SMCI · US
SMCI.US: Q4 FY26 preliminary — order explosion & conversion risk
Super Micro issued a preliminary Q4 FY2026 update on 2026-07-21: revenue estimated near the low end of guide ~$11B (prior $11–12.5B), GM estimated 15–17% (well above prior 8.2–8.4% guide). Q4 new orders >$60B with record backlog, but deliveries span future quarters. Market cap ~$19–20B vs a massive order book → conversion and execution risk. AI server / liquid-cooling building blocks thesis is clear, but historical accounting/governance overhang remains. Formal Q4 print on 2026-08-11. See CLS, NVDA, Vertiv, AVGO.
Cite a section with a deep link, e.g. /en/r/smci-us-research#thesis
Q4 FY26 preliminary snapshot
- Q4 revenue (est.)
- USD 11.0B
- GM (est.)
- 0
- Q4 new orders
- USD 60.0B
- Market cap (approx.)
- USD 19.5B
Guide vs preliminary
- Prior revenue guide midpoint
- USD 11.8B
- Prior GM guide
- 0.1
- Formal earnings date
- 2026-08-11
- Backlog status
- record
Thesis
Super Micro (SMCI) sits in the assembly and system integration layer of the AI server chain—integrating NVIDIA GPUs, Broadcom networking, and liquid-cooling components into deliverable AI racks. The 2026-07-21 preliminary update reveals a striking contrast: Q4 new orders >$60B vs market cap ~$19–20B.
Dual-sided investment logic:
- Order strength — >$60B Q4 bookings and record backlog validate real AI server demand; GM estimated 15–17% (vs prior 8.2–8.4% guide) reflects favorable mix (liquid cooling / high-config racks);
- Conversion risk — orders ≠ revenue; $60B+ backlog converts over future quarters, dependent on capacity, GPU allocation (NVDA), and liquid-cooling supply (Vertiv);
- Governance overhang — 2024 audit delay and restatement history remain valuation discounts; the 8/11 formal print is the validation window.
Versus Celestica (CLS): SMCI wins on speed, customization, and Direct Liquid Cooling; CLS wins on large-scale ODM and Cisco-aligned enterprise. They are the primary comps in AI server assembly.
Business
Model: AI server/rack assembly and system integration including liquid cooling.
Flagships: GPU servers / SuperCluster; DLC; edge/storage base.
Competitiveness: Customization speed + DLC; faster AI rack delivery vs classic OEMs.
Strategy: Digest >$60B bookings / record backlog; formal guide and controls are validation points.
Value chain
Mid/downstream system integration between GPUs and powered racks.
Upstream: NVDA GPUs, networking, cooling parts. Downstream: hyperscalers / sovereign AI. Risks: GPU allocation; governance overhang on customer adoption.
Valuation
SMCI's valuation frame centers on the order book / market cap ratio—>$60B orders vs ~$20B mkt cap implies 3x+ theoretical revenue pipeline, but conversion rate and GM sustainability are discount sources.
Valuation frame
| Metric | Status | Read |
|---|---|---|
| 订单/booking | 爆发 | Q4 新订单 >$60B；record backlog |
| Q4 营收 | 指引低端 | 预估 ~$11B vs 指引 $11–12.5B |
| 毛利率 | 超预期 | 预估 15–17% vs 原指引 8.2–8.4% |
| 市值 | ~$19–20B | 相对订单簿显著偏低 → 转化风险 |
| 治理/会计 | 历史 overhang | 2024 审计延迟；需持续验证 |
| 催化剂 | 临近 | 2026-08-11 正式 Q4 财报 |
Key valuation questions:
- Revenue conversion slope: Q4 ~$11B (low end of guide) vs >$60B orders—backlog needs 5–6+ quarters to fully convert;
- GM sustainability: 15–17% preliminary vs historical ~8–10% band—is this mix improvement or one-off?
- Governance discount: post-2024 audit crisis, SMCI still carries significant governance discount vs CLS / Dell / HPE.
Financial trend (~24 months)
Eight-quarter revenue and gross margin with YoY and QoQ. Semi/AI hardware seasonality is secondary to supply, ASP, and order timing.
Revenue · last 8 quarters
Interactive chart available in the reader.
Gross margin · last 8 quarters
Interactive chart available in the reader.
Operations
Q4 FY26 preliminary operations
Q4 FY26 preliminary data
| Item | Preliminary est. | Note |
|---|---|---|
| Q4 营收（预估） | ~$11.0B | 接近指引低端 $11–12.5B |
| 毛利率（预估） | 15–17% | 优于原指引 8.2–8.4%；mix 改善 |
| Q4 新订单 | >$60B | record；跨季度交付 |
| Backlog | record | 未来数季逐步确认营收 |
| 财季 | Q4 FY2026 | 截至 2026-06-30 |
| 更新日期 | 2026-07-21 | preliminary update |
Segment mix (illustrative)
| Segment | Mix / trait | Watchpoint |
|---|---|---|
| GPU 服务器 | 主导 | NVDA H100/B200 平台 |
| 液冷系统 | 增长 | vs 风冷；ASP 更高 |
| Storage / Edge | 次要 | 非 AI 叙事核心 |
| Hyperscaler | 大客户 | MSFT / AMZN / GOOGL / META |
| Enterprise / Sovereign | 上升 | 主权 AI 项目 booking |
| San Jose 总部 + 全球制造 | — | 美国 + 台湾/马来西亚产能 |
Preliminary read:
- Revenue ~$11B near low end of guide—not a beat narrative, but GM well above guide (15–17% vs 8.2–8.4%) is the positive surprise;
- New orders >$60B is the headline—needs decomposition: repeat vs new hyperscaler / sovereign AI wins;
- Record backlog implies FY2027 revenue path may be significantly raised—but only confirmable on 8/11 formal guidance.
Orders & backlog
SMCI's Q4 2026 order dynamics are core to the thesis:
| Metric | Value | Meaning |
|---|---|---|
| Q4 new orders | >$60B | Real AI server demand strength |
| Backlog | Record | Multi-quarter revenue coverage |
| Market cap | ~$19–20B | Significantly below order book |
| Conversion cycle | Multi-quarter | GPU allocation + capacity constraints |
"Orders >> market cap" logic: if SMCI sustains 15%+ GM and converts backlog within 4–6 quarters, current valuation has significant upside. Conversely, if GPU allocation is constrained or GM reverts to 8–10%, the order book is "paper wealth."
Key dependencies: NVIDIA GPU allocation (who gets Blackwell B200 priority?), liquid-cooling supply (Vertiv CDU / CLS liquid racks), SMCI's own US/Taiwan/Malaysia manufacturing capacity.
Products & liquid cooling
SMCI's product positioning:
- GPU servers — 1U/2U/4U to full-rack AI clusters; NVIDIA HGX / NVL platforms are core;
- Direct Liquid Cooling (DLC) — at B200 ~1000W+ power density, air cooling is infeasible; DLC is mandatory; SMCI's DLC capability is a differentiator;
- SuperCluster — turnkey AI datacenter solutions including compute + networking + cooling + software stack;
- Edge / Storage — non-AI core but provides base revenue.
Liquid cooling is the key SMCI vs CLS battleground—both expanding DLC capacity, but SMCI leads on "speed + customization" while CLS leads on "scale + yield."
Competition
Strategy / competition / risk map
| Category | Detail | Implication |
|---|---|---|
| AI 服务器 | GPU 机柜 / 液冷 rack 级交付 | NVDA Blackwell 放量直接受益 |
| 液冷 | DLC / CDU 系统 | 高功率密度集群必需；vs CLS 竞争 |
| 订单转化 | >$60B Q4 booking | 产能/供应链决定 revenue 斜率 |
| 治理 | 2024 审计延迟 / 重述历史 | 估值折价来源；需 8/11 验证 |
| 竞争 | CLS、Dell、HPE、Foxconn | SMCI 以速度 + 定制见长 |
| 供应链 | GPU 分配 + 液冷组件 | 瓶颈不在设计而在交付能力 |
SMCI vs CLS comparison:
| Axis | SMCI | CLS |
|---|---|---|
| Positioning | Speed + custom AI racks | Large-scale ODM + enterprise |
| Liquid cooling | DLC leadership narrative | Liquid rack ramping |
| Customers | Hyperscaler + Sovereign AI | Hyperscaler + Cisco enterprise |
| Governance | 2024 audit crisis overhang | Relatively clean |
| Valuation | Order book / low mkt cap | Steadier but lower growth |
SMCI's 2024 audit delay (delayed 10-K), independent investigation, and CFO change caused major stock volatility—governance issues largely repaired in 2025 but the market still carries a discount. The 8/11 formal print's audit opinion and internal controls assessment are critical validation points.
Peer comparison
| Company | Share / role | GM | Strength | Weakness |
|---|---|---|---|---|
| Super Micro (self) | AI server speed/custom | ~15–17% prelim | DLC + rack speed | Governance overhang |
| Celestica | ODM scale | Adj OM ~8% | Scale + yield | Lower absolute GM |
| Dell / OEM peers | Enterprise AI servers | Server mid-teens | Brand + channel | Slower custom racks |
Management
Founder-CEO Charles Liang remains; CFO/controls bench stabilized after 2024 audit crisis. Governance events are the main 24m change. Stability: still transitional — watch audit opinion on formal filing.
Key management (24m)
| Role | Name | Since | 24m change |
|---|---|---|---|
| CEO / President | Charles Liang | 1993 | No change (founder-led) |
| CFO | David Weigand | 2021 | CFO bench stabilized post-2024 audit crisis |
| Chair / CEO | Charles Liang | 1993 | Combined roles; governance discount remains |
Outlook
SMCI's Q4 2026 order dynamics are core to the thesis:
| Metric | Value | Meaning |
|---|---|---|
| Q4 new orders | >$60B | Real AI server demand strength |
| Backlog | Record | Multi-quarter revenue coverage |
| Market cap | ~$19–20B | Significantly below order book |
| Conversion cycle | Multi-quarter | GPU allocation + capacity constraints |
"Orders >> market cap" logic: if SMCI sustains 15%+ GM and converts backlog within 4–6 quarters, current valuation has significant upside. Conversely, if GPU allocation is constrained or GM reverts to 8–10%, the order book is "paper wealth."
Key dependencies: NVIDIA GPU allocation (who gets Blackwell B200 priority?), liquid-cooling supply (Vertiv CDU / CLS liquid racks), SMCI's own US/Taiwan/Malaysia manufacturing capacity.
Scenarios
| Scenario | Conditions | Implication |
|---|---|---|
| Bull | 8/11 confirms GM 15%+; FY27 guide >$50B rev; fast backlog conversion | Order book narrative delivers; doubling potential |
| Base | 8/11 roughly confirms preliminary; FY27 guide $40–45B | Steady growth; hold for conversion |
| Bear | GM reverts to 8–10%; GPU allocation constrained; new governance issues | Order book ≠ revenue; -30–40% possible |
Positioning frame (research only): AI server satellite allocation; avoid aggressive adds before the 8/11 formal print. Preliminary data is a positive signal but not final.