SMCI · Super Micro Computer, Inc. · Stock research

Current Thesis

The dilution leg has now fully overwhelmed the order-book leg. Supermicro priced its $7B financing on June 11 — 45,454,545 common shares at $27.50 plus 75,000,000 depositary shares at $50 representing 1/20th of a 7.0% Series A mandatory convertible preferred — and the stock has spent every session since trading toward and then through that clearing price. At $24.18 (July 18), the deal buyers from five weeks ago are underwater by roughly 12%, and the $2B at-the-market program that opened in fiscal Q3 sits on top of that as live, price-insensitive supply. The July 16 AI-hardware unwind took the stock down 6–11% in a session and left it 13.6% under the 20-day SMA ($28.98) and 26.7% under the 200-day ($34.18). The 52-week low of $19.48 is now only ~19% away, versus a 52-week high of $62.36. Underneath the tape, the cost structure is moving the wrong way: DDR contract prices are modeled up 32% quarter-over-quarter in Q3 2026 and a further 18% in Q4, which lands directly in the COGS line of a business that guided Q4 gross margin to 8.2–8.4%. A box assembler with single-digit margins and no pricing power against Dell and HPE does not absorb a 32% memory step-up quietly. The Aug 11 FY26 print is the binary, and the setup going into it is a stock making lower lows under all major averages with a hold-rated sell side.

Bullish and bearish views on Super Micro Computer, Inc.

The model's bull view on Super Micro Computer, Inc. (SMCI), in brief: The order book is contracted and specific. The bear view: Price is below the deal. The offering cleared at $27.50 on June 11; spot is $24.18. Every syndicate account is offside, and the $2B ATM program active from fiscal Q3 supplies stock into any rally without regard to level. Known forward supply of this size caps rallies… Both cases follow in full.

Bull Case

• The order book is contracted and specific. The June 9–11 raise was explicitly sized to buy components against ~$39B of AI-server orders booked from more than 20 customers. That is signed demand, and it is the single reason the equity has not simply broken to the 52-week low outright.
• Product cadence has not slowed. July 15: rear-door heat exchanger portfolio expanded to 10 models spanning 10kW to 120kW per rack — the thermal envelope Rubin-class racks actually require. July 8: Kubernetes Edge AI appliances validated with Red Hat and Everpure, which lifted the stock 3.56% to $27.18 on the day. June 22: Vera Rubin NVL4 blueprint at ISC Hamburg, scaling to 1,152 Rubin GPUs in liquid-cooled racks for H2 2026 deployment.
• Margin direction did improve last print. GAAP gross margin recovered to ~10% from ~6% sequentially in fiscal Q3 2026; adjusted gross margin printed 10.1% against a 6.75% estimate and adjusted EPS $0.84 versus $0.63.
• Sell-side targets sit above spot. Citi raised its target to $33 on July 13 (Asiya Merchant, Hold maintained, 8x FY28 EPS), GF Securities carries a $48 Buy from June 22, and the average 12-month target is ~$37.38 with a $58 Street high — roughly 55% above the July 18 quote.
• The balance sheet risk was pre-funded. Whatever the equity cost, the company now holds roughly $1.22B net from the common offering and ~$3.68B from the preferred, so component procurement against the $39B pipeline is no longer a financing question.

Bear Case

• Price is below the deal. The offering cleared at $27.50 on June 11; spot is $24.18. Every syndicate account is offside, and the $2B ATM program active from fiscal Q3 supplies stock into any rally without regard to level. Known forward supply of this size caps rallies mechanically.
• Share count expansion is structural. The combined common plus mandatory-convertible issuance represents dilution equivalent to roughly 35% of the pre-announcement market cap, with modeled share-count expansion as high as ~38%. The 7.0% preferred coupon is a permanent cash drag on a business earning single-digit gross margins.
• Input costs are inflecting against the model. DDR contract prices modeled +32% QoQ in Q3 2026 and +18% in Q4 hit a company that already guided Q4 gross margin to 8.2–8.4%. The USITC instituted a Section 337 investigation into certain DRAM devices on July 15 — another vector for memory supply friction.
• The last guide was cut. Fiscal Q3 revenue of $10.24B (+123% YoY) still missed by roughly $2.2B (~17.7%); FY26 guidance was reduced to $38.9–40.4B and Q4 framed at $11.0–12.5B. Consensus for the Aug 11 print sits near $0.59 EPS.
• Every bounce has been sold. The June 22 GF upgrade spike to $35.46 round-tripped within four sessions. The July 8 Red Hat pop to $27.18 was gone inside a week. On July 16 the AI-hardware complex sold off and SMCI led the decline.
• Governance discount is unresolved. The 2024 Hindenburg report, E&Y auditor resignation, near-delisting, and charges tied to illegal NVIDIA-chip shipments to China still sit in the multiple. The stock trades at ~8x on Citi's FY28 numbers because the market prices a risk premium, and that premium has not compressed.
• Retail saturation is documented. The GraniteShares leveraged single-stock SMCI ETF launched June 1, within days of the local high near $52 — a durable marker of where retail participation peaked. Leveraged wrappers on a name that then fell 50% amplify the downside flow.

Setup & Price Structure

• Spot $24.18 (July 18), previous close $24.68; session range $23.38–$25.17.
• 20-day SMA $28.98 — price is 13.6% below it. 200-day SMA $34.18 — price is 26.7% below it. No moving average is offering support; all of them are overhead resistance.
• 52-week range $19.48–$62.36. Spot is ~61% off the high and ~24% above the low, down nearly 50% over the trailing year.
• The $27.50 offering price is the first meaningful overhead level and now functions as supply, since it is where the newest large holder base is anchored.
• The June 10 capitulation close of $33.87 and the late-June $30–$31 shelf both failed. Structure is a sequence of lower highs — $52 (June 1), $35.46 (June 22), $27.18 (July 8) — and the July 16–17 break through $25 extended it.
• Volume expanded on the July 16 decline, consistent with distribution rather than exhaustion. There is no higher low, no base, and no volatility contraction to trade against.
• Behaviourally this is a retail-squeeze profile: leveraged-ETF wrapper, scandal history, elevated overnight gap risk. Sizing appropriate to the name is a small probe at most, however attractive the forward multiple screens.

Catalyst Calendar (next 30 days)

• 2026-08-11 (confirmed, after close) — Q4/FY26 results. Consensus ~$0.59 EPS against a Q4 guide of $11.0–12.5B revenue and 8.2–8.4% gross margin. This print is the entire binary: the first quarter that shows whether the $39B order book converts at a margin capable of servicing the new preferred coupon.
• Ongoing through the window — the $2B ATM program is live in fiscal Q3, making share issuance a continuous, undated overhang rather than a single event.
• Undated, in-window — USITC Section 337 DRAM proceeding (instituted July 15) procedural milestones; exclusion-order noise raises memory input risk further.
• No scheduled investor day and no analyst-day catalyst in the window. The stretch between now and Aug 11 is an empty tape dominated by AI-capex sentiment and rate headlines.

Elapsed catalysts

• ~2026-07-22 to 2026-08-05 (est.) — peer prints from the AI-hardware complex (Dell, and the memory names) set the read-through on server margin and DRAM pass-through before SMCI reports. (passed 4d ago)

What Would Change Our Mind

• A weekly close back above $27.50 — reclaiming the offering price would mean the deal book is whole and the ATM is no longer selling into weakness. That is the first structural repair worth respecting.
• A base rather than a bounce. A higher low above the $23–$24 zone followed by a reclaim of the 20-day near $29 would convert a falling knife into a tradable setup. Until a higher low prints, every rally is a supply event.
• Q4 gross margin above 9% with FY27 revenue guided above $50B on Aug 11. That combination would show memory cost pass-through working and reframe the $39B backlog as profitable revenue. A print at or below the 8.2–8.4% guide, with DDR up 32% QoQ, confirms the margin problem is structural.
• An upgrade cycle rather than a lone outlier. Citi at $33 Hold and GF at $48 Buy do not constitute a turn. Two or more firms moving to Buy with targets above $40 would mark sell-side capitulation to the upside.
• Evidence the ATM is finished or unused. A filing showing the $2B program shelved or substantially undrawn removes the largest mechanical cap on the stock.
• What would deepen the bearish read: a weekly close under $19.48 breaks the 52-week low and puts the post-2024 structure in play, with no technical reference beneath it.

Correlation Notes

• Trades as a basket with DELL and HPE on the AI-server-integrator theme. Note the divergence: Dell surged on July 9 and drew attention again July 15 while SMCI faded — relative weakness inside its own cluster points to a company-specific problem layered on a sector move.
• Functions as a high-beta, leveraged proxy on NVDA GB200/GB300 and Rubin rack volume. It captures the upside of GPU shipment growth with almost none of the margin, which is why it falls harder than NVDA on any AI-capex scare; July 16 is the current example.
• Inversely exposed to memory pricing. Rising DRAM/HBM contract prices are a tailwind for MU and the memory complex and a direct margin headwind here. When memory names rally on pricing, the gross-margin math at SMCI worsens.
• The GraniteShares leveraged single-stock wrapper (June 1) increases intraday reflexivity: rebalancing flow amplifies moves in both directions and raises gap risk around the Aug 11 print.
• Theme status: SATURATED, edging toward dead for the AI-server-integrator expression specifically. Mainstream coverage arrived (leveraged ETF launch, retrospective "if you invested $1,000 15 years ago" pieces on July 16), the sell side sits at Hold, and price structure rolled over months ago. The broader AI-datacenter buildout remains alive; this particular vehicle has already had its retail phase.