Super Micro Computer

01 / The central investment debate

The opportunity is a large AI-infrastructure revenue ramp valued like a vulnerable hardware supplier. Management guides FY2027 revenue to $65–72bn; accessible consensus is $67.13bn. The provider’s FY2027 EPS estimate is $4.34, apparently adjusted, versus reported FY2026 adjusted EPS of $3.63. Revenue can rise roughly 72% without comparable per-share earnings growth. [S2] [S4]

*Rebuilt using 656.965m common shares at 31 July, June debt/cash and the $4.313bn preferred liquidation claim. The quote feed’s $24.67bn market cap does not reconcile; $23.41bn is a dated filing-based estimate, not a verified live share count. Normalised P/E uses the $2.84 economic EPS basis described below. [S1] [S3] [S9]

What may be mispriced: profitable rack-scale integration and liquid-cooling demand outlasting near-term funding stress. Why the discount may be rational: expensive components inflate sales; customers and suppliers retain bargaining power; capital issuance can transfer growth away from existing owners. Q4’s 17.5% gross margin is not a sensible permanent assumption: the preceding quarter was 9.9%. [S2]

02 / Competitive position

SMCI’s strength is product integration and deployment speed, not ownership of the accelerator economics. One supplier represented 63.1% of FY2026 purchases; the three largest receivable balances represented 52.6% of receivables. A broad AI market does not remove supplier-allocation or customer-credit risk. [S1]

Peer prices: $543.51 Dell and $55.88 HPE, 16 September snapshots. P/E uses adjusted/provider EPS; definitions and fiscal periods differ. Cash figures are not annualised or like-for-like annual yields. [S4] [S7] [S8] [S9]

Dell offers scale and broader distribution; HPE has substantial networking and enterprise exposure. Their mixes justify different margins—not automatic valuation equivalence. Dell’s quarter diluted share count fell about 5%, whereas SMCI’s annual diluted count rose about 11%. For near-term opportunity sizing, the $65–72bn company revenue guide is more useful than an unsupported “AI TAM” multiple. Sustained share gains, financing and deployment capacity—not the headline market size—constrain the bull case. [S1] [S2] [S7] [S8]

03 / Earnings are not distributable cash

Three fiscal years, ending June; growth from FY2024 to FY2026 spans two intervals. Revenue CAGR was 61.4%, versus approximately 30.3% for diluted EPS. Reported FCF means operating cash flow less capital expenditure. [S1] [S2]

Capex was only 0.4% of sales, but the business was not capital-light: year-end inventory reached $12.90bn and receivables $6.13bn. Cash of $7.52bn sat against $8.72bn of financial debt, before preferred claims. The positive Q4 operating cash inflow of $747m is encouraging; it does not undo the annual cash burn. [S1] [S2]

What $100 buys: approximately $9.47 of trailing common accounting profit, but −$29.78 of reported FCF. Common dividends and gross buybacks contributed $0. All common accounting profit was retained; that is the same profit, not an additional return. Gross cash shareholder yield was 0%; net cash shareholder yield was about −6.0% after common equity issuance. Period-end common shares increased 10.6%—a negative share-reduction yield, not a buyback benefit.

Per-$100 figures use current filing-derived common market capitalisation. Reported diluted-EPS yield is separately 9.1% because it uses the annual weighted-average diluted denominator. These are look-through corporate economics, not payouts. SBC was $412m, or roughly 1.8% of common market value. [S1] [S2]

The dilution that matters: $4.313bn of mandatory preferred carries an approximately $302m annual coupon and converts in 2029 into about 131–157m common shares, before other dilution. That is roughly 20–24% of July’s common count. A further $1.25bn ATM programme exists; post-July use is not verified. The model separately captures the notes’ financing costs and net conversion premiums rather than counting the same notes as both extinguished debt and fully issued shares. [S1]

Approximate FY2026 ROIC is 20.6% on average invested capital, but 14.1% on ending capital; incremental NOPAT / incremental capital is about 11.5%. Rapid balance-sheet expansion makes the average-capital headline flattering. These are analytical proxies, not a clean peer ROIC ranking.

04 / Our five-year operating scenarios

The bear combines a guidance miss, weaker pricing and maximum preferred conversion. The base broadly follows near-term Street revenue, then slows to single-digit growth, with a 6.5% terminal operating margin. The bull requires $148bn revenue and durable 9% operating margins—not merely an AI-themed re-rating. No probabilities are assigned.

Flat-multiple test: hold the initial economic P/E at 12.54×. Base returns become 16.8%, so the base case does not depend on multiple expansion. The starting economic EPS of $2.84 equals FY2026 net income less a full preferred coupon, divided by July common shares plus the existing award allowance. This is not the 8.2× provider-EPS valuation.

Economic forecast EPS uses year-end diluted shares, cash-settled/refinanced note principal and net premium dilution after assumed capped-call offsets. It is not a GAAP EPS prediction. Future SBC is treated as equivalent cash compensation, without adding a second ongoing dilution charge; existing award overhang is retained. All common dividends and buybacks are zero. The bear case is not a worst-case loss limit.

05 / The base-case path

The FY2029 EPS dip reflects preferred conversion, not falling revenue. Trade working capital must fall from 36.3% of FY2026 sales to 28% in FY2027 and 23% by FY2031. Even then, FY2027 owner FCF is approximately −$2.16bn. The base remains above a $2bn cash buffer, but assumes existing facilities and $4.725bn of note principal can be refinanced.

Owner FCF = net income − preferred dividends + depreciation − capex − change in trade working capital. SBC stays expensed; other operating balances are held constant. This is not reported CFO. FY2028 visible consensus is $78.76bn revenue / $5.33 provider EPS; the model’s earnings, margins and shares remain our own assumptions. No FY2028 company EPS guide is assumed. [S4]

DCF cross-check: roughly $33, not an unambiguous bargain

A 13% discount rate and 3% terminal growth produce $33.31 per share. A 12–15% discount-rate / 2–3.5% growth envelope spans approximately $25–40; about 79.1% of enterprise value comes from the terminal value. Terminal growth still consumes working capital and capex.

The bridge deducts debt at face value, adds only cash above the operating buffer, subtracts pre-conversion preferred coupons and applies full preferred dilution immediately. It does not subtract the preferred liquidation value again. The P/E scenario is more generous because its terminal earnings valuation is stronger than the DCF’s sustainable-cash valuation. That disagreement is a reason for caution, not a figure to average away.

Full use of the ATM at $35.64 lowers base earnings-driven CAGR to about 14.8% before any benefit from deploying proceeds. Adding both its shares and $1.25bn gross proceeds leaves the DCF near $33.4. Neither calculation asserts that the ATM has already been issued.

06 / Back-solving the 20% hurdle

At 12× earnings, the required $7.39 EPS is about 19.5% above the $6.18 base case. Holding base revenue, shares, tax and interest constant requires approximately 7.7% operating margin, versus 6.5% in the base. A 15× exit reduces the required EPS to $5.91, but then part of the return comes from re-rating.

Cash-based hurdle: at 15× terminal owner FCF, the same 20% target requires approximately $4.79bn of annual common owner FCF, versus $3.57bn in the base. A conventional FCF CAGR from today’s negative reported cash flow is not meaningful. The workbook shows the positive-FY2028 recovery hurdle and the full EPS/FCF × multiple × entry-price sensitivities.

07 / What would confirm or break the thesis?

The audited accounts are available, but one IT general-controls material weakness remained at June 30. On 20 August, the company reported that its board-led investigation found no evidence that current senior management knew of the alleged export-diversion scheme; SMCI was not named in the March indictment. A company-led investigation is not government clearance. [S1] [S6]

Thresholds are our monitoring rules, not management commitments. Potential catalysts are shipments and cash collection, a credible refinancing plan, margin durability and control remediation—not merely another large order announcement. [S2] [S5]

08 / Conclusion and limitations

The stock is cheap on earnings, not demonstrably cheap on sustainable cash. Around $30, the unchanged base case approaches the requested 20% hurdle; at $35.64 it falls short. A lower price caused by weaker margins, failed collections or fresh dilution would require a new model—not blind use of the old entry threshold.

Evidence confidence: relatively high for audited historical numbers; lower for current post-July capitalisation and future order conversion. Underwriting status: conditional watchlist, with medium-to-low forecast confidence. Peer organic growth, financing-adjusted ROIC and comparable annual FCF yields were not independently harmonised. No probability-weighted expected return is claimed. The five fiscal years are an annualised proxy; no intra-quarter stub is modelled. Returns exclude tax, fees and AUD/USD effects.