46.4% | 39.0% | 54.3% | 85.8% | 58.6% | 271.4% | 360.5% | 372.7% | 300.6% | 326.9% | 145.8% | 83.7% | 58.6% | 47.9% | 74.1% | |
| Supply (Shipments Based) (M 1Gb) | Supply (Shipments Based) (M 1Gb) | 64,317 | 75,938 | 85,576 | 87,400 | 313,229 | 89,789 | 96,744 | 102,723 | 107,178 | 396,433 | 109,881 | 116,451 | 123,213 | 128,510 | 478,055 |
| %QoQ | -2.6% | 18.1% | 12.7% | 2.1% | 2.7% | 7.7% | 6.2% | 4.3% | 2.5% | 6.0% | 5.8% | 4.3% | ||||
| %YoY | 3.4% | 14.2% | 28.9% | 32.3% | 19.9% | 39.6% | 27.4% | 20.0% | 22.6% | 26.6% | 22.4% | 20.4% | 19.9% | 19.9% | 20.6% | |
| Demand (End Consumption) (M 1Gb) | Demand (End Consumption) (M 1Gb) | 69,879 | 73,066 | 87,229 | 94,120 | 324,295 | 89,603 | 91,394 | 100,943 | 109,547 | 391,487 | 112,620 | 125,154 | 144,918 | 165,927 | 548,619 |
| %QoQ | 0.2% | 4.6% | 19.4% | 7.9% | -4.8% | 2.0% | 10.4% | 8.5% | 2.8% | 11.1% | 15.8% | 14.5% | ||||
| %YoY | 17.5% | 21.0% | 35.1% | 34.9% | 27.6% | 28.2% | 25.1% | 15.7% | 16.4% | 20.7% | 25.7% | 36.9% | 43.6% | 51.5% | 40.1% | |
| DDR Demand (M 1Gb) | 306,809 | 359,316 | 487,099 | |||||||||||||
| %YoY | 24.9% | 17.1% | 35.6% | |||||||||||||
| Inventory adjustment | -6,150 | 3,250 | 200 | -700 | -3,400 | 17,300 | 13,850 | 6,900 | -750 | 37,300 | -1,200 | -8,300 | -21,400 | -36,600 | -67,500 | |
| Inventory Adj. Demand (M 1Gb) | 63,729 | 76,316 | 87,429 | 93,420 | 320,895 | 106,903 | 105,244 | 107,843 | 108,797 | 428,787 | 111,420 | 116,854 | 123,518 | 129,327 | 481,119 | |
| %QoQ | -5.8% | 19.8% | 14.6% | 6.9% | 14.4% | -1.6% | 2.5% | 0.9% | 2.4% | 4.9% | 5.7% | 4.7% | ||||
| %YoY | 1.6% | 10.5% | 28.7% | 38.1% | 20.0% | 67.7% | 37.9% | 23.3% | 16.5% | 33.6% | 4.2% | 11.0% | 14.5% | 18.9% | 12.2% | |
| Wafer out (k wpm, 12" equi.) | 1,716 | 1,764 | 1,804 | 1,854 | 1,784 | 1,889 | 1,907 | 1,929 | 1,961 | 1,921 | 2,009 | 2,066 | 2,134 | 2,203 | 2,103 | |
| %QoQ | 2.5% | 2.8% | 2.3% | 2.8% | 1.9% | 1.0% | 1.2% | 1.7% | 2.4% | 2.9% | 3.3% | 3.2% | ||||
| %YoY | 17.9% | 16.5% | 13.4% | 10.8% | 14.5% | 10.1% | 8.1% | 7.0% | 5.8% | 7.7% | 6.3% | 8.4% | 10.6% | 12.3% | 9.4% | |
| %Sufficiency | Ratio | 0.9% | -0.5% | -2.1% | -6.4% | -2.4% | -16.0% | -8.1% | -4.7% | -1.5% | -7.5% | -1.4% | -0.3% | -0.2% | -0.6% | -0.6% |
| Blended ASP incl. HBM ($/Gb) | Blended ASP incl. HBM ($/Gb) | 0.41 | 0.42 | 0.46 | 0.61 | 0.48 | 1.09 | 1.51 | 1.81 | 1.98 | 1.62 | 2.19 | 2.31 | 2.39 | 2.44 | 2.34 |
| %QoQ | -4.8% | 1.9% | 9.6% | 32.2% | 80.3% | 38.4% | 19.4% | 9.6% | 10.9% | 5.2% | 3.4% | 2.3% | ||||
| %YoY | 41.5% | 21.8% | 19.7% | 40.4% | 32.2% | 166.0% | 261.4% | 293.8% | 226.6% | 237.3% | 100.9% | 52.6% | 32.2% | 23.4% | 44.4% | |
| DDR ASP ($/Gb) | 0.33 | 0.35 | 0.38 | 0.53 | 0.40 | 1.05 | 1.51 | 1.81 | 1.99 | 1.61 | 2.11 | 2.21 | 2.28 | 2.32 | 2.24 | |
| %QoQ | -6.2% | 7.0% | 7.0% | 42.0% | 97.0% | 43.0% | 20.0% | 10.0% | 6.0% | 5.0% | 3.0% | 2.0% | ||||
| %YoY | 20.3% | 10.6% | 11.8% | 52.5% | 26.4% | 220.3% | 328.0% | 380.0% | 271.9% | 297.2% | 100.1% | 46.9% | 26.1% | 16.9% | 39.1% | |
| HBM ASP ($/Gb) | 1.42 | 1.55 | 1.56 | 1.52 | 1.52 | 1.57 | 1.59 | 1.79 | 1.91 | 1.74 | 2.89 | 3.06 | 3.20 | 3.31 | 3.13 | |
| %QoQ | 2.7% | 9.4% | 0.4% | -2.5% | 3.0% | 1.4% | 12.7% | 6.7% | 51.4% | 5.7% | 4.8% | 3.2% | ||||
| %YoY | 25.6% | 28.3% | 16.9% | 10.0% | 15.6% | 10.3% | 2.3% | 14.8% | 25.6% | 14.8% | 84.6% | 92.5% | 79.0% | 73.2% | 79.4% |
Source: Company data, UBS estimates
Figure 34: DRAM vendors' market share in bit shipments
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | 39.7% | 37.5% | 38.3% | 38.2% | 38.4% | 37.6% | 38.7% | 38.3% | 36.0% | 37.6% | 37.0% | 37.7% | 37.7% | 37.4% | 37.4% |
| SK Hynix | 29.1% | 30.9% | 29.6% | 29.2% | 29.7% | 28.3% | 28.3% | 29.4% | 31.5% | 29.5% | 29.8% | 29.3% | 29.2% | 29.1% | 29.3% |
| Micron | 23.0% | 23.4% | 23.9% | 23.5% | 23.5% | 24.1% | 23.0% | 22.4% | 22.4% | 22.9% | 22.7% | 22.5% | 22.3% | 22.2% | 22.4% |
| Nanya Tech | 1.0% | 1.4% | 1.6% | 1.8% | 1.5% | 1.6% | 1.4% | 1.3% | 1.1% | 1.4% | 1.1% | 1.1% | 1.3% | 1.4% | 1.2% |
| CXMT | 6.8% | 6.5% | 6.4% | 7.0% | 6.7% | 8.2% | 8.3% | 8.4% | 8.8% | 8.4% | 9.2% | 9.2% | 9.4% | 9.7% | 9.4% |
| Others | 0.3% | 0.3% | 0.2% | 0.2% | 0.3% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% | 0.2% |
| Total Supply | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% |
Source: Gartner, Company data, UBS estimates
Figure 35: DRAM vendors' bit shipments QoQ and YoY growth
| Q125 QoQ | Q225 QoQ | Q325 QoQ | Q425 QoQ | 2025 YoY | Q126 QoQ | Q226 QoQ | Q326E QoQ | Q426E QoQ | 2026E YoY | Q127E QoQ | Q226E QoQ | Q327E QoQ | Q427E QoQ | 2027E YoY | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | 1.5% | 11.5% | 15.0% | 2.0% | 9.7% | 1.0% | 11.0% | 5.0% | -2.0% | 24.0% | 5.4% | 7.9% | 5.8% | 3.5% | 20.0% |
| SK Hynix | -7.0% | 25.0% | 8.0% | 0.8% | 24.3% | -0.3% | 7.5% | 10.3% | 12.0% | 25.6% | -3.1% | 4.1% | 5.4% | 4.0% | 20.0% |
| Micron | -8.0% | 20.3% | 15.0% | 0.5% | 21.8% | 5.1% | 3.2% | 3.4% | 4.0% | 23.5% | 4.0% | 5.0% | 5.0% | 4.0% | 17.9% |
| Nanya Tech | 8.4% | 68.0% | 26.4% | 11.9% | 54.3% | -6.2% | -3.0% | -6.7% | -6.9% | 15.3% | 2.5% | 5.6% | 17.6% | 14.9% | 9.7% |
| CXMT | 14.0% | 12.5% | 10.8% | 12.8% | 67.5% | 19.2% | 9.4% | 7.6% | 9.1% | 59.7% | 7.2% | 6.8% | 7.9% | 7.9% | 34.8% |
| Others | -4.8% | 2.5% | 2.4% | 2.4% | 3.1% | -2.3% | 0.0% | 2.4% | 0.0% | 2.4% | -2.3% | 0.0% | 2.4% | 0.0% | 0.0% |
| Bit supply Growth | -2.6% | 18.1% | 12.7% | 2.1% | 19.9% | 2.7% | 7.7% | 6.2% | 4.3% | 26.6% | 2.5% | 6.0% | 5.8% | 4.3% | 20.6% |
Source: Gartner, Company data, UBS estimates
Figure 36: DRAM vendors' wafers out (k wpm, 12" equivalent)
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | 663 | 667 | 668 | 674 | 668 | 676 | 678 | 685 | 689 | 682 | 695 | 707 | 719 | 731 | 713 | |
| SK Hynix | 475 | 499 | 519 | 544 | 509 | 564 | 572 | 583 | 600 | 580 | 618 | 636 | 654 | 672 | 645 | |
| Micron | 315 | 313 | 311 | 309 | 312 | 307 | 307 | 307 | 307 | 307 | 317 | 327 | 337 | 347 | 332 | |
| Nanya Tech | 49 | 53 | 59 | 60 | 57 | 60 | 60 | 60 | 60 | 60 | 60 | 63 | 70 | 78 | 70 | |
| CXMT | 164 | 182 | 198 | 217 | 190 | 232 | 240 | 244 | 256 | 243 | 269 | 284 | 304 | 325 | 295 | |
| Other | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | 50 | |
| Total | 1,716 | 1,764 | 1,804 | 1,854 | 1,784 | 1,889 | 1,907 | 1,929 | 1,961 | 1,921 | 2,009 | 2,066 | 2,134 | 2,203 | 2,103 | |
| %YoY | 17.9% | 16.5% | 13.4% | 10.8% | 14.5% | 10.1% | 8.1% | 7.0% | 5.8% | 7.7% | 6.3% | 8.4% | 10.6% | 12.3% | 9.4% |
Source: Gartner, Company data, UBS estimates
Figure 37: DRAM bit demand growth by application
| Q125 QoQ | Q225 QoQ | Q325 QoQ | Q425 QoQ | 2025 YoY | Q126 QoQ | Q226 QoQ | Q326E QoQ | Q426E QoQ | 2026E YoY | Q127E QoQ | Q226E QoQ | Q327E QoQ | Q427E QoQ | 2027E YoY | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total PC DRAM | -6.1% | 13.8% | 18.5% | 3.0% | 20.2% | -15.2% | -2.0% | 1.2% | -1.4% | -4.6% | -10.3% | 8.8% | 17.1% | 7.1% | 5.0% |
| Desktops | 0.4% | 10.8% | 9.3% | 5.6% | 22.8% | -12.9% | -5.4% | -6.1% | 0.2% | -9.7% | -3.9% | 8.4% | 15.2% | 10.7% | 8.3% |
| Notebooks | -8.1% | 14.6% | 20.7% | 2.2% | 24.3% | -15.3% | -0.2% | 2.5% | -1.1% | -2.3% | -10.3% | 9.6% | 18.3% | 7.7% | 7.9% |
| Upgrade Modules | -8.1% | 17.3% | 27.5% | 1.9% | -10.9% | -20.0% | -8.3% | 12.1% | -8.3% | -8.7% | -30.0% | 0.0% | 10.0% | -18.6% | -32.2% |
| Servers | 5.4% | 7.2% | 35.8% | 11.1% | 57.3% | 0.9% | 7.1% | 18.6% | 9.1% | 49.9% | 16.8% | 19.7% | 21.3% | 16.8% | 83.1% |
| Graphics | 13.2% | 13.4% | 17.5% | -1.0% | 26.8% | -9.7% | 5.3% | 13.4% | 0.0% | 10.5% | -8.4% | 4.9% | 13.1% | -0.2% | 8.8% |
| Game Consoles | 22.7% | 4.2% | 4.0% | 1.9% | 18.3% | 0.3% | 2.3% | 4.2% | -8.5% | 6.5% | 2.3% | 2.7% | 4.3% | -3.4% | 1.5% |
| HDTVs | -17.5% | 1.8% | 21.2% | 5.8% | 9.7% | -21.5% | 1.3% | 16.2% | 14.5% | 2.0% | -21.1% | 3.4% | 15.7% | 12.2% | 7.2% |
| Total Mobile DRAM | -5.6% | -1.0% | 14.1% | 10.2% | 15.6% | -13.2% | -6.3% | 4.4% | 12.6% | -0.6% | -7.5% | -2.9% | 8.1% | 16.0% | 7.5% |
| Smartphones | -4.8% | -1.4% | 14.6% | 10.1% | 16.7% | -13.2% | -7.1% | 4.4% | 12.8% | -1.2% | -6.8% | -3.5% | 8.1% | 16.6% | 7.8% |
| Apple | -31.8% | -5.6% | 29.4% | 80.5% | 31.3% | -32.6% | -5.8% | 4.7% | 74.7% | 28.6% | -32.1% | -5.8% | 1.1% | 87.0% | 17.5% |
| Samsung | 26.2% | -7.0% | 11.5% | -2.8% | 17.6% | 8.3% | 4.5% | 6.9% | -10.0% | 14.5% | 16.0% | -8.1% | 9.5% | -9.5% | -15.9% |
| Non-Apple/Samsung | -3.0% | 1.7% | 11.8% | -6.1% | 12.0% | -9.7% | -12.4% | 3.0% | -3.8% | -16.7% | 3.2% | 0.2% | 11.1% | -3.6% | 13.8% |
| Tablet PCs | -16.0% | 5.0% | 8.9% | 10.2% | 3.2% | -13.1% | 4.1% | 5.3% | 10.2% | 6.9% | -16.3% | 5.3% | 8.4% | 8.9% | 3.2% |
| Others | 4.3% | 4.0% | 2.8% | 2.9% | 12.2% | 3.0% | 2.6% | 1.8% | 7.3% | 12.8% | -11.2% | 2.7% | 3.6% | 8.8% | 0.8% |
| Total Bit Demand | 0.2% | 4.6% | 19.4% | 7.9% | 27.6% | -4.8% | 2.0% | 10.4% | 8.5% | 20.7% | 2.8% | 11.1% | 15.8% | 14.5% | 40.1% |
Source: Gartner, Company data, UBS estimates
Figure 38: DRAM content assumptions (MB/unit) by application
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total PC | 12,038 | 12,569 | 12,982 | 13,302 | 12,766 | 13,312 | 13,548 | 13,745 | 13,965 | 13,637 | 14,008 | 14,214 | 14,697 | 15,231 | 14,579 | |
| %YoY | 9% | 12% | 15% | 16% | 13% | 11% | 8% | 6% | 5% | 7% | 5% | 5% | 7% | 9% | 7% | |
| Desktops | 11,860 | 12,370 | 12,422 | 12,422 | 12,283 | 12,422 | 12,622 | 12,822 | 13,022 | 12,708 | 13,022 | 13,222 | 13,722 | 14,222 | 13,589 | |
| Notebooks | 12,401 | 13,001 | 13,601 | 14,081 | 13,328 | 14,081 | 14,281 | 14,481 | 14,681 | 14,379 | 14,681 | 14,881 | 15,381 | 15,881 | 15,252 | |
| Upgrade Modules | 9,900 | 10,100 | 10,300 | 10,500 | 10,232 | 10,500 | 10,700 | 10,900 | 11,100 | 10,794 | 11,100 | 11,100 | 11,100 | 11,300 | 11,144 | |
| Servers | 667,847 | 707,847 | 770,462 | 836,694 | 752,827 | 871,694 | 916,694 | 966,694 | 1,026,434 | 949,280 | 1,176,434 | 1,376,434 | 1,576,434 | 1,792,029 | 1,490,002 | |
| %YoY | 17% | 20% | 26% | 34% | 26% | 31% | 30% | 25% | 23% | 26% | 35% | 50% | 63% | 75% | 57% | |
| Smartphones | 7,739 | 7,829 | 8,255 | 8,611 | 8,126 | 8,664 | 8,692 | 8,988 | 9,333 | 8,924 | 9,240 | 9,104 | 9,321 | 10,014 | 9,438 | |
| %YoY | 11% | 11% | 13% | 15% | 13% | 12% | 11% | 9% | 8% | 10% | 7% | 5% | 4% | 7% 12,800 | 6% | |
| Apple | 7,600 | 7,700 | 8,940 | 9,590 | 8,652 | 9,690 | 9,690 | 10,400 | 10,800 | 10,227 | 11,000 | 11,000 | 11,400 | 11,711 | ||
| Samsung | 8,619 | 8,359 | 8,866 | 8,764 | 8,655 | 9,573 | 9,202 | 9,643 | 9,577 | 9,496 | 9,989 | 9,538 | 10,126 | 9,917 | 7,749 |
Source: Gartner, Company data, UBS estimates
Figure 39: UBS server DRAM bottom-up model
| 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|
| Total Server shipments (K unit) | 13,550 | 13,967 | 14,943 | 14,087 | 16,025 | 20,078 | 23,874 | 27,844 | |
| %YoY | 7.5% | 3.1% | 7.0% | -5.7% | 13.8% | 25.3% | 18.9% | 16.6% | |
| Traditional enterprise | 5,403 | 5,147 | 5,028 | 4,217 | 4,283 | 4,214 | 4,298 | 4,384 | |
| %YoY | -5.9% | -4.7% | -2.3% | -16.1% | 1.6% | -1.6% | 2.0% | 2.0% | |
| %of total | 39.9% | 36.8% | 33.6% | 29.9% | 26.7% | 21.0% | 18.0% | 15.7% | |
| Hyperscalers | 8,147 | 8,822 | 9,915 | 9,883 | 11,102 | 15,864 | 19,576 | 23,460 | |
| %YoY | 25.2% | 8.3% | 12.4% | -0.3% | 12.3% | 42.9% | 23.4% | 19.8% | |
| %of total | 60.1% | 63.2% | 66.4% | 70.2% | 69.3% | 79.0% | 82.0% | 84.3% | |
| Standard servers | 7,355 | 7,914 | 8,916 | 8,841 | 9,265 | 14,363 | 18,164 | 21,427 | |
| %of | %YoY | 23.8% | 7.6% | 12.7% | -0.8% | 4.8% | 55.0% 90.5% | 26.5% | 18.0% 91.3% |
| total hyperscalers %of total servers | 90.3% | 89.7% | 89.9% | 89.5% | 83.5% | 92.8% | |||
| 54.3% | 56.7% | 59.7% | 62.8% | 57.8% | 71.5% | 76.1% | 77.0% | ||
| AI Servers | %YoY | 520 | 630 | 730 | 783 | 1,591 | 1,243 | 1,146 | 1,759 |
| %of | hyperscalers | 73.3% | 21.2% | 15.9% | 7.3% | 103.1% | -21.8% | -7.8% | 53.5% |
| total %of | total servers | 6.4% 3.8% | 7.1% 4.5% | 7.4% 4.9% | 7.9% 5.6% | 14.3% 9.9% | 7.8% | 5.9% 4.8% | 7.5% 6.3% |
| Database servers | 272 | 270 | 259 | 246 | 6.2% 258 | 266 | 274 | ||
| %YoY | 278 | 5.0% | 3.0% | ||||||
| %of total | hyperscalers | 2.0% | 2.0% | -3.0% | -4.0% | -5.0% | 3.0% | ||
| %of | total servers | 3.3% | 3.1% | 2.7% | 2.6% | 2.2% | 1.6% | 1.4% | 1.2% |
| Avg. DRAM content (GB) | 392 | 457 | 461 | 524 | 576 | 735 | 927 | 1,455 | |
| %YoY | 32.5% | 16.7% | 0.9% | 13.5% | 9.9% | 27.7% | 26.1% | 57.0% | |
| Traditional enterprise | 356 | 417 | 420 | 464 | 470 | 480 | 500 | 520 | |
| %YoY | 27.4% | 17.3% | 0.7% | 10.3% | 1.4% | 2.1% | 4.2% | 4.0% | |
| Hyperscalers Blended | 416 | 480 | 482 | 549 | 650 | 803 | 1,021 | 1,630 | |
| %YoY | 27.3% | 15.5% | 0.4% | 13.8% | 18.4% | 23.5% | 27.1% | 59.7% | |
| Standard servers | %YoY | 28.3% | 7.3% | ||||||
| 11.7% | 0.5% | 6.3% | 1.1% | 3.4% | 7.8% | ||||
| AI servers | 600 | 840 | 860 | 1,348 | 1,647 | 4,435 | 8,961 | 14,754 | |
| %YoY | 7.1% | 40.0% | 2.4% | 56.8% | 22.1% | 169.3% | 102.1% | 64.7% | |
| Database Servers | %YoY | 900 | 1,100 | 1,120 1.8% | 1,200 7.1% | 1,210 0.8% | 1,280 5.8% | 1,340 4.7% | 1,420 |
| Total DRAM Demand (mn Gb) | 36.4% 42,491 | 22.2% 51,095 | 55,158 | 59,039 | 73,831 | 118,088 | 177,057 | 6.0% 324,124 | |
| %YoY | 42.3% | 20.2% | 8.0% | 7.0% | 25.1% | 59.9% | 49.9% | 83.1% | |
| Total DDR Demand (mn Gb) | 42,491 | 51,095 | 55,158 | 56,355 | 65,224 | 100,602 | 144,886 | 262,604 | |
| Traditional enterprise | %YoY | 42.3% | 20.2% | 8.0% | 2.2% | 15.7% | 54.2% | 44.0% | 81.2% |
| %YoY | 15,382 | 17,191 11.8% | 16,908 -1.6% | 15,641 -7.5% | 16,106 3.0% | 16,181 0.5% | 17,193 6.3% | 18,238 | |
| Hyperscalers | 19.8% 27,110 | 33,904 | 38,250 | 43,398 | 57,725 | 101,907 | 159,864 | 6.1% 305,886 | |
| Standard | %YoY | 59.3% | 25.1% | 13.5% | 76.5% | 91.3% | |||
| servers | 22,652 | 27,225 | 12.8% 30,813 | 32,465 | 33.0% 34,393 | 55,152 | 56.9% 74,835 | 95,136 | |
| %YoY | 58.9% | 20.2% | 13.2% | 5.4% | 5.9% | 60.4% | 35.7% | 27.1% | |
| servers | 2,496 | 4,234 | 5,022 | 8,448 | 20,952 | 44,111 | 82,179 | 207,639 | |
| AI | %YoY | 69.6% | |||||||
| HBM DRAM | 85.7% | 152.7% | |||||||
| 18.6% | 86.3% | ||||||||
| 110.5% | |||||||||
| Demand | 68.2% | 148.0% | 17,486 | 32,171 | 61,520 | ||||
| 2,684 | 8,607 220.7% | 103.2% 26,625 | 84.0% 50,009 | 91.2% 146,119 | |||||
| DDR DRAM Demand | %YoY | 12,345 | 87.8% | 192.2% | |||||
| LPDDR / SOCAMM | %YoY | 6,081 | 103.0% | 115.7% | 14,626 | 38,001 | |||
| Demand | 468 | 7,613 | 159.8% | ||||||
| %YoY | 1,962 | 2,445 | 2,415 -1.2% | 1527.9% 2,643 | 92.1% 2,850 | 3,111 | |||
| Database Servers | %YoY | 39.1% | 24.7% | 2,484 | 2,380 | 11.1% | 7.8% | 9.1% | |
| 2.9% | |||||||||
| -4.2% |
Source: Dell'Oro Group, Gartner, Company data, UBS estimates
Figure 40: UBS bottom-up server procurement model (k units)
| (K Units) | 2022 | 2023 | 2024 | 2025 | 1Q26 | 2Q26 | 3Q26E | 4Q26E | 2026E | 1Q27E | 2Q27E | 3Q27E | 4Q27E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Quarterly | |||||||||||||||
| Alphabet | 1,039 | 1,085 | 1,247 | 1,845 | 530 | 600 | 580 | 640 | 2,350 | 720 | 720 | 760 | 750 | 2,950 | |
| AWS | 2,237 | 2,131 | 2,276 | 3,087 | 940 | 960 | 1,020 | 1,000 | 3,920 | 1,150 | 1,180 | 1,250 | 1,240 | 4,820 | |
| Microsoft | 1,056 | 1,274 | 1,468 | 2,457 | 680 | 760 | 830 | 820 | 3,090 | 850 | 920 | 1,000 | 970 | 3,740 | |
| Meta | 2,192 | 2,027 | 2,188 | 3,409 | 950 | 1,080 | 1,100 | 1,050 | 4,180 | 1,200 | 1,280 | 1,250 | 1,260 | 4,990 | |
| Baidu | 71 | 70 | 37 | 30 | 10 | 10 | 11 | 10 | 41 | 13 | 14 | 13 | 12 | 52 | |
| Alibaba | 376 | 188 | 305 | 456 | 135 | 145 | 155 | 160 | 595 | 158 | 167 | 175 | 180 | 680 | |
| Tencent | 118 | 149 | 303 | 268 | 75 | 82 | 88 | 82 | 327 | 82 | 92 | 100 | 95 | 369 | |
| ByteDance | 358 | 550 | 770 | 1,306 | 360 | 420 | 430 | 440 | 1,650 | 450 | 480 | 520 | 490 | 1,940 | |
| Total Top 8 | 7,448 | 7,473 | 8,594 | 12,858 | 3,680 | 4,057 | 4,214 | 4,202 | 16,153 | 4,623 | 4,853 | 5,068 | 4,997 | 19,541 | |
| %YoY | 13.5% | 0.3% | 15.0% | 49.6% | 28.2% | 31.7% | 25.0% | 18.8% | 25.6% | 25.6% | 19.6% | 20.3% | 18.9% | 21.0% | |
| %QoQ | 4.1% | 10.2% | 3.9% | -0.3% | 10.0% | 5.0% | 4.4% | -1.4% | |||||||
| Others | 2,468 | 2,410 | 2,901 | 1,879 | 787 | 839 | 889 | 908 | 3,423 | 922 | 999 | 1,021 | 977 | 3,919 | |
| Total hyperscale | 9,915 | 9,883 | 11,494 | 14,737 | 4,467 | 4,896 | 5,103 | 5,110 | 19,576 | 5,545 | 5,852 | 6,089 | 5,974 | 23,460 | |
| %YoY | 12.4% | -0.3% | 16.3% | 28.2% | 38.8% | 51.2% | 26.5% | 20.3% | 32.8% | 24.1% | 19.5% | 19.3% | 16.9% | 19.8% | |
| %QoQ | 5.2% | 9.6% | 4.2% | 0.1% | 8.5% | 5.5% | 4.0% | -1.9% | |||||||
| YoY (%) | |||||||||||||||
| Alphabet | 24.0% | 4.4% | 15.0% | 48.0% | 51.8% | 6.8% | 26.9% | 34.1% | 27.3% | 35.8% | 20.0% | 31.0% | 17.2% | 25.5% | |
| AWS | 24.5% | -4.7% | 6.8% | 35.6% | 40.0% | 63.4% | 20.0% | 2.3% | 27.0% | 22.3% | 22.9% | 22.5% | 24.0% | 23.0% | |
| Microsoft | -2.9% | 20.6% | 15.3% | 67.4% | 17.9% | 29.3% | 20.4% | 35.9% | 25.7% | 25.0% | 21.1% | 20.5% | 18.3% | 21.0% | |
| Meta | 20.7% | -7.5% | 8.0% | 55.8% | 23.0% | 21.3% | 23.9% | 22.4% | 22.6% | 26.3% | 18.5% | 13.6% | 20.0% | 19.4% | |
| Baidu | -34.0% | -1.6% | -47.9% | -16.5% | 44.8% | 42.5% | 3.4% | 68.6% | 34.5% | 30.0% | 40.0% | 18.2% | 20.0% | 26.8% | |
| Alibaba | -10.6% | -50.2% | 62.5% | 49.6% | 81.9% | 10.3% | 41.3% | 13.8% | 30.5% | 17.0% | 15.2% | 12.9% | 12.5% | 14.3% | |
| Tencent | -56.1% | 25.9% | 103.3% | -11.6% | 2.3% | 42.6% | 103.2% | -12.2% | 22.2% | 9.3% | 12.2% | 13.6% | 15.9% | 12.8% | |
| ByteDance | 60.6% | 53.8% | 39.9% | 69.7% | 4.0% | 63.3% | 32.9% | 16.0% | 26.3% | 25.0% | 14.3% | 20.9% | 11.4% | 17.6% |
Source: Dell'Oro Group, company data, UBS estimates
Figure 41: UBS NAND Flash supply/demand summary
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Revenues ($ mil) | 12,302 | 14,726 | 16,969 | 21,320 | 65,317 | 39,006 | 70,368 | 89,070 | 102,157 | 300,602 | 109,320 | 118,629 | 129,841 | 134,658 | 492,447 | |
| %QoQ | -20.1% | 19.7% | 15.2% | 25.6% | 83.0% | 80.4% | 26.6% | 14.7% | 7.0% | 8.5% | 9.5% | 3.7% | ||||
| %YoY | -1.0% | -1.1% | 4.9% | 38.5% | 10.9% | 217.1% | 377.8% | 424.9% | 379.2% | 360.2% | 180.3% | 68.6% | 45.8% | 31.8% | 63.8% | |
| Supply (Shipments Based) (M 1GB) | 201,828 | 247,795 | 271,936 | 273,324 | 994,882 | 270,305 | 288,547 | 304,362 | 323,223 | 1,186,438 | 335,811 | 350,391 | 372,338 | 386,153 | 1,444,692 | |
| %QoQ | -6.0% | 22.8% | 9.7% | 0.5% | -1.1% | 6.7% | 5.5% | 6.2% | 3.9% | 4.3% | 6.3% | 3.7% | ||||
| %YoY | -6.3% | 15.1% | 25.6% | 27.3% | 15.4% | 33.9% | 16.4% | 11.9% | 18.3% | 19.3% | 24.2% | 21.4% | 22.3% | 19.5% | 21.8% | |
| Demand (End Consumption) (M 1GB) | 235,552 | 242,791 | 280,969 | 321,309 | 1,080,621 | 301,290 | 309,481 | 337,580 | 378,354 | 1,326,704 | 369,711 | 389,868 | 430,103 | 485,995 | 1,675,677 | |
| %QoQ | -8.1% | 3.1% | 15.7% | 14.4% | -6.2% | 2.7% | 9.1% | 12.1% | -2.3% | 5.5% | 10.3% | 13.0% | ||||
| %YoY | 11.4% | 13.0% | 20.1% | 25.4% | 17.9% | 27.9% | 27.5% | 20.1% | 17.8% | 22.8% | 22.7% | 26.0% | 27.4% | 28.4% | 26.3% | |
| Inventory adjustment | -39,600 | 3,700 | -6,200 | -33,500 | -75,600 | 25,000 | 22,000 | -20,000 | -50,000 | -23,000 | -31,000 | -36,000 | -55,000 | -100,000 | -222,000 | |
| Inventory Adj. Demand (M 1GB) | 195,952 | 246,491 | 274,769 | 287,809 | 1,005,021 | 326,290 | 331,481 | 317,580 | 328,354 | 1,303,704 | 338,711 | 353,868 | 375,103 | 385,995 | 1,453,677 | |
| %QoQ | -8.0% | 25.8% | 11.5% | 4.7% | 13.4% | 1.6% | -4.2% | 3.4% | 3.2% | 4.5% | 6.0% | 2.9% | ||||
| %YoY | -14.3% | 9.9% | 24.6% | 35.1% | 13.4% | 66.5% | 34.5% | 15.6% | 14.1% | 29.7% | 3.8% | 6.8% | 18.1% | 17.6% | 11.5% | |
| Wafer out (k wpm, 12" equi.) | 1,418 | 1,288 | 1,268 | 1,352 | 1,331 | 1,366 | 1,333 | 1,338 | 1,351 | 1,347 | 1,331 | 1,307 | 1,303 | 1,299 | 1,310 | |
| %QoQ | -0.4% | -9.2% | -1.5% | 6.6% | 1.1% | -2.5% | 0.4% | 1.0% | -1.5% | -1.8% | -0.3% | -0.3% | ||||
| %YoY | 23.7% | 2.5% | -5.8% | -5.1% | 2.9% | -3.6% | 3.5% | 5.5% | -0.1% | 1.2% | -2.6% | -2.0% | -2.6% | -3.9% | -2.8% | |
| %Sufficiency | Ratio | 3.0% | 0.5% | -1.0% | -5.0% | -1.0% | -17.2% | -13.0% | -4.2% | -1.6% | -9.0% | -0.9% | -1.0% | -0.7% | 0.0% | -0.6% |
| NAND ASP ($/1GB) | 0.06 | 0.06 | 0.06 | 0.08 | 0.07 | 0.14 | 0.24 | 0.29 | 0.32 | 0.25 | 0.33 | 0.34 | 0.35 | 0.35 | 0.34 | |
| %QoQ | -15.0% | -2.5% | 5.0% | 25.0% | 85.0% | 69.0% | 20.0% | 8.0% | 3.0% | 4.0% | 3.0% | 0.0% | ||||
| %YoY | 5.8% | -14.1% | -16.5% | 8.8% | -3.9% | 136.7% | 310.4% | 369.0% | 305.2% | 285.9% | 125.6% | 38.8% | 19.2% | 10.3% | 34.5% |
Source: Company data, UBS estimates
Figure 42: NAND Flash vendors' market share in bit shipments
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | 31.0% | 31.8% | 31.8% | 28.5% | 30.7% | 31.0% | 29.6% | 29.5% | 31.4% | 30.4% | 30.0% | 30.5% | 30.0% | 29.4% | 30.0% |
| Kioxia | 15.3% | 12.5% | 15.5% | 16.2% | 14.9% | 14.7% | 14.2% | 14.8% | 14.9% | 14.7% | 15.7% | 14.3% | 14.2% | 14.3% | 14.6% |
| SanDisk | 13.2% | 11.2% | 11.7% | 12.0% | 12.0% | 11.6% | 12.7% | 12.6% | 10.7% | 11.9% | 10.8% | 11.4% | 12.3% | 12.7% | 11.8% |
| SK Hynix | 12.8% | 18.2% | 15.7% | 17.1% | 16.1% | 15.5% | 16.7% | 16.3% | 16.3% | 16.2% | 16.0% | 16.3% | 16.3% | 16.1% | 16.2% |
| Micron | 14.1% | 14.3% | 12.6% | 13.1% | 13.5% | 13.7% | 13.5% | 13.6% | 13.6% | 13.6% | 13.4% | 13.3% | 12.9% | 12.8% | 13.1% |
| YMTC | 12.5% | 11.3% | 11.9% | 12.2% | 12.0% | 12.7% | 12.5% | 12.5% | 12.4% | 12.5% | 13.3% | 13.6% | 13.7% | 14.0% | 13.7% |
| Total Bit Supply | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% |
Source: Gartner, company data, UBS estimates
Figure 43: NAND Flash bit shipments QoQ and YoY growth
| Q125 QoQ | Q225 QoQ | Q325 QoQ | Q425 QoQ | 2025 YoY | Q126 QoQ | Q226 QoQ | Q327E QoQ | Q428E QoQ | 2026E YoY | Q127E QoQ | Q227E QoQ | Q327E QoQ | Q427E QoQ | 2027E YoY | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | -10.0% | 25.9% | 10.0% | -10.0% | 3.0% | 7.5% | 2.0% | 5.0% | 13.2% | 18.0% | -0.8% | 5.9% | 4.7% | 1.8% | 20.0% |
| Kioxia | -10.0% | 0.0% | 36.0% | 5.0% | 13.5% | -10.0% | 3.0% | 10.0% | 7.0% | 17.4% | 9.6% | -5.0% | 5.0% | 5.0% | 21.4% |
| SanDisk | -6.0% | 4.0% | 15.0% | 3.0% | 17.7% | -5.0% | 17.0% | 5.0% | -10.0% | 18.0% | 5.0% | 10.0% | 15.0% | 7.0% | 21.6% |
| SK Hynix | -16.9% | 73.6% | -5.0% | 9.6% | 11.8% | -10.4% | 15.0% | 2.7% | 6.5% | 20.0% | 1.7% | 6.5% | 6.0% | 2.5% | 21.3% |
| Micron | 3.5% | 24.3% | -3.3% | 5.0% | 25.6% | 3.4% | 5.0% | 6.0% | 6.0% | 20.4% | 3.0% | 3.0% | 3.0% | 3.4% | 17.2% |
| YMTC | 15.7% | 10.6% | 15.5% | 3.5% | 57.3% | 2.3% | 5.5% | 5.3% | 5.2% | 24.5% | 12.0% | 6.2% | 7.6% | 5.5% | 33.1% |
| Total Bit Supply | -6.0% | 22.8% | 9.7% | 0.5% | 15.4% | -1.1% | 6.7% | 5.5% | 6.2% | 19.3% | 3.9% | 4.3% | 6.3% | 3.7% | 21.8% |
Source: Gartner, company data, UBS estimates
Figure 44: NAND Flash vendors wafers out (k wpm, 12" equivalent)
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Samsung | 471 | 449 | 423 | 409 | 438 | 412 | 405 | 400 | 400 | 404 | 398 | 393 | 388 | 383 | 390 | |
| Kioxia / SanDisk | 470 | 400 | 400 | 480 | 437 | 480 | 460 | 460 | 450 | 463 | 420 | 400 | 400 | 400 | 405 | |
| SK Hynix | 216 | 189 | 185 | 186 | 194 | 185 | 178 | 183 | 201 | 186 | 208 | 204 | 200 | 196 | 202 | |
| Micron | 132 | 111 | 111 | 117 | 118 | 130 | 130 | 130 | 130 | 130 | 130 | 130 | 130 | 130 | 130 | |
| YMTC | 130 | 140 | 150 | 160 | 145 | 160 | 160 | 165 | 170 | 164 | 175 | 180 | 185 | 190 | 183 | |
| Total | 1,418 | 1,288 | 1,268 | 1,352 | 1,331 | 1,366 | 1,333 | 1,338 | 1,351 | 1,347 | 1,331 | 1,307 | 1,303 | 1,299 | 1,310 | |
| %YoY | 23.7% | 2.5% | -5.8% | -5.1% | 2.9% | -3.6% | 3.5% | 5.5% | -0.1% | 1.2% | -2.6% | -2.0% | -2.6% | -3.9% | -2.8% |
Source: Gartner, company data, UBS estimates
Figure 45: NAND Flash bit demand growth by application
| Q125 QoQ | Q225 QoQ | Q325 QoQ | Q425 QoQ | 2025 YoY | Q126 QoQ | Q226 QoQ | Q327E QoQ | Q428E QoQ | 2026E YoY | Q127E QoQ | Q227E QoQ | Q327E QoQ | Q427E QoQ | 2027E YoY | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total Embedded | -8.3% | 3.1% | 16.1% | 14.8% | 18.5% | -6.2% | 2.8% | 9.3% | 12.4% | 23.6% | -2.2% | 5.6% | 10.5% | 13.3% | 27.2% |
| Smartphones | -11.3% | -6.3% | 17.3% | 26.6% | 16.9% | -21.7% | -6.4% | 2.4% | 30.6% | 2.1% | -16.7% | -4.6% | 6.3% | 32.3% | 8.4% |
| iPhones | -29.0% | -7.7% | 24.8% | 69.7% | 28.1% | -33.6% | -5.0% | -1.6% | 70.4% | 15.5% | -32.8% | -5.0% | 1.3% | 73.4% | 9.5% |
| Samsung | 42.2% | -13.4% | 10.1% | -8.2% | 15.0% | 16.2% | -3.9% | 7.6% | -12.2% | 7.4% | 30.6% | -14.1% | 7.9% | -11.0% | 9.7% |
| Non- iPhone/Samsung | -5.6% | -0.9% | 13.2% | -5.0% | 5.5% | -15.1% | -10.2% | 5.5% | -1.6% | -18.2% | -2.7% | 3.9% | 13.7% | -1.5% | 5.3% |
| Tablet PCs | -21.6% | 10.0% | 9.5% | 17.2% | 6.4% | -20.2% | 12.2% | 4.5% | 17.0% | 11.3% | -24.7% | 13.1% | 10.6% | 14.3% | 6.6% |
| SSD-Total | -6.9% | 7.5% | 18.9% | 11.2% | 21.8% | 1.4% | 5.5% | 12.9% | 7.3% | 36.7% | 5.2% | 8.9% | 12.7% | 9.6% | 38.6% |
| SSD (notebooks) | -13.7% | 9.3% | 15.3% | -1.3% | 9.0% | -14.1% | -0.3% | 2.4% | 0.0% | -6.2% | -10.2% | 9.4% | 15.7% | 6.7% | 6.9% |
| SSD (servers) | -4.0% | 9.4% | 36.0% | 22.6% | 45.1% | 4.0% | 7.4% | 18.5% | 8.1% | 65.7% | 10.7% | 10.5% | 14.0% | 10.2% | 53.3% |
| SSD (storage) | 10.5% | 7.9% | 8.0% | 10.1% | 47.9% | 15.2% | 11.4% | 10.7% | 10.2% | 53.6% | 9.7% | 9.3% | 8.8% | 8.5% | 44.8% |
| SSD (others) | -14.6% | 1.4% | -0.5% | 0.2% | -12.8% | 1.3% | -0.5% | 8.8% | 8.8% | 8.0% | -6.0% | 0.2% | 9.4% | 12.2% | 12.5% |
| Total Flash cards | -0.4% | 1.6% | 2.2% | -3.4% | -0.4% | -5.4% | -2.8% | -1.6% | -2.9% | -9.8% | -8.1% | -6.8% | -4.4% | -10.8% | -20.0% |
| Mobile Phones - Flash cards | -6.3% | -1.5% | 5.3% | -1.6% | -0.8% | -3.7% | -6.8% | 1.5% | -7.4% | -8.8% | 7.1% | -0.3% | 6.7% | -6.1% | 1.4% |
| Total Bit Demand | -8.1% | 3.1% | 15.7% | 14.4% | 17.9% | -6.2% | 2.7% | 9.1% | 12.1% | 22.8% | -2.3% | 5.5% | 10.3% | 13.0% | 26.3% |
Source: Gartner, company data, UBS estimates
Figure 46: NAND Flash content assumptions (GB/unit) by application
| Q125 | Q225 | Q325 | Q425 | 2025 | Q126 | Q226 | Q326E | Q426E | 2026E | Q127E | Q227E | Q327E | Q427E | 2027E | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Smartphones - Blended | 198 | 193 | 216 | 251 | 216 | 223 | 223 | 227 | 276 | 238 | 243 | 235 | 238 | 293 | 253 | |
| %YoY | 14% | 14% | 11% | 22% | 15% | 12% | 16% | 5% | 10% | 10% | 9% | 5% | 5% | 6% | 6% | |
| iPhones | 517 | 512 | 573 | 579 | 538 | 576 | 581 | 586 | 594 | 572 | 599 | 604 | 627 | 653 | 610 | |
| Samsung | 225 | 203 | 213 | 198 | 210 | 232 | 206 | 217 | 210 | 216 | 247 | 220 | 230 | 222 | 230 | |
| Non- iPhone/Samsung | 127 | 127 | 132 | 137 | 131 | 127 | 132 | 137 | 142 | 134 | 132 | 137 | 142 | 146 | 139 | |
| SSD - Notebook PCs | 643 | 639 | 674 | 676 | 659 | 663 | 658 | 686 | 706 | 678 | 695 | 681 | 710 | 734 | 706 | |
| %YoY | 2% | 3% | 2% | 0% | 2% | 3% | 3% | 2% | 4% | 3% | 5% | 3% | 4% | 4% | 4% | |
| SSD - Servers | 3,433 | 3,775 | 4,117 | 4,780 | 4,083 | 4,975 | 5,170 | 5,366 | 5,561 | 5,286 | 5,951 | 6,342 | 6,733 | 7,123 | 6,563 | |
| %YoY | 12% | 23% | 34% | 55% | 33% | 45% | 37% | 30% | 16% | 29% | 20% | 23% | 25% | 28% | 24% | |
| SSD - Storage | 12,383 | 13,359 | 14,434 | 15,885 | 14,015 | 16,862 | 17,838 | 18,815 | 19,791 | 18,386 | 20,768 | 21,744 | 22,721 | 23,697 | 22,282 | |
| %YoY | 57% | 47% | 41% | 39% | 45% | 36% | 34% | 30% | 25% | 31% | 23% | 22% | 21% | 20% | 21% |
Source: Gartner, company data, UBS estimates
Valuation Method and Risk Statement
Investing in the technology sector involves a high degree of risk. Rapid technological changes, increasing competition and exposure to macroeconomic cycles are among the many risks faced by investors in technology stocks. Moreover, it is extremely difficult to project the financial results of tech companies since their operating models are highly volatile and unpredictable. Finally, valuing technology stocks can prove challenging, as neither traditional nor non-traditional valuation measures have provided much insight into how these stocks trade.
Required Disclosures
This document has been prepared by UBS Securities Pte. Ltd., an affiliate of UBS AG. UBS AG, its subsidiaries, branches and affiliates, including former Credit Suisse AG and its subsidiaries, branches and affiliates are referred to herein as "UBS".
For information on the ways in which UBS manages conflicts and maintains independence of its UBS Global Research product; historical performance information; certain additional disclosures concerning UBS Global Research recommendations; and terms and conditions for certain third party data used in research report, please visit https://www.ubs.com/disclosures. Unless otherwise indicated, information and data in this report are based on company disclosures including but not limited to annual, interim, quarterly reports and other company announcements. The figures contained in performance charts refer to the past; past performance is not a reliable indicator of future results. Additional information will be made available upon request. UBS Securities Co. Limited is licensed to conduct securities investment consultancy businesses by the China Securities Regulatory Commission. UBS acts or may act as principal in the debt securities (or in related derivatives) that may be the subject of this report. This recommendation was finalized on: 07 August 2026 09:18 AM GMT. UBS has designated certain UBS Global Research department members as Derivatives Research Analysts where those department members publish research principally on the analysis of the price or market for a derivative, and provide information reasonably sufficient upon which to base a decision to enter into a derivatives transaction. Where Derivatives Research Analysts coauthor research reports with Equity Research Analysts or Economists, the Derivatives Research Analyst is responsible for the derivatives investment views, forecasts, and/or recommendations. Quantitative Research Review: UBS Global Research publishes a quantitative assessment of its analysts' responses to certain questions about the likelihood of an occurrence of a number of short term factors in a product known as the 'Quantitative Research Review'. Views contained in this assessment on a particular stock reflect only the views on those short term factors which are a different timeframe to the 12-month timeframe reflected in any equity rating set out in this note. For the latest responses, please see the Quantitative Research Review Addendum at the back of this report, where applicable. For previous responses please make reference to (i) previous UBS Global Research reports; and (ii) where no applicable research report was published that month, the Quantitative Research Review which can be found at https://neo.ubs.com/ quantitative, or contact your UBS sales representative for access to the report or the Quantitative Research Team on ubs-quantanswers@ubs.com. A consolidated report which contains all responses is also available and again you should contact your UBS sales representative for details and pricing or the Quantitative Research team on the email above.