2026-08-31 · Fiscal year ended 30 June; common shares dated 31 July. Reviewed 15 October.
https://www.sec.gov/Archives/edgar/data/1375365/000137536526000022/smci-20260630.htmHigh-risk watchlist — cash conversion is the gate
The investment decision
The retained base produces 15.7% annualised from $35.78 and needs about $29.80 for a conditional 20% return. Product momentum is real company evidence; incremental sales, margin and collections remain to be demonstrated. The capital structure keeps this from being a simple low-P/E thesis.
Edition update. Added September 23 rack shipments, October 8 board update and October 13 OCP portfolio news. Reviewed FY2026 filing and financing terms. September operating/financing paths retained, with October price replay and new working-capital/ATM stresses. Current post-July capitalisation remains unverified.
01. Integration speed must translate into cash per share
Supermicro integrates servers, racks, networking and cooling for customers building AI infrastructure. Its opportunity is speed and system-level execution: getting a demanding installation configured, cooled and operating reliably. The company does not own the economics of every expensive accelerator it resells. Revenue can therefore grow much faster than the value retained by common shareholders.
The latest financial release guides FY2027 revenue to $65–72bn against FY2026 sales of approximately $39.1bn. That is a substantial operational task involving component purchases, customer acceptance and collection. The Q4 gross margin of 17.5% was much stronger than the preceding quarter’s 9.9%; extrapolating the high quarter indefinitely would make the valuation look easier than the evidence supports. [S2]
The October quote is only slightly above the original report’s $35.64 reference. The main conclusion is consequently unchanged: a low earnings multiple is not enough when growth consumes working capital and financing dilutes existing holders. The retained base illustrates about 15.7% annualised at $35.78, below a 20% hurdle. An investor needs either a more favourable entry or better cash economics, with evidence for the latter.
Our stance is a high-risk watchlist, with cash conversion and capital structure as the decision gates. The upside is that deployment skill and service breadth earn durable integration margins. The downside is that customers, component suppliers and capital providers capture most of the growth while SMCI bears inventory and execution risk. Both are plausible even in a growing AI spending market.
02. What changed since the September report
Three subsequent company developments have been incorporated. On 23 September SMCI announced shipments of NVIDIA Vera Rubin NVL72 racks. On 8 October it announced Shesha Krishnapura’s independent-director appointment and Judy Lin’s retirement. On 13 October it expanded its Open Compute Project portfolio across rack-scale systems, power, networking, cooling and management software. [S4][S5][S6]
These are relevant signals of execution and product breadth, but they do not disclose incremental recognised revenue, realised margin or cash collection. We have not inserted a speculative sales uplift into the model or treated shipping availability as customer payment. The board change likewise does not establish that internal-control deficiencies have been remediated.
The competitive advantage to test is time-to-operation and accountability across a full system. If a more integrated offering reduces deployment friction, customers may pay for a broader solution rather than the lowest-cost server. The competing explanation is that integration becomes a required service in a fiercely competitive market, with little durable premium. Dell, HPE and other system suppliers also bring customer relationships, service reach and financing options.
Supplier concentration reinforces this distinction. The annual filing reports one supplier at 63.1% of purchases. Product breadth does not eliminate dependence on accelerator allocation or pricing. A company can gain shipments while its supplier captures much of the economic rent. That makes margin durability and cash terms better leading indicators of equity value than a count of new product announcements. [S1]
03. The earnings and cash records tell different stories
| Reported annual, USD bn | FY2024 | FY2025 | FY2026 |
|---|---|---|---|
| Revenue | 14.99 | 21.97 | 39.06 |
| Operating income | 1.21 | 1.25 | 2.77 |
| Net income | 1.15 | 1.05 | 2.23 |
| Operating cash flow | -2.49 | 1.66 | -6.81 |
| Cash capex | 0.12 | 0.13 | 0.16 |
| Reported FCF | -2.61 | 1.53 | -6.97 |
These are rounded audited annual figures ending in June; reported FCF equals operating cash less cash capex. Revenue grew much faster than cash available after investment. The chart retains both negative years because averaging them away would hide the funding cycle. The latest quarter’s positive $747m operating cash flow is encouraging, but cannot erase the full-year outflow. [S1][S2]
Low property capex does not make the business capital-light. Servers require expensive components before the customer pays. Inventory, receivables, supplier terms and customer deposits can dominate the cash cycle. Revenue recognised before collection is economically different from revenue that immediately replenishes the cash needed for the next order.
The right earnings-quality test is not to reject accounting profit, but to reconcile it. Track how much profit becomes cash after components, receivables and the operating buffer. Rising profits alongside worsening receivable age or inventory turnover would reduce confidence in both growth and valuation. A cash recovery caused by a temporary inventory unwind is helpful, yet also differs from recurring margin improvement.
The original report’s historical peer comparisons had different fiscal periods and cash definitions. This edition does not roll those old screen multiples forward as current valuations. A meaningful peer comparison would require matching financing-adjusted margins, cash conversion and share claims, not merely declaring the lowest displayed P/E the cheapest company.
04. Preferred stock changes the common-share equation
The June balance sheet had roughly $7.52bn cash and $8.72bn financial debt, before the preferred claim. July 31 common shares were 656.965m. These are dated filing observations, not a verified October capitalisation. The financing programme’s final terms matter: the priced June transaction used a $1.25bn ATM, rather than blindly carrying the earlier $2bn proposal forward. [S1][S3]
Completed mandatory preferred had about $4.313bn liquidation value and a 7% coupon, approximately $302m annually. Conversion in 2029 adds roughly 131–157m common shares depending on the contractual conversion price, before other dilution. The model’s economic share counts include the preferred mechanics and existing award allowance. A low P/E based on an earlier weighted-average share count understates the future common claim. [S1][S3]
The retained model treats note principal as cash-settled or refinanced, and captures net conversion premiums after assumed capped-call offsets. It does not both eliminate debt and add the full principal-equivalent common shares. Future SBC stays expensed as cash-equivalent labour cost without an additional recurring share grant charge; existing award overhang is retained. These conventions are economic approximations, not a prediction of GAAP diluted EPS.
Post-July ATM usage and a fully current share count were not verified. Rather than assert the programme is unused or already issued, this edition adds a clearly hypothetical full-ATM stress. Refinancing access also remains assumed. A failure to refinance note maturities or maintain customer/supplier confidence can force terms worse than those in any of the three operating cases.
05. The base needs working capital to become more efficient
| September house model | FY2027 | FY2028 | FY2029 | FY2030 | FY2031 |
|---|---|---|---|---|---|
| Revenue, $bn | 67.13 | 78.76 | 89 | 98 | 105 |
| Operating margin | 6.0% | 6.3% | 6.5% | 6.5% | 6.5% |
| Economic shares, m | 678.7 | 678.7 | 809.4 | 809.4 | 809.4 |
| Economic EPS, $ | 3.94 | 5.03 | 4.97 | 5.74 | 6.18 |
| Common owner FCF, $bn | -2.16 | 1.5 | 2.43 | 3.13 | 3.57 |
This path is retained from the September 16 house model after reviewing the annual filing and subsequent company updates. It is not a newly downloaded consensus curve. The October calculator reproduces the published rounded valuation inputs and adds funding stresses; it does not recreate the entire original financing workbook. The model vintage is part of the investment information, not hidden in a footnote. [S7]
The base revenue assumption begins within the company’s FY2027 range, then growth slows as scale increases. Terminal operating margin is 6.5%, below the exceptional Q4 level. The dip in FY2029 economic EPS reflects preferred conversion even while revenue rises. This is precisely why total sales growth alone can exaggerate the benefit to existing holders.
The owner-cash definition is net income less preferred cash coupons, plus depreciation, less capex and the change in trade working capital. SBC remains expensed; other operating balances are held constant. It is not reported cash from operations. The retained forecast assumes trade working capital falls from 36.3% of sales in FY2026 to 28% in FY2027 and 23% by FY2031. That efficiency improvement is a substantive underwriting assumption, not an automatic consequence of faster growth.
06. Wide outcomes, with no implied probabilities
| House case | FY31 revenue, $bn | FY31 op margin | Economic shares, m | Economic EPS | Exit P/E | 5Y CAGR |
|---|---|---|---|---|---|---|
| Bear | 56 | 4.5% | 835.5 | $1.73 | 9x | -15.3% |
| Base | 105 | 6.5% | 809.4 | $6.18 | 12x | 15.7% |
| Bull | 148 | 9.0% | 827.7 | $12.44 | 15x | 39.1% |
Bear case: five-year total-wealth CAGR is -15.3%, using FY2031 economic EPS of $1.73, a 9x exit P/E and $15.57 terminal share value from a $35.78 reference price.
Base case: five-year total-wealth CAGR is 15.7%, using FY2031 economic EPS of $6.18, a 12x exit P/E and $74.16 terminal share value from a $35.78 reference price.
Bull case: five-year total-wealth CAGR is 39.1%, using FY2031 economic EPS of $12.44, a 15x exit P/E and $186.60 terminal share value from a $35.78 reference price.
No-rerating base case: five-year total-wealth CAGR is 16.8%, holding the current economic P/E of 12.60x based on $2.84 coupon-normalised starting EPS and $6.18 terminal economic EPS.
The starting $2.84 economic EPS normalises a full preferred coupon and an existing award allowance rather than using a provider’s adjusted EPS. The base 12 times exit is slightly below the current economic multiple, so the base does not rely on rerating. Its difficult assumption is cash-funded execution with dilution contained, not a higher P/E.
CAGR means ((terminal share value plus cumulative cash distributions) / entry price)^(1/5) minus one. Common dividends and buybacks are zero. The five fiscal-year observations approximate a five-year holding period without a precise quarterly stub. No scenario probabilities are assigned. The bear case is not a total-loss bound, and dilution, loss of financing or adverse legal outcomes could produce worse results.
Terminal prices are recomputed from the earlier published rounded EPS, giving a $74.16 base rather than the old $74.19 unrounded-model output. The small rounding difference is disclosed; it is not presented as a change in business outlook.
07. The funding sensitivity is larger than the capex line suggests
| FY2027 trade WC / sales | Extra cash use versus 28% | Common owner FCF |
|---|---|---|
| 28% | $0.00bn | $-2.16bn |
| 30% | $1.34bn | $-3.50bn |
| 32% | $2.69bn | $-4.85bn |
| 34% | $4.03bn | $-6.19bn |
At the fixed $67.13bn revenue assumption, each two percentage points of additional working-capital intensity consumes about $1.34bn. A 32% ratio makes the base $2.16bn owner-cash outflow roughly $4.85bn. This isolates working capital: profits, capex, opening balances and financing terms are held fixed. It is not a prediction that all those factors remain unchanged in a real stress.
Separately, hypothetical use of the full $1.25bn ATM at $35.78 adds 34.9m shares. Holding terminal earnings constant and assuming no benefit from the proceeds reduces base EPS to $5.92 and the five-year CAGR to 14.7%. Fees are excluded. The exercise quantifies sensitivity; it does not assert that this stock has been issued.
Funding shocks can be correlated. A supplier demanding faster payment can raise cash needs just as a customer delays acceptance and a weaker share price increases dilution per dollar raised. The clean sensitivity tables deliberately show one variable at a time, but the investment downside can combine them. The most important next disclosure is therefore a credible bridge from orders to collected cash and the financing needed along the way.
08. The 20% threshold requires more than the retained base
The conditional 20% entry ceiling is $29.80, equal to the $74.16 base terminal share value divided by 1.20^5. It holds the September operating, refinancing, preferred-conversion and share-count path fixed; weaker cash conversion or extra issuance requires a new valuation.
At $35.78 the investor needs $89.03 terminal wealth to compound at 20%. At a 12x exit, that requires $7.42 economic EPS versus $6.18 in the base. On a separate 15x terminal owner-cash basis it requires about $4.80bn annual common owner cash, versus the retained $3.57bn. These are alternative valuation lenses, not additive return streams.
The old $33.31 owner-cash DCF, using a 13% discount rate and 3% terminal growth, remains a September-vintage cross-check and has not been rebuilt in this edition. It was heavily terminal-dependent and required working-capital efficiency and refinancing. We do not relabel it current fair value merely because the quote is updated. [S7]
The website’s price control changes the entry denominator while holding terminal outcomes fixed. It does not rerun the price at which future equity is sold, preferred conversion or lender terms. A drop below the conditional entry ceiling caused by worsening cash conversion requires a fresh model. PEG is not decisive: a low growth-adjusted P/E misses the capital and dilution needed to fund that growth.
09. What would strengthen or invalidate the thesis
| Gate | Evidence to demand | Failure signal |
|---|---|---|
| Execution | Delivery and collection against the $65–72bn FY27 guide | Announcements grow while revenue acceptance or collections lag |
| Margins | Broadly 6% operating margin with a sustainable customer mix | Two quarters below 10% gross margin without a credible bridge |
| Working capital | Trade WC intensity trends toward the 28% base | Persistent ratio above 32% or deteriorating receivables |
| Funding | Current share count, ATM usage, preferred coupons and note refinancing | Unexpected dilution, expensive refinancing or reduced liquidity |
| Controls | Documented remediation and effective oversight | New material findings or customer/export restrictions |
These thresholds are house monitoring rules. The June filing still identified an IT general-controls material weakness. The August company-led investigation reported findings favourable to current senior management, but that announcement is not government clearance. The October board appointment is an event to monitor, not proof that all control or compliance risks are resolved. [S1][S5][S9]
Current institutional flows, short interest and options positioning were not independently verified, so no squeeze or crowding thesis is presented. Financial actuals stop at June 30 and the verified common count at July 31; later product disclosures are included through October 13. The reference quote is the October 14 close. Returns exclude tax, fees and currency effects. The investor decision remains conditional: durable operating cash and a credible common-share funding bridge must catch up with the revenue story.
Financial exhibits
Exact observations accompany every chart. Actuals, management guidance and Capital Thesis assumptions are labelled separately; the chart shapes do not imply probabilities.
Accounting profits coexist with large cash outflows
The annual cash record is volatile because growth requires working capital.
| Observation | Net income | Operating cash flow | Reported FCF |
|---|---|---|---|
| FY2024 | 1.15 | -2.49 | -2.61 |
| FY2025 | 1.05 | 1.66 | 1.53 |
| FY2026 | 2.23 | -6.81 | -6.97 |
Rounded audited actuals, fiscal years ending June. FCF=CFO less cash capex; negative values retained. Sources: [S1] [S2]
The retained base still burns owner cash in the first year
Cash recovery relies on declining working-capital intensity, not revenue alone.
| Observation | Common owner FCF |
|---|---|
| FY2027E | -2.16 |
| FY2028E | 1.5 |
| FY2029E | 2.43 |
| FY2030E | 3.13 |
| FY2031E | 3.57 |
September house forecast retained. Owner FCF is after preferred coupons and cash-equivalent SBC, not reported CFO. Sources: [S7]
A few percentage points of working capital change the funding need
At fixed revenue and profit, 32% trade WC intensity adds approximately $2.69bn cash use versus the 28% base.
| Observation | FY2027 owner FCF |
|---|---|
| 28% WC / sales | -2.16 |
| 30% WC / sales | -3.5026 |
| 32% WC / sales | -4.8452 |
| 34% WC / sales | -6.1878 |
October one-variable house sensitivity on retained $67.13bn FY27 sales; all other operating/funding assumptions fixed. Not four future time periods. Sources: [S7]
Source register and research basis
Evidence checked for this edition on 15 October 2026. Dates below identify source vintages. Company guidance and house scenarios are conditional; source links and the downloadable model allow the assumptions to be inspected.
2026-08-11 · Reported financials and FY2027 guidance; annual filing takes precedence.
https://ir.supermicro.com/news/news-details/2026/Supermicro-Announces-Fourth-Quarter-and-Full-Fiscal-Year-2026-Financial-Results/default.aspx2026-06-11 · Final priced transaction terms; subsequent 10-K confirms completed amounts.
https://ir.supermicro.com/news/news-details/2026/Supermicro-Announces-Pricing-of-Equity-and-Equity-Linked-Financing-Transactions-To-Fund-AI-Orders/default.aspx2026-09-23 · Company product/shipping announcement, not financial revenue disclosure.
https://www.supermicro.com/en/pressreleases/supermicro-now-shipping-nvidia-vera-rubin-nvl72-racks2026-10-08 · Board appointment and retirement announcement; not evidence controls were remediated.
https://www.supermicro.com/en/pressreleases/supermicro-appoints-shesha-krishnapura-board-directors-independent-director2026-10-13 · Product portfolio update, not quantified incremental sales or profit.
https://www.supermicro.com/en/pressreleases/supermicro-expands-integrated-open-compute-project-portfolio-accelerate-rack-scale-ai2026-09-16 · Archived house operating, dilution and owner-cash scenarios retained with explicit vintage.
https://capital-thesis.pages.dev/research/smci-2026-09-16-full-c39932f132ef44f92fc02026-10-14 · Public-provider close $35.78 at 20:00 UTC, captured 15 October; may be delayed.
https://stockanalysis.com/stocks/smci/statistics/2026-08-20 · Company board-investigation findings; not government clearance.
https://ir.supermicro.com/news/news-details/2026/Supermicro-Announces-Completion-of-Independent-Investigation-and-Continued-Enhancement-of-Export-Compliance-Program/default.aspx