| 共识点 | HSBC 表述 | UBS 表述 | MS 表述 |
|---|---|---|---|
| ① Server CPU 短缺是真的 | "underlying demand potentially up 60% y-o-y but shipments likely constrained to 20-21% growth" | "undershipping demand by at least $1B ... server CPU market demand 20% above shipments" | "1Q guidance 3 months ago was too pessimistic ... mgmt says undershipping demand by $1bn plus" |
| ② 价格会涨 | 2026 ASP +20%, 2027 +10% | "pricing tailwinds and ongoing supply improvements" | "much of the incremental y/y growth in DCAI is price" |
| ③ Foundry 财务难看但消息会继续正向 | "timing and revenue opportunity remain uncertain" | "newsflow on foundry should also remain positive" | "Foundry losses ... remain stubbornly high at $2.4bn" |
| ④ 至少对 EPS 估计要大幅上调 | 2026e EPS $0.80→$1.10 (+37.5%) | 2026e EPS $0.44→$1.39 (+219%) | 2026e EPS $0.68→$1.12 (+65%) |
| ⑤ CCG 是结构性逆风 | PC 市场走弱 | "CCG sets up as a headwind now into C2027 ... units are actually down" | "PC TAM that seems likely to be down 10%+ in 2026" |
"Intel should still grow server CPU shipments 20% y-o-y in 2026e and 2027e as it reallocates foundry capacity from client to server CPUs on Intel 3 and 7"
DCAI 2026e/27e 超共识 16%/33%。GM 40.9%/45.0% 超共识 37.1%/41.2%。
"NVIDIA is moving some CPU racks from Intel to Vera, Trainium is moving head node from Intel to ARM, both TPU and Trainium are moving host node from x86 to ARM. AMD is likely to favor AMD CPUs. So most of the important CPUs in AI are facing share headwinds."
"The shortage is likely to last for a while ... but the shortage will be critical to market share prospects until Intel can deliver leadership"
"We do not include the foundry business in our valuation as external customer related uncertainty remains."
$95 TP 全由 core business (server CPU 涨价) 驱动。2027e 26x PE on core EPS。
"EV per capacity of $232 (vs. prior $220) ... applying a 45% haircut to reflect node and yield gaps, customer stickiness, and utilization risk"
SOTP: Products 23x EV/EBIT + Foundry 按 $232/wafer 减 45% → $83。
"Foundry losses in the quarter remain stubbornly high at $2.4bn - roughly the same level as the last 4 quarters despite top line growth. There remains an official forecast of breakeven just 7 quarters from now, which seems challenging to us"
"we would love more detail on Terafab structure" — 明确质疑
| 券商 | 估值方法 | 应用到哪年 EPS | 倍数 | 得到 TP | 对应 EPS |
|---|---|---|---|---|---|
| HSBC | SOTP PE | 2027e (core only) | 26x | $95 | Core EPS ~$3.65 (implied) |
| UBS | SOTP (EV/EBIT + wafer-EV) | 2027e (blended) | 23x EV/EBIT | $83 | EPS $1.95 (UBS 2027e) |
| MS | PE | CY2027 | 42x | $73 | EPS $1.73 (MS 2027e) |
吊诡的事实：MS 给了 最高的 PE 倍数（42x）但得到 最低的 TP $73，因为 MS 的 CY2027 EPS 估计只有 $1.73（反映对份额/foundry 的悲观）。
HSBC 给 26x 却得 $95，因为假设 Core EPS ~$3.65 —— 两者差异的 75% 来自 EPS 假设，不是倍数。
| 维度 | HSBC (Buy $95) | UBS (Neutral $83) | MS (EW $73) |
|---|---|---|---|
| 发布日期 | 2026-04-21 (preview) | 2026-04-23 (post-Q1) | 2026-04-24 (post-Q1) |
| 当前股价参考 | $65.70 (4/20 收盘) | $80.07 (4/23 盘后) | $66.78 (4/23 收盘) |
| 上方空间 | +45% | +4% | +9% |
| 2026e Revenue | $56.7B | $59.5B | $53.1B |
| 2027e Revenue | $64.5B | $65.4B | n/a (implicit ~$58B) |
| UBS model：2026E 季度营收 | — | Q1/Q2/Q3/Q4 = $13.58B / $14.75B / $15.29B / $15.92B | — |
| UBS model：2027E 季度营收 | — | Q1/Q2/Q3/Q4 = $15.68B / $15.76B / $16.69B / $17.24B | — |
| 2026e Non-GAAP EPS | $1.10 | $1.39 | $1.12 |
| 2027e Non-GAAP EPS | $2.09 | $1.95 | $1.73 |
| 2026 ASP 假设 | +20% YoY | + (pricing tailwinds, 未量化) | mostly price, 未量化 |
| 2026 DCAI 增速 | +20% shipments × +20% ASP | +42% YoY revenue | ~+22% (Q1 实际延伸) |
| GM 2026e | 40.9% | ~40% | 39.0% (MS estimate) |
| GM 2027e | 45.0% | ~44% | ~42% |
| UBS model：Non-GAAP GM | — | 2026E 42.3% / 2027E 43.8% | — |
| UBS model：External Foundry Rev | — | 2026E $1.19B / 2027E $2.19B | — |
| Foundry 估值 | 不计入 | $232/wafer × 45% 折 | 不明（42x PE 中含一些） |
| 对 AMD 份额假设 | 中性 | "AMD 今年增长 ~45%，远超 INTC" | "AMD continues to be in share gain position" |
| 对 14A 外部客户假设 | 有但不量化 | "foundry deals with NVDA, AAPL (M-Series), maybe Amazon" | "no real color on Terafab" |
| Intel IR / 10-Q 新风险披露 | 未纳入估值，但构成风险 | 未显著展开 | 未显著展开 |
|
10-Q 原文：Intel 明确披露，若无法通过重大外部客户 design wins + Intel 自身产品路线拿到足够 committed demand，可能会 "pause or discontinuation of our pursuit of Intel 14A and other next-generation leading-edge process technologies"。 这意味着：14A 不是低成本免费期权，而是需要订单验证的高资本开销承诺。UBS 的乐观期权、HSBC 的忽略处理，都应该打一个折扣系数。 |
|||
| Bull Case TP | >$100 | ~$100+ (implied) | $110 |
| Bear Case TP | 未明示 | 未明示 | $43 |
观察：UBS 的 2026e 预测（$1.39 EPS）已经明显高于管理层 Q2 指引隐含的全年节奏（$14.3B × 4 ≈ $57.2B，$0.20/Q × 4 ≈ $0.80 EPS）。HSBC 的 $1.10 最接近指引节奏。MS 的 $1.12 保守一些。
按 UBS model 的季度拆分：Q2/Q3/Q4 2026E 营收为 $14.75B / $15.29B / $15.92B，说明 UBS 核心判断不是仅仅 beat 一季，而是认为 server 供给改善 + pricing tailwind + DCAI 欠发货释放会贯穿全年。
解读：UBS 认为 管理层 Q2 指引"implied 2H flat"是保守的，因为 Q1 upside 大超预期 + DCAI 欠发货 $1B+ 应会随供给改善释放。HSBC 主要价格假设还未完全体现在 Q2 指引中。
| 时间 | 事件 | 潜在影响 | 哪家券商最敏感 |
|---|---|---|---|
| 2026-Q2 | Q2 实际业绩 vs 指引 $14.3B / EPS $0.20 | DCAI 增速能否突破 25% QoQ | UBS（预期最高） |
| 2026-Q3 | Q3 指引 + AMD 服务器份额数据 | 印证/打脸 MS "AMD 抢份额" 论 | MS |
| 2026-2H | Intel 14A PDK 1.0 发布 | UBS 认为这是 foundry 签约催化 | UBS (期权假设) |
| 2026-2H / 2027-1H | NVDA / AAPL / Amazon 14A 签约 | 若任一落地 → HSBC 可进一步上调 | 三家都会重估 Foundry 部分 |
| 2027-Q2 | PC TAM 见底 vs CCG revenue | CCG 是否从逆风转顺风 | MS |
| 2027-Q4 | Foundry 能否达到 breakeven 指引 | management 承诺 "7 quarters from now" | MS (最怀疑) |
| 2028-Q1 | Clearwater Forest / Diamond Rapids 量产 | 18A server CPU 完整年度利润贡献 | HSBC 2027/28 ASP +10% 假设 |