20 October 2026 · Primary preliminary unaudited ranges, drivers, limitations and scheduled results call.
https://www.supermicro.com/en/pressreleases/supermicro-provides-first-quarter-fiscal-year-2027-preliminary-business-updateWait for proof · conditional opportunity
The investment decision
The update strengthens the demand case, but cash collection and dilution decide the investment outcome. At the frozen $37.60 after-hours observation, the base scenario returns 17.9% a year over five years. A $34.39 entry clears a 20% mathematical hurdle only if the stated operating and financing path holds.
Edition update. Fresh 21 October post-catalyst model replaces the prior edition’s retained September paths. Market freeze: 20 October, 7:41 pm EDT after-hours $37.60; regular close $32.47. No unreported September cash or current issuance is invented.
01. Demand improved. The cash-conversion test did not disappear.
Research stance: wait for proof / selective watchlist. The 20 October preliminary update is real evidence of stronger deliveries and a richer mix. It raises our operating opportunity, but it does not establish that receivables turn into cash quickly enough to finance the next wave. At the frozen after-hours price of $37.60, our conditional base produces 17.9% a year over five years, below a 20% hurdle. The illustrative downside is severe. A lower price helps only if the operating and financing assumptions survive. [S1] [S4]
The potential mispricing is not simply “AI grows.” The investable question is whether an integrator can turn surging rack demand into sustainable profit per common share after expensive components, working capital, preferred dividends and dilution. The market reaction has already incorporated a major revenue surprise. There is no verified post-announcement consensus estimate set in this report; it would be misleading to call our new assumptions a consensus beat.
Evidence confidence is high for the disclosed preliminary ranges and audited fiscal 2026 statements; underwriting readiness is moderate. September cash, final earnings, shipment acceptance, current share issuance and full-year guidance remain missing. The next hard evidence event is the 3 November results release/call. We would become more constructive if margins hold near 12% while receivables and inventory release cash, and less constructive if financing needs keep rising despite accounting earnings. [S1] [S2]
02. The surprise is larger gross profit, not just a bigger sales number
| Q1 FY27, ended 30 September | 11 August guidance | 20 October preliminary |
|---|---|---|
| Revenue | $14.5–15.5bn | $18.5–19.5bn |
| Gross margin | 10.4–10.8% | 12.0–12.5% |
| Revenue midpoint | $15.0bn | $19.0bn |
| Gross-margin midpoint | 10.60% | 12.25% |
| Illustrative midpoint gross profit | $1.590bn | $2.328bn |
Revenue midpoint increases 26.7%. Multiplying each revenue midpoint by its corresponding gross-margin midpoint gives a 46.4% increase in illustrative gross profit, or about $738m. This is our arithmetic, not a separately guided profit figure, and it is before operating costs, financing, taxes and preferred claims. Management attributes the sales improvement mainly to AI-platform demand and the margin improvement mainly to customer and product mix. [S1] [S3]
The update is preliminary and unaudited; the auditor has not performed procedures on these preliminary numbers. It contains no new EPS, operating-cash-flow or balance-sheet guidance and does not raise the previously issued $65–72bn FY27 revenue range. The underlying quarter has already ended, so this is a preliminary result update rather than a new multi-year contract or a newly booked order. [S1] [S3]
The frozen market observation is $37.60 at 7:41 pm EDT on 20 October 2026, versus the $32.47 regular-session close at 4 pm: a 15.8% after-hours rise. In Sydney, the snapshot is 10:41 am on 21 October. It is not a final closing price or a guarantee of an executable quote. Earlier supplied news screenshots captured smaller moves at earlier times; those observations do not conflict. [S4]
03. The product is a working AI rack, and the bottleneck moves
Supermicro combines processors and accelerators with boards, chassis, memory, networking, power, cooling and integration. Customers buy systems that can be deployed, not merely a box of chips. Modular designs, fast platform qualification and liquid-cooling integration can shorten the path from a new accelerator to a functioning data centre. Its DCBBS offering broadens the scope from a server to more of the deployment. These capabilities support demand, but do not give it the economics of the underlying chip designer. [S2] [S6]
Customer value comes from time to operation, workload performance and total power/cooling cost. Shareholder value comes from the spread retained after purchased components, assembly, support, warranty and financing. More expensive accelerators can inflate revenue without a proportional increase in retained margin. Power availability, networking, buildings and customer acceptance can delay revenue or collection even after hardware has been bought. This is why a backlog headline is not cash.
The August release described more than $60bn of new orders over the preceding year and a record backlog. Neither that wording nor the supplied broker excerpts establishes that all orders are incremental to the sales forecast, non-cancellable, fully funded, or immediately deliverable. Do not add the order number to expected revenue. The 23 September Vera Rubin shipping announcement provides product-readiness evidence, not a quantified customer contract. [S3] [S6]
04. A guidance beat can still be a sequential margin decline
The new gross-margin midpoint is 12.25%, comfortably above the old 10.6% guide. It remains below the June quarter’s 17.5% actual GAAP gross margin. The June quarter was unusually favourable: the supplied August bank notes attribute much of the uplift to customer/product mix, with the rest linked to lower tariff costs and inventory reserves. Those notes are dated views, not new October management guidance. [S1] [S3] [S7] [S8]
An AI-heavy shipment mix can carry lower percentage margins because costly purchased components dominate the bill. A larger enterprise contribution or a better customer mix can improve the percentage, while lower reserves can help one quarter without repeating. Our base terminal operating margin is 7%, close to fiscal 2026’s roughly 7.1% GAAP outcome; it does not extrapolate the June gross-margin spike. Operating margin is after operating expenses and must never be substituted for gross margin. [S2]
For a stockholder, the evidence to watch is gross profit dollars alongside margin, operating costs and cash collection. A lower margin can still create more earnings on much larger sales; a high margin can still fail to create cash when inventory and invoices grow faster. A one-percentage-point operating-margin change on our $117bn terminal sales assumption changes pretax profit by $1.17bn.
05. The profit-to-cash gap is the central financial fact
| USD bn, fiscal years ended June | FY2024 actual | FY2025 actual | FY2026 actual |
|---|---|---|---|
| Revenue | 14.989 | 21.972 | 39.063 |
| Net income | 1.153 | 1.049 | 2.230 |
| Operating cash flow | −2.486 | 1.660 | −6.810 |
| Cash capital expenditure | 0.124 | 0.127 | 0.162 |
| Free cash flow: CFO less capex | −2.610 | 1.532 | −6.972 |
Fiscal 2026 earned $2.230bn of net income but used $6.810bn of operating cash. The cash-flow statement records $8.877bn of inventory absorption and $3.922bn from receivables. Customer advances/deferred revenue supplied $1.881bn and payables supplied $963m, offsetting part of the drain. Other noncash and operating movements reconcile the remainder. These are annual cash-flow movements; the $12.896bn closing inventory balance is a different measure. [S2]
The mechanism is simple: buy GPUs and other components, build systems, deliver or obtain acceptance, then collect. Revenue and earnings can be recorded before customer cash arrives. If customers pay deposits or suppliers extend terms, they fund part of the cycle. If collection slows, the balance sheet must fund more. This cash gap can reflect rapid growth rather than worthless inventory, but investors still bear obsolescence, cancellation and financing risk.
The June quarter generated $747m of operating cash. That is encouraging, but one positive quarter does not erase a $6.81bn annual outflow. Conventional free cash flow is CFO less cash capex and is not the same as owner cash after preferred dividends. Net income is also not identical to income attributable to common shares; the preferred securities participate in the earnings calculation. [S2] [S3]
06. Growth needs capital, and common shareholders are not the only claim
| 30 June 2026, unless indicated | Amount | Interpretation |
|---|---|---|
| Cash | $7.521bn | Liquidity before the next growth cycle |
| Loans and credit facilities | $4.056bn | About $2.040bn current |
| Convertible notes, carrying value | $4.664bn | $4.725bn face value in the model |
| Total financial debt, carrying value | $8.720bn | Excludes preferred equity and leases |
| Mandatory preferred, liquidation preference | $4.3125bn | 7% coupon; expected June 2029 conversion |
| Preferred annual dividend run rate | $301.875m | Cash or shares under the terms |
| Common shares, 31 July | 656.965m | Dated count, not verified October count |
Fiscal 2026 financing activities provided $9.479bn. Equity offerings and borrowing replenished the cash balance while operations consumed cash. A large cash balance is therefore not proof that the underlying business has self-funded its growth. Debt carrying value differs from contractual face value because of accounting adjustments; the DCF deducts face principal conservatively. [S2]
The mandatory preferred converts into roughly 131–157m common shares, depending on the contractual price formula. The dividend costs about $302m per full year until conversion. Payment in shares would reduce cash use but add dilution. It is not ordinary common debt, yet it is a real prior claim on value. Our model pays the coupon in cash and then includes conversion shares; it does not deduct preferred principal again from equity value. [S2]
The $1.25bn ATM programme is authorisation, not evidence that the entire amount has been issued. A full hypothetical issue at $37.60 would add 33.24m shares before fees. Current usage after the latest reported capital data is unverified. Convertible-note principal is assumed refinanced or settled at par; conversion premiums are modelled as shares, net of applicable capped-call protection. Financing access and counterparties must actually perform. [S2]
07. The failure modes are operational, financial and institutional
Concentration and bargaining power. One supplier accounted for 63.1% of FY26 purchases. Three customers represented 23.0%, 17.1% and 12.5% of June receivables. Those receivable percentages are not revenue shares. A delayed large customer payment, a component allocation change or stronger pricing from a supplier can weaken cash and margins simultaneously. Competitors and customers can also capture the benefits of a booming market. [S2]
Controls and export risk. The 10-K filed 31 August says internal control over financial reporting and disclosure controls were not effective at 30 June because of a material weakness. It also says government investigations and inquiries remain ongoing and are not bound by the completed internal investigation. The company’s review reported no evidence that current senior management knew of the alleged diversion scheme. That is the company’s reported review conclusion, not government clearance. No subsequent government closure was verified for this edition. [S2]
The appointment of a new CFO announced on 15 October is a potential execution improvement, not proof that controls have been remediated. A filed remediation conclusion, more reliable cash forecasting and consistent delivery of statements would be stronger evidence. New product announcements and senior appointments should not be given the same evidential weight as audited cash generation. [S5]
What kills the thesis: sustained gross margin below 10%, weaker delivery/collection despite a large order narrative, material inventory write-downs, recurring equity issuance needed to bridge routine operations, or an adverse legal or reporting development. These are monitoring thresholds and failure mechanisms, not predictions. Fresh government developments should be checked at the next earnings event; we do not infer an outcome from a stock-price rally.
08. The base assumes a strong year and a slower, more cash-efficient expansion
| Independent house model | FY27 | FY28 | FY29 | FY30 | FY31 |
|---|---|---|---|---|---|
| Revenue, $bn | 75 | 87 | 98 | 108 | 117 |
| Operating margin | 7.0% | 7.0% | 7.0% | 7.0% | 7.0% |
| Net income, $bn | 3.808 | 4.416 | 4.945 | 5.459 | 5.957 |
| Owner cash after preferred, $bn | -3.519 | 0.537 | 1.344 | 2.389 | 3.144 |
| Cash at period end, $bn | 4.002 | 4.539 | 5.883 | 8.272 | 11.416 |
Our $75bn FY27 revenue assumption is above the unchanged August $65–72bn company range. It consists of the $19bn preliminary first-quarter midpoint plus $56bn over the remaining three quarters, or $18.67bn per quarter. That is a house judgement about sustained deliveries, not newly issued guidance. Subsequent annual growth slows to 16%, 12.6%, 10.2% and 8.3%, reaching $117bn. It assumes platform demand remains strong but that the initial step-up is not repeated indefinitely.
Trade working capital is defined consistently as receivables plus inventory, less payables and current/noncurrent deferred revenue. It was $14.162bn, or 36.3% of FY26 revenue, at June. The base requires it to improve to 28% of annual sales. Even then, the first year absorbs about $6.84bn of incremental trade working capital. The resulting initial owner-cash deficit is financed from existing cash in this model, not hidden by adding back the working-capital investment. [S2]
A 21% tax assumption, depreciation at 0.35% of revenue and capex at 0.60% apply to all cases. Net financing cost rises from $430m to $650m in the base as the business scales and note maturities approach. Operating margins retain compensation as a cash-equivalent expense; we do not add back future compensation as free money or charge a second stream of future award dilution. Existing award claims remain explicit. This is an economic scenario model, not formal GAAP diluted EPS.
The cash bridge holds a $3bn liquidity floor, borrowing if the forecast breaches it. Surpluses remain cash; there are no common dividends or buybacks. Interest is an explicit scenario assumption, not automatically repriced to every debt balance. Other operating assets/liabilities, taxes and precise intra-year timing are simplified. The stress cases show where these simplifications matter.
09. Three paths from the post-news price
| House case | FY31 sales | Op margin | Economic shares | EPS | Exit P/E | Future price | Annual return |
|---|---|---|---|---|---|---|---|
| Bear | $60bn | 4.5% | 832.0m | $1.61 | 9x | $14.53 | -17.3% |
| Base | $117bn | 7.0% | 835.3m | $7.13 | 12x | $85.57 | +17.9% |
| Bull | $160bn | 9.0% | 876.6m | $12.26 | 15x | $183.85 | +37.4% |
Bear five-year total-wealth CAGR: -17.3%. Base five-year total-wealth CAGR: 17.9%. Bull five-year total-wealth CAGR: 37.4%. Return = (terminal share value / $37.60)^(1/5) − 1. No common dividends, no reinvestment receipts, no probabilities and no separate buyback yield are assumed. Returns are nominal USD before investor tax, fees and currency changes. Five fiscal forecasts are mapped to a stylised 20 October 2026–20 October 2031 holding period.
The bear assumes the step-up in demand fades: FY27 sales reach $65bn, then retreat to $60bn by FY31. A 4.5% operating margin, $1bn net financing cost, 21% tax and a 9x multiple produce about $14.53 per share. The base requires $117bn revenue and a 7% operating margin. After $650m net financing cost and tax, $5.957bn of earnings is spread across roughly 835.3m economic shares: $7.13 EPS at 12x gives $85.57. The bull requires $160bn revenue, a 9% margin and a richer 15x valuation, producing $183.85.
The denominator matters. The model starts with 656.965m July common shares, adds 18.210m existing RSUs, treasury-method option dilution using 34.704m options at a $25.67 weighted exercise price, preferred conversion, and note-conversion premiums after capped calls. Conversion uses the terminal price as a proxy for prices at earlier maturity dates. Actual paths, vesting, tax withholding and settlement elections can change the result materially. The price and diluted share count are solved together rather than assumed independently. [S2]
The P/E exit capitalises earnings only. Although the cash bridge retains a terminal cash balance, that balance is not added again to the share price. This avoids counting retained cash both inside the selected multiple and as a second payoff. No new ATM shares are assumed in the central cases; the separate stress shows their effect.
10. Margin and valuation move returns; working capital moves funding
Base five-year total-wealth CAGR without rerating: 17.0%. This comparison freezes the observed $37.60 / $3.26 trailing reported EPS multiple of 11.53x. It is close to the base 12x exit, so the base does not rely on a large multiple expansion. However, the trailing GAAP denominator and the model’s economic share treatment are different; this is a valuation discipline check, not perfectly matched accounting. [S2] [S4]
| Change to base | Annual return | 20% entry ceiling | Peak additional borrowing |
|---|---|---|---|
| Exit P/E 10x | 13.8% | $28.85 | $0.00bn |
| Operating margin 6% | 14.1% | $29.23 | $0.00bn |
| Working capital 32% | 17.9% | $34.39 | $2.00bn |
Moving trade working capital from 28% to 32% consumes another $3bn in FY27. Holding earnings and the explicit interest path fixed leaves the P/E return unchanged, but triggers extra borrowing to preserve the cash floor. This is a funding stress, not evidence that worse cash conversion is harmless. If lenders charge more or the company issues shares, returns fall as well. The DCF includes the larger reinvestment burden directly.
Issuing the full hypothetical $1.25bn ATM at the reference price, with no earnings benefit from proceeds, reduces base annual return to 17.0%. This is a deliberately isolated dilution test, not a claim that issuance has occurred. A 10x terminal P/E or a 6% operating margin is more damaging than this single financing event. Combinations of these stresses can be worse than any individual sensitivity.
11. Cash valuation provides a deliberately different cross-check
The base unlevered economic cash-flow DCF is approximately $43.62 per share at a 12% discount rate and 3% perpetual growth. It uses the same five-year operating path, taxes and working-capital investment, then funds continuing working-capital growth in the terminal year. The terminal cash flow is $5.380bn; it does not assume growth with zero reinvestment.
| Discount rate / terminal growth | 2% | 3% | 4% |
|---|---|---|---|
| 10% | $55.86 | $60.50 | $66.61 |
| 12% | $41.38 | $43.62 | $46.43 |
| 14% | $31.73 | $33.08 | $34.46 |
Enterprise value is reduced by $8.781bn of starting debt at face value, increased by $7.521bn of cash, and reduced by the present value of preferred coupons. Preferred conversion is included in the economic share count rather than subtracted again as principal. Future financing flows are excluded from enterprise cash flow. Conversion prices and aggregate option strikes are simplified, so this cross-check is not a precise appraisal of each security. [S2]
The DCF is much lower than the $85.57 base five-year exit price because an exit value is a future amount and the DCF discounts interim reinvestment and a cash-intensive perpetuity. The DCF is modestly above the reference price; the entry for a 20% return is below it. These answer different questions. Neither should be presented as a guaranteed price target.
12. The price that clears a 20% hurdle is conditional
The base-case conditional entry for a 20% five-year total-wealth CAGR is $34.39 per share, calculated as $85.57 terminal wealth divided by 1.20^5. The calculation uses unrounded model outputs. At $37.60 the same frozen path produces 17.9% per year; paying about $34.39 would increase the mathematical return to 20% if that path occurs. Buying at that price does not make the outcome certain.
The purchase price changes the investor’s return calculation, not the business forecast, financing terms or frozen share path. A lower share price can itself change future dilution and financing access in reality; the entry explorer intentionally isolates the price effect. Before acting on a lower price, confirm that margin, cash collection, customer readiness and regulatory evidence have not deteriorated. A cheap entry into a broken path is not a margin of safety.
13. The next evidence should decide the next action
| Evidence window | What to inspect | Investment consequence |
|---|---|---|
| 3 November 2026, 5 pm ET | Final September revenue/GM, EPS, cash and revised annual guide | Confirm the preliminary beat; do not assume it guarantees cash |
| Next quarterly filing | Receivables, inventory, customer advances, debt and shares | Re-underwrite working capital and dilution |
| Next 2–4 quarters | Recurring margin near 12%, cash conversion and deployment readiness | More constructive if growth increasingly self-funds |
| As filed or announced | Control remediation and authoritative legal developments | Reassess confidence and required return immediately |
The 3 November call is 9 am on 4 November in Sydney under the scheduled US/Sydney daylight-saving offsets. The company may release additional details before or with the call; the event is an evidence window rather than a guaranteed positive catalyst. [S1]
This is an absolute-return common-equity research idea with substantial AI infrastructure, component-cycle, financing and event exposure. Current short interest, borrow terms, options positioning, passive ownership and a fresh institutional-holder register were not verified. No conclusion about a short squeeze, crowded positioning or appropriate portfolio weight is supported. The practical action is to wait for cash evidence or a better conditional entry, then re-underwrite.
14. What was checked, and what remains unresolved
All six supplied PDFs were extracted with page markers and SHA-256 hashes; key pages were rendered and inspected. The one-page news exports corroborate the company update. The three-page 8-K export refers to Exhibit 99.1 but does not include the full release, so the original company release supplies the details. A stray currency symbol before the 12.5% margin in the news export is treated as a transcription error, not a financial amount. [S1]
The JPMorgan and Barclays notes are from 12 August, and the UBS HOLT sector note is from 15 September. Their price targets, share counts and valuation dates are not current consensus. They provide useful context about June margin mix, funding and competitive pressure. Differences in reported adjusted EPS and share definitions are not averaged together; audited company financials and explicitly defined house claims take priority. No proprietary bank charts or long excerpts are reproduced. [S7] [S8] [S9]
The FY26 10-K is the primary source for accounting and capital terms. The earlier Capital Thesis edition is preserved as an archive, but this edition builds a fresh operating/cash/claim model after the October update; it does not relabel September scenarios as new research. Missing September cash, current issuance and post-news consensus remain explicit limitations. Sources support factual inputs; forecasts, hurdle prices and investment judgement are Capital Thesis assumptions. [S2] [S10]
Financial exhibits
Exact observations accompany every chart. Actuals, management guidance and Capital Thesis assumptions are labelled separately; the chart shapes do not imply probabilities.
A $4bn increase in the quarterly revenue midpoint
The new preliminary range reflects a completed quarter; it is not an order backlog.
| Observation | Q1 FY27 revenue midpoint |
|---|---|
| 11 Aug guidance midpoint | 15 |
| 20 Oct preliminary midpoint | 19 |
Ranges $14.5–15.5bn and $18.5–19.5bn. Preliminary unaudited; midpoint house arithmetic. Sources: [S1] [S3]
Above guidance, below the unusually strong June quarter
The comparison base changes the conclusion.
| Observation | Gross margin |
|---|---|
| Q4 FY26 GAAP actual | 17.5 |
| Old Q1 FY27 guide midpoint | 10.6 |
| New Q1 FY27 prelim midpoint | 12.25 |
Different quarters and evidence types are labelled. Gross margin is before operating expenses. Sources: [S1] [S3]
Accounting profit and cash generation diverged
Working capital, not large fixed-asset capex, drove the FY26 funding gap.
| Observation | Net income | Operating cash flow | CFO less cash capex |
|---|---|---|---|
| FY2024 actual | 1.1527 | -2.486 | -2.6103 |
| FY2025 actual | 1.0489 | 1.6595 | 1.5323 |
| FY2026 actual | 2.2305 | -6.8099 | -6.9719 |
Fiscal years end June. Net income includes all claims before common allocation. FCF excludes preferred dividends and other investments. Sources: [S2]
Inventory and invoices absorbed cash before collection
Deposits and supplier terms offset only part of the annual absorption.
| Observation | FY2026 operating cash effect |
|---|---|
| Inventory | -8.8767 |
| Receivables | -3.9219 |
| Deferred revenue | 1.8807 |
| Payables | 0.9633 |
Selected cash-flow movements, not closing balances and not a complete CFO reconciliation. Sources: [S2]
Future earnings require scale and retained margin
The house scenarios are operating conditions, not probabilities.
| Observation | Terminal economic EPS |
|---|---|
| Bear FY31 house | 1.6142 |
| Base FY31 house | 7.1312 |
| Bull FY31 house | 12.2565 |
Capital Thesis assumptions: sales $60/$117/$160bn; operating margins 4.5%/7%/9%; explicit financing, tax and dilution. See model. Sources: [S1] [S2]
Source register and research basis
Evidence checked for this edition on 21 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.
Filed 31 August 2026; year ended 30 June · Statements pp 58–63; working capital/credit pp 71–72; debt Notes 8–9; awards Note 12; preferred/ATM Note 13; investigations Note 15; controls Item 9A. Current common count 31 July.
https://www.sec.gov/Archives/edgar/data/1375365/000137536526000022/smci-20260630.htm11 August 2026 · Quarter margin/cash, prior Q1 and FY27 guidance, management order commentary. Annual actuals checked against subsequently audited 10-K.
https://www.supermicro.com/en/pressreleases/supermicro-announces-fourth-quarter-and-full-fiscal-year-2026-financial-results20 October 2026, 7:41 pm EDT after-hours; regular close 4 pm · Frozen $37.60 after-hours versus $32.47 close. Page later changes; not an executable or final-close quote. Screen multiples based on the regular close are not reused.
https://stockanalysis.com/stocks/smci/15 October 2026 · Appointment is an operating development, not proof of remediation.
https://www.supermicro.com/en/pressreleases/supermicro-appoints-mark-long-chief-financial-officer23 September 2026 · Product/rack integration capability; no quantified incremental revenue assumed from the announcement alone.
https://www.supermicro.com/en/pressreleases/supermicro-now-shipping-nvidia-vera-rubin-nvl72-racks12 August 2026 · Private supplied document, pp 1–2; Neutral and $45 December 2027 target are historical. Portal link identifies provenance, not a publicly accessible report. Used only as dated secondary context.
https://app.godelterminal.com/12 August 2026 · Private supplied document, pp 1–2; Equal Weight and $39 target are historical. Margin mix and deployment discussion; no charts reproduced.
https://app.godelterminal.com/15 September 2026 · Private supplied 31-page sector note; p 16 relative-return/valuation context, not a company forecast or October consensus.
https://app.godelterminal.com/15 October 2026 · Earlier edition retained. Current operating paths and cash valuation are freshly recalculated.
https://capital-thesis.pages.dev/report?slug=smci-2026-10-15-full-877987083896e8787761