Super Micro Computer (SMCI) — Deep Dive

Price context: $40.81 (−0.95%) at 2026-09-22T15:09Z [verified: data/smci-1y.json]. Closed Sep 21 at $41.20 (~+5.4% on a chip-sector rally; no company-specific catalyst per mission). 52w 19.48–58.78. Returns: 1w +12.14%, 1mo +10.63%, 3mo +16.19%, 6mo +90.92%, 1y −10.06%. Market cap ~$27B, P/E ~12.1, EPS $3.20, beta 2.19 [index: finnhub, read 2026-09-22].

Summary (SMCI)

A record FY2026 ($39.1B revenue, +78%) with a stunning FQ4 margin spike (non-GAAP GM 17.6% vs ~10% prior quarters) and a record $60B+ backlog produced a huge Q1/FY2027 guide-up — but management itself calls the margin spike partly nonrecurring, expects GM to normalize to ~10.4–10.8% in Q1 FY27, and the business is a cash furnace ($6.8B negative FCF in the 12 months to Mar 2026) that forced a $7B dilutive equity package in June 2026. Wall Street is split: consensus Hold, average target ~$41–43.

1. Earnings reports

FQ4 2026 (reported Aug 11, 2026) [index: businesswire PR; marketbeat call highlights; financefeeds; bingx] - Net sales $11.1B (+93% YoY vs $5.8B; +9% vs $10.2B Q3). Near low end of $11.0–12.5B guide; slightly missed consensus ~$11.3–11.55B. CFO Weigand blamed customer delays tied to power availability, cooling and networking readiness; delayed revenue to be recognized in later quarters; Liang: timing, not demand change. - GAAP GM 17.5% (vs 9.9% Q3, 9.5% year-ago). Non-GAAP GM 17.6% (vs 9.6% year-ago; guide was 8.2–8.4%; prelim Jul 21 was 15–17%). Non-GAAP diluted EPS $1.70 vs consensus ~$0.92–0.94 (~84% beat) and guide $0.65–0.79. GAAP EPS $1.62. Net income $1,178M (vs $483M Q3, $195M year-ago). - Margin drivers (CFO): ~75% of sequential improvement = customer/product mix including deferral of several contracts into FY27; remainder = lower tariff costs and lower inventory reserves — described as "potentially nonrecurring benefits." No tariff rebate was booked (refunds being pursued). - FY2026: record revenue $39.1B (+78% vs $22.0B); non-GAAP EPS $3.63 (+76%); GAAP EPS $3.26 (vs $1.68); FY non-GAAP GM 10.9% (vs 11.2% FY25 — full-year margin FELL); non-GAAP op margin 8.1% (vs 7.1%). - Guidance: Q1 FY2027 revenue $14.5–15.5B (consensus $11.8B — beat by ~$3.2B at midpoint); non-GAAP EPS $1.01–1.10; FY2027 revenue $65–72B reiterated (Wall St ~$52.5–53B — ~$16B above at midpoint). Liang: demand "stronger than ever"; >80% of future revenue AI-related; $60B+ new orders; record backlog; nine customers >$1B annual revenue (vs four a year ago). - Stock: +9–10% after-hours to ~$34.74 (off a $31.60 close) [index: coincentral]; later rallied ~19% in subsequent sessions per bingx.

FQ3 2026 (reported May 5, 2026) [index: coincentral; fool transcript; stocktitan 8-K; barchart] - Revenue $10.2–10.24B (+123% YoY vs $4.6B; down from $12.7B in Q2) — missed consensus ~$12.33–12.39B and the earlier $12.3B guide (component shortages, customer site-readiness delays). Adjusted EPS $0.84 beat $0.62; GAAP EPS $0.72 (guide ≥$0.52). GAAP GM 9.9% (vs 6.3% Q2, 9.6% year-ago); non-GAAP GM 10.1% (vs 6.4% Q2). Net income $483M (+344% YoY). Stock +19% after-hours. - Cash: operating cash flow −$6.6B (driven by $10B AP reduction, $581M inventory build, offset by $2.6B AR reduction); 9-month FY26 operating cash used $7.56B (vs +$796M prior year). As of 3/31/26: cash $1.29B; bank debt + converts $8.8B; total liabilities $15.88B; equity $7.58B. Inventory $11.1B. Net debt $7.5B (vs $787M prior quarter). Cash conversion cycle 106 days (from 54); DOI 106, DSO 85. - Segments: enterprise channel $2.8B (28% of revenue, +46% YoY, +45% QoQ); OEM appliance/large data center $7.4B (72%); AI GPU platforms >80% of revenue; software revenue >$46M (vs <$10M/quarter historically). - Concentration: two customers >10% — a large data center customer at 27%, an enterprise customer at 10%. - Q4 guide given then: revenue $11–12.5B (above $11.07B consensus); non-GAAP EPS $0.65–0.79 (above $0.55); FY26 revenue $38.9–40.4B (below $40.9B consensus); Q4 GM guided 8.2–8.4%. - Regulatory: "Super Micro Computer, Inc. is not a defendant nor a target of a grand jury investigation"; investigations continue; "co-founder smuggling case" did not affect Nvidia/AMD/Intel vendor relationships.

Gross-margin trajectory (the controversy): Non-GAAP GM sat between 6.3% and 10% for six straight quarters, then printed 17.5–17.6% in FQ4 [index: financefeeds]. Management's explanation: mix — enterprise buyers taking full Data Center Building Block Solutions (DCBBS) pay more than hyperscalers; plus deferred contracts, lower tariffs, lower inventory reserves (partly nonrecurring). Bears' counter: FY2026 full-year non-GAAP GM fell to 10.9% from 11.2%, i.e., the full-year mix shift "was worth nothing," and management expects Q1 FY27 GM to normalize to 10.4–10.8% [index: bingx, Sep 15 2026]. One quarter ≠ trend.

2. Business metrics

• Blackwell ramp: Sep 11, 2026 — announced broad availability of NVIDIA Blackwell Ultra solutions: Plug-and-Play HGX B300 systems and GB300 NVL72 racks [index: ainvest]. Record backlog >$60B anchored by next-gen accelerator platforms (Blackwell) [index: fxleaders, Sep 2 2026]. Manufacturing capacity >6,000 high-end racks/month [index: fool transcript, May 2026].
• Data-center/mix: Enterprise + channel customers ~half of recent quarterly revenue; nine customers >$1B/year; $39B in "recent weeks" AI server orders (June) from 20+ customers; DCBBS/software pivot (software $46M+ in Q3 FY26).
• Backlog: Record entering FY2027; $60B+ new orders in FQ4.
• Cash/cash burn: Q4 operating CF +$747M (vs −$6.6B in Q3). But 12 months trailing Mar 2026: −$6.8B free cash flow [index: marketwatch/morningstar, Jun 9 2026]. Cash $1.29B at 3/31/26 vs $8.8B debt+converts.
• Raises/dilution (last 12 months):
• Jun 9, 2026: $7B equity/equity-linked package — $1.25B common stock + $3.75B depositary shares (Series A mandatory convertible preferred, converts ~Jun 1, 2029, Nasdaq: SMCIP) + $2B ATM program starting Q3 2026. Proceeds: component purchases for $39B order pipeline. JPM, Goldman, Citi lead bookrunners. Stock −7% AH, −30% over five days [index: marketwatch; morningstar; techtimes; ainvest].
• Mar 5, 2026: up to $2B convertible senior notes due 2030 (Rule 144A + $300M option), with capped-call hedges and a concurrent $200M buyback; stock −7% [index: stocktwits.com news].
• Also: $700M convertible senior notes (2.25% new notes due 2028; amended notes at 3.50%) [index: kalkinemedia]; and a prior $2B converts (June 2025-era, 100% premium capped call at $81.78, 35% conversion premium at $55.20) [index: ainvest].
• Accounting/governance overhang: EY resigned Oct 2024 after Hindenburg short-seller report; company refiled 10-K/10-Qs, appointed new auditor, regained Nasdaq compliance through 2025–2026; shares had fallen from >$60 toward the high teens [index: gotrade/heygotrade, May 2026]. Q3 FY26 call: not a defendant/target of grand jury investigation. Overhang is diminished but not fully erased — governance was still being cited as a valuation drag in Sep 2026 commentary [index: bingx].

3. Earnings calls & conferences

• Liang (FQ4, Aug 2026): "The demand for our AI IT solutions is even stronger than ever before"; positioned as a "one-stop shop"/"one-stop supplier" for customers building data centers and AI factories; record backlog entering FY27; >80% of future revenue AI-related [index: marketbeat; financefeeds; businesswire].
• Weigand (CFO, FQ4): detailed the margin bridge (~75% mix incl. deferrals; rest tariffs/inventory reserves), flagged nonrecurring elements, no tariff rebate booked.
• Analyst Q&A tone: On the FQ3 call, analysts pressed on margin sustainability and working capital; one older Stocktwits cheatsheet (FQ1 FY26) noted analysts "directly challenged" management on whether gigawatt-scale deployments would ever yield margins (citing the xAI Colossus precedent), with Liang responding vaguely and admitting they'd "control revenue based on cash flow" — i.e., may not fund the guided growth [index: cheatsheets.stocktwits.com]. Mizuho expected Q4 margins to decline ~180bps sequentially on mix even after the Q3 beat [index: newsbreak].

4. Sentiment (public posts/indexed only)

• Retail sentiment broadly bullish but softening: AltIndex sentiment score 68/100 (bullish; down from 30-day average) across Reddit/Stocktwits [index: altindex, read 2026-09-22]. Stocktwits retail sentiment was "extremely bullish" with "extremely high" message volumes after the Q3 print [index: newsbreak, May 2026]. SMCI buzz +5.4 (Sep 2) and +12.8 (Aug 11) on Reddit-signal trackers, "positive momentum" [index: github adanos bot].
• Dominant bullish arguments: record $60B backlog and $39B order intake prove AI-server demand; liquid-cooling (DLC-2) leadership (cuts DC power usage up to 40%) gives a structural edge on Blackwell/GB300 racks; margin recovery to 17.6% shows the mix pivot to enterprise/DCBBS is working; cheap at ~12x earnings vs hyperscaler growth.
• Dominant bearish arguments: margin spike is one quarter of deferral/tariff/inventory timing; management guides GM back to ~10.4–10.8%; cash furnace (−$6.8B FCF) forced a $7B dilutive raise; 106-day cash conversion cycle; two customers = 37% of revenue; Dell/HPE competition compressing pricing; accounting/governance stigma lingers; stock is 39% below its $58.78 52-week high despite the hype.
• tradestie (Sep 2026): mixed 52/100 "bottoming-setup" gauge — price "extended above range," near Wall St target $42.38 [index].

5. Analyst landscape

• Consensus (as of Sep 22, 2026): Hold. Aggregators disagree on counts: 27 analysts / Hold 3.33 with base target $42.38, bull $60, bear $15 [index: techi]; 25 analysts: 0 Strong Buy / 9 Buy / 14 Hold / 2 Sell, avg $41.44, range $35–48 [index: tickergate]; 19 analysts: median $43, range $15–$60, 5 Buy / 11 Hold / 3 Sell [index: tickernerd]; 13 analysts per wallstreetzen: target $36.55, 2 SB / 1 Buy / 8 Hold / 1 Sell / 1 SS [index]; MarketBeat: Hold, $42.13 [index: transcriptdaily].
• Notable bulls: Northland's Nehal Chokshi — Buy, $59 target (reiterated Sep 17, 2026), confident in growth trajectory and Blackwell Ultra lineup [index: ainvest]; Rosenblatt — Buy, raised $45→$51 (Aug 12) [index: transcriptdaily]; GF Securities — upgraded Hold→Buy, $48 target (Jun 22, 2026), arguing dilution was absorbed and backlog converts to H2 revenue [index: marketbeat]; Zacks — upgraded to Strong Buy (Aug 20) [index: transcriptdaily].
• Notable bears/cautious: Bernstein SocGen's Mark Newman — initiated Sep 16, 2026: Market Perform, $46, citing slowing growth, aggressive guidance, weak cash conversion [index: ainvest]; Mizuho — Neutral, expected sequential Q4 margin decline; Citi — Neutral, $31 post-Q3 [index: newsbreak]; Goldman — $30 target referenced in Stocktwits cheatsheet context.

6. Growth drivers vs risks

• Hyperscaler capex: the tailwind behind the $60B backlog; AI GPU platforms >80% of revenue.
• Liquid cooling: DLC-2 leadership is the differentiator bulls cite for Blackwell-density racks.
• Competition: Dell, HPE (and white-box/ODM) pressure pricing — the structural reason margins sat at 6–10% for six quarters.
• Customer concentration: 27% + 10% from two customers (FQ3 FY26).
• Margin compression: the core debate — see §1.
• Accounting credibility: EY resignation/refilings/Nasdaq scare are resolved procedurally but remain a valuation discount factor.
• Dilution: $7B June package (common + mandatory converts + $2B ATM) plus $2B converts in March; converts use capped calls but ATM and mandatory preferred will dilute.
• Macro: 10Y ~4.9%, Fed 3.75–4.00% after the Sep 16 hike — a headwind for a cash-burning, debt-heavy growth name; also supports the "AI capex funded by equity" trend (Alphabet $80B equity raise cited by MarketWatch).

7. Analyst reliability & bias

• Northland (Chokshi): persistently bullish; Buy $59 maintained even as Bernstein initiated cautiously the day before and the stock sat ~$41. Targets have been serially above the stock; accuracy hinges on the margin-recovery thesis proving durable. No evidence of manipulation found — just a growth-optimist stance. Note: Northland also covers INTC bullishly (see Memo 3) — a general pro-cyclical-tech posture, not SMCI-specific bias per se.
• Bernstein's Newman: newly initiated (Sep 16, 2026) — the most substantive bearish initiation, grounded in cash-conversion math rather than price momentum.
• Zacks: upgraded to Strong Buy Aug 20 — quant-driven, typically late-cycle.
• Consensus accuracy: consensus has been poor at SMCI's turning points — targets clustered $25–31 after Q3 (all wrong-footed by the Q4 margin spike), then chased up to $45–59. Wide dispersion ($15–$60) signals low conviction, not informed disagreement. Speculation vs fact: claims that the margin spike is "sustainable" or "one-time" are both estimates; the only hard facts are management's own normalization guide (10.4–10.8%) and the FY26 full-year GM decline.

(a) BULL CASE (SMCI)

1. Demand is real and enormous: $60B+ new orders, record backlog, $65–72B FY27 revenue guide (~$16B above Street). Nine $1B+ customers diversifies concentration.
2. Margin pivot evidence: 17.6% non-GAAP GM in FQ4; enterprise/DCBBS mix and liquid-cooling leadership (Blackwell Ultra, GB300 NVL72) structurally favor higher-margin configurations.
3. Cash flow turned: +$747M operating CF in FQ4 after −$6.6B in FQ3.
4. Valuation: ~12x trailing earnings for a company guiding ~70% revenue growth — cheap if FY27 guide is real.
5. Catalyst path: Q1 FY27 results (Nov 2026) showing $14.5–15.5B revenue conversion of backlog.

(b) BEAR CASE (SMCI)

1. Margin spike likely ephemeral: management attributes ~25%+ to nonrecurring items and guides GM back to 10.4–10.8%; full-year FY26 GM fell YoY.
2. Cash furnace + dilution: −$6.8B FCF (TTM Mar 2026); $7B June raise (common + mandatory converts + $2B ATM) with more ATM selling ahead; net debt $7.5B; 106-day cash conversion cycle.
3. Customer/concentration & competition: 37% of revenue from two customers; Dell/HPE pricing pressure is structural.
4. Governance discount: EY resignation/refiling history keeps a multiple discount; any accounting wobble would be punished severely.
5. Macro: 4.9% 10Y + Fed at 3.75–4.00% punishes cash-burning growth; AI-capex-funded-by-equity raises the bar for everyone.

(c) Probability-weighted view, 3–6 months (SMCI)

Base case (50%): stock chops $35–48 into Q1 FY27 earnings (Nov 2026). The FY27 guide is aggressive; backlog conversion depends on customer site/power readiness — the same factor that pushed FQ4 revenue to the low end of guide. Margins normalize to ~10–11% as guided, which the bull case needs to be wrong about for real upside. Bull case (25%): backlog converts cleanly, enterprise mix sustains GM >13%, stock re-tests $55–60 (Northland's target). Bear case (25%): working-capital strain forces slower revenue recognition or another raise; GM reverts to high-single digits; stock re-tests low $30s (Bernstein/Mizuho zone). Single most important data point: Q1 FY2027 gross margin and the revenue-conversion rate of backlog (Nov 2026 print) — it adjudicates whether FQ4's 17.6% was mix-shift or timing.

Could not verify / open questions (SMCI)

• Whether the $7B June package's ATM ($2B) has been utilized and at what prices (would confirm ongoing dilution).
• The exact customer identity behind the 27% concentration.
• Tariff-refund amounts being pursued (management says none booked yet).