The $116 Billion Question

Microsoft spent USD 115.9 billion on capital equipment in one year — up 80% — kept Big Tech's widest server depreciation range, and reported $17.95 of earnings per share. On our three-year server clock that number re-marks to $16.96. At $451.10, the market is paying a bigger multiple on the bigger number.

TON618 Capital Research · As of July 31, 2026 · Price $451.10 (Jul 30, 2026 close) · Download PDF

Every filed figure in this note is drawn from Microsoft's FY2026 Form 10-K (filed July 29, 2026) and other SEC filings, pulled and verified July 31, 2026, and is as-of dated. Fact ("the filing says…") is separated from estimate ("our model implies…") and opinion ("we believe…") throughout. Model figures are estimates and are labelled as such. Prepared to CFA Institute analytical standards (methodology only); independently sourced against primary filings; §9 documents every vendor, accession number and formula.

Executive summary

Fair value $403 per share against the $451.10 close of 30 July 2026 — an 11% gap. Probability-weighted expected value $398; bull $516, base $403, bear $269 at 25/50/25. Twelve-month evaluation window.

• The operating business is superb, and it is in the price. Revenue +18% to $331.8bn, Azure +41%, commercial RPO +84% to $678bn, operating margin 46.8%.
• The bill is arriving fastest here. Depreciation grew 56% off the biggest investment year in the company's history — $115.9bn of cash capex, +80% — with $26.7bn of equipment bills unpaid at year-end.
• The clock still defers about $9bn of FY2026 expense (est.). Microsoft held its two-to-six-year server schedule; on our three-year clock FY2026 EPS of $17.95 re-marks to $16.96, and $37.7bn is deferred across FY2026–28.
• Both valuation bases land below the price. $427 on reported earnings, $403 on the re-marked basis. The reverse DCF says today's price needs ~20% revenue growth next year fading to ~12%.

Where to find the work: the re-mark is §4, the valuation §6, the case against us §7, and the falsifiers we will be scored on §8.

The story so far

On 17 July we published The Depreciation Question, which argued that Big Tech is expensing a record AI hardware build on schedules that flatter current earnings, deferring roughly $221bn of depreciation out of 2026–28. On 30 July, The First Invoice scored that claim against the Q2 prints: depreciation accelerated at all four hyperscalers, every capex assumption held, and no company shortened a clock. This note is the first single-name valuation built on that framework. Microsoft's chapter role: the name that held its schedule — and whose bill is arriving fastest. It reported FY2026 on 29 July, filing a full 10-K the same day, and the shares rose 15.5% the following session to $451.10.

§1 The thesis in five entries

We would revisit on a price below ~$400, on evidence that server economics genuinely support six-year lives, or on capex-to-revenue turning down while Azure growth holds (§8).

1. The operating business is superb, and that is in the price. Revenue +18% to USD 331.8bn, Azure +41%, commercial RPO +84% to USD 678bn, operating margin 46.8%. Data shows a franchise compounding at scale; we do not dispute it — we value it, at a cost of equity of 8.9% with a 4.67% ten-year Treasury underneath.
2. The bill is arriving fastest here. Depreciation grew 56% — the fastest of the four hyperscalers — off the biggest investment year in Microsoft's history (USD 115.9bn of cash capex, +80%). The server gross book grew USD 83bn to USD 215.9bn in twelve months, USD 26.7bn of equipment purchases sat unpaid in accounts payable at year-end (3.9× the prior year), and another USD 34.6bn of construction is contracted. The spend is not decelerating: FY2027 capex is guided higher (press-sourced, §3), with first-quarter spend incl. finance leases above USD 50bn.
3. The clock still defers ~USD 9bn of FY2026 expense (est.). Microsoft held its two-to-six-year server schedule — to its credit, with upgraded disclosure. But on our three-year server clock the schedule deferred roughly USD 9.1bn of pre-tax depreciation out of FY2026 alone, ~USD 37.7bn across FY2026–28 (model estimates, §4). Re-marked EPS: $16.96 vs $17.95 reported.
4. The day it filed "no change," it stretched a different clock. The FY2026 10-K discloses no change in estimate. On the same day's earnings call, management extended datacenter and office building lives from 15 to 25 years effective FY2027 and reclassified future datacenter leases — moving ~USD 15bn out of the headline capex-plus-leases number (press-sourced from three carriers, §3). Legal, disclosed, and directionally the same ratchet the series documents.
5. Both valuation bases land below the price. On reported earnings our blended fair value is $427; on the re-marked basis $403. The market closed at $451.10 after a 15.5% one-day repricing. A reverse DCF says today's price requires ~20% revenue growth next year fading to ~12% — meaningfully above our base path and, we believe, above what a USD 331.8bn revenue base sustains for five years (§6).

Fund context. TON618 Capital is a Bitcoin fund; Microsoft is not a BTC-correlated instrument, and this work supports the fund's uncorrelated equity research and hedging mandate. The Fund holds no position in any company named here.

§2 The machine: three businesses, one invoice

(If you know Microsoft's segments, skip to §3. The reader's guide to depreciation mechanics — clocks, vintages, and the dose — lives in the series scoreboard, The First Invoice, §0.)

Microsoft sells software subscriptions and rents computing. Three reported segments:

• Productivity and Business Processes — Office/Microsoft 365, LinkedIn, Dynamics. FY2026 revenue $140.0bn (+16%), operating income $83.9bn at a 59.9% margin. The subscription annuity; the profit engine that funds everything else.
• Intelligent Cloud — Azure, servers, enterprise services. Revenue $137.8bn (+30%), operating income $57.0bn at a 41.3% margin, down from 42.0%: segment cost of revenue grew 44%, which the filing attributes to "investments in AI infrastructure to support growing customer demand." Azure and other cloud services grew 41%. This is where the invoice lands.
• More Personal Computing — Windows, Xbox, Surface, search advertising. Revenue $54.1bn (−1%). Mature; a cash source, not a story.

Across segments, Microsoft Cloud revenue reached $214.4bn, +27% — now 65% of the company. Commercial RPO — contracted revenue not yet delivered — grew 84% to $678bn, with a weighted-average duration of ~2.3 years and ~30% expected to convert within twelve months. That backlog is the strongest fact the bull case owns (§7), and we state it here rather than bury it.

The print and the pop. The FY2026 10-K was filed July 29 — a full-year filing, same day as the release. Revenue +17.8%, net income +31% to $133.7bn, diluted EPS $17.95. The stock, which had de-rated from a 52-week high of $555 to a late-June trough near $349, jumped +15.5% the next day to $451.10. One footnote belongs next to the +31%: FY2026 pre-tax income includes $11.0bn of net recognized gains on investments and derivatives (a +$15.6bn nonoperating swing year over year, including equity-method gains related to OpenAI). Operating income — the line the clock actually touches — grew 20.8%. Both facts are in the filing; only one made the headline.

§3 The invoice: USD 115.9bn, terms two to six years

Source: Microsoft Forms 10-K, FY2019–FY2026 (cash flow statements; depreciation expense per PP&E note), pulled via SEC EDGAR July 31, 2026.

The gap between the gold bars and the dark bars is the series' whole subject. Capital spending nearly doubled to $115.9bn — the largest fiscal year in Microsoft's history and, on an annual filing basis, second in the group only to Amazon's current run-rate — while the depreciation recognized against it reached $34.3bn. The difference is not missing; it is scheduled. Under straight-line accounting, the 2024–26 vintages will surface as expense through 2028–2032 at whatever pace the useful-life estimate dictates. The estimate Microsoft uses for servers and network equipment: two to six years — the widest disclosed range in Big Tech, with no point estimate.

Three facts from the new 10-K say the bill is arriving faster than the schedule alone would suggest:

• Depreciation expense grew 56% ($22.0bn → $34.3bn) — the fastest of the four hyperscalers this season (Meta +46%, Alphabet +43%, Amazon +31%).
• The server book grew $83bn in one year — the "Servers, network equipment, and software" gross line went from $132.8bn to $215.9bn. Most of that hardware started depreciating within the year.
• $26.7bn of equipment purchases sat unpaid in accounts payable at June 30 — versus $6.9bn a year earlier, a 3.9× jump. Machines are being installed faster than they are being paid for, let alone expensed. Contracted construction adds $34.6bn; total purchase commitments stand at $194.1bn, $169.0bn of it due within a year.

Source: Microsoft FY2026 Form 10-K (0001193125-26-323660), PP&E note: "purchases of property and equipment remaining in accounts payable." FY2024–FY2026.

Credit where due — and a new entry in the ledger. In the July note Microsoft was the least transparent name in the group: server hardware buried in "computer equipment." The FY2026 10-K names the category for the first time — policy note and PP&E table both — a real disclosure upgrade, with no change in estimate disclosed: the server clock held. But the same day, on the earnings call, CFO Amy Hood announced that the useful life of datacenter and office buildings will be extended from 15 to 25 years effective FY2027 — "reflects our operating history and expected use of these assets" — with more future datacenter leases classified as operating rather than finance leases as a consequence. The mechanical result, per the company: calendar-2026 capex including finance leases is now expected around $175bn versus roughly $190bn prior — with Hood explicit that "outside of this useful-life impact, our calendar year 2026 capex investment expectations remain unchanged." (These are call statements, not filing text; we source them from three named carriers in §9 and will verify against the first FY2027 10-Q.) We report it straight: buildings are not servers, and management guided that the change provides only "a minimal benefit" to FY2027 operating income. But the series' July count stood at thirteen server-life extensions against one reversal since 2020 across this group; add this buildings change and the tally is fourteen life extensions against one reversal, the newest arriving the same day the 10-K said "no change." The ratchet turns quietly, and it turns one way.

And FY2027 spending goes up, not down. Guidance from the same call: capital expenditures grow year over year in FY2027, with fiscal Q1 spend including finance leases above $50bn (press-sourced; the only forward number in an issuer document is the $34.6bn construction commitment). The v3 model's deferral math below assumes FY2027–28 capex flat at $115bn. Data now says that assumption is conservative — which makes the deferral estimates a floor, not a ceiling.

§4 The re-mark: $17.95, counted honestly

Everything in this section below the first paragraph is a model estimate, stated as such, from the July note's v3 vintage model — construction, validation against five disclosed life-change benefits, and limitations documented in that note's §2 and §7. Nothing was refit for Microsoft's FY2026 print; the model was published July 17 and its FY2026 capex assumption ($115.0bn) missed the filed actual ($115.9bn) by 0.8%.

The question the series asks of every name: what would earnings be if server hardware were expensed over the ~three years the technology cycle actually gives it? NVIDIA ships a new accelerator architecture roughly every two years; each generation resets the economics of the last. A six-year server clock does not make the hardware last six years — it makes the expense arrive later. The model takes Microsoft's own capex vintages, assumes servers are ~45% of capex (sensitivity 40–50% shown), and re-runs depreciation on a three-year straight-line clock against the reported schedule.

Sources: Microsoft FY2026 Form 10-K (EPS, tax rate, share count, investment gains); TON618 v3 vintage model, published July 17, 2026 (server-depreciation delta, estimate). Band reflects the 40–50% server-share sensitivity.

Stated plainly: we estimate reported FY2026 EPS is overstated by about 5.8% against a three-year server clock — about $0.99 a share, $7.4bn after tax. And because the 2024–26 vintages are already bought, the gap widens mechanically before it closes: the model puts the deferral at USD 9.1bn (FY2026), 13.3bn (FY2027E) and 15.3bn (FY2028E) pre-tax — roughly USD 37.7bn cumulative (band 33.5–41.9), on capex assumptions guidance has already exceeded.

Source: TON618 v3 vintage model (server-only depreciation, both clocks — model estimates); reported total depreciation per Microsoft 10-Ks for scale. FY2027–28 model years assume USD 115bn capex, below current guidance direction.

Three honesty notes, carried in every piece of this series. First, nothing here alleges fraud or non-compliance: Microsoft's schedules are disclosed, audited, and legal, and its estimates sit inside industry practice. The question is earnings quality, not legality. Second, we do not claim the distortion wrecks returns-on-capital arithmetic: our own July finding is that ROIC levels barely move under the re-mark — the damage is concentrated in EPS quality and in the 2026–28 vintage burn-through. Third, the re-mark does not change cash. It changes when cash already spent shows up as expense — which matters precisely because the market pays a multiple of the expense-timing number, not of cash (§5), and because a three-year economic life implies structurally higher replacement capex forever (§6).

§5 What the market pays for the bigger number

At $451.10, Microsoft trades at 25.1× its reported FY2026 EPS — after the one-day 15.5% repricing. On the re-marked $16.96 that is 26.6×; on core re-marked earnings of $15.77, 28.6× (both estimates). The de-rate from the 2025 highs (when the stock briefly paid ~37× trailing) has happened; the premium to clean earnings has not gone.

Sources: closing prices July 30, 2026 (Massive Market Data, adjusted); trailing diluted EPS from SEC XBRL company facts via edgartools, consecutive four-quarter windows (Meta through Mar 31, 2026 — its Q2 10-Q was not yet in the facts feed). Alphabet and Amazon shown ex disclosed non-operating investment gains, tax-effected at the 21% statutory rate (estimates); their headline P/Es are 16.8× and 18.9×. MSFT re-marked bars are house-model estimates.

Two readings of the same chart, both honest. The generous one: on clean trailing earnings Microsoft is mid-pack — cheaper than Apple (38×) and NVIDIA (30×), near Alphabet ex-gains (30×), richer than Meta (~20×) and Oracle (22×). The stricter one: the peer multiples themselves are quoted on the same stretched clocks this series documents — Alphabet's six-year servers, Meta's 5.5, Amazon's five — so "mid-pack against peers" partly means mid-pack against the same flattery. We anchor instead on a justified multiple: for a franchise with Microsoft's backlog and margins but capex intensity of 35% of revenue and a 4.67% risk-free rate underneath, we apply 24× trailing (a discount to the peer ex-gain median of ~30×, a premium to the market's long-run average): $431 on reported EPS; $407 on the re-marked $16.96. We believe the second number is the honest one; we show both.

What is actually free about the free cash flow. Cash FCF (operating cash flow $182.9bn minus capex $115.9bn) was $67.0bn — a 2.0% yield on the market cap. On an accrual basis the number is worse: the unpaid-equipment balance in accounts payable grew $19.8bn during the year, meaning roughly $47bn (est.) of FY2026 economic FCF — a 1.4% yield, before counting $20.4bn growth in finance-lease obligations (now $66.6bn, 13-year weighted term). Shareholders received $48.7bn in dividends and buybacks — more than accrual FCF. Data shows the equipment vendors are financing part of the shareholder return.

§6 What the arithmetic says it is worth

Method choice. Microsoft is a mature-growth, dividend-paying, buyback-heavy equity with positive and forecastable free cash flow to equity — the house FCFE framework fits (a supernormal-DDM collapses to the same arithmetic at MSFT's 27% payout; FCFF adds nothing since leverage is trivial: $40.3bn debt against $76.8bn cash and investments). We run a five-year explicit FCFE model plus terminal value, twice: once with a terminal economy on Microsoft's reported clock, once on ours — because if servers truly last ~three years, replacement capex is structurally higher forever, and terminal FCFE conversion is lower. That — not the EPS optics — is how the clock enters a DCF: cash is clock-agnostic in any single year, but the clock is a claim about how much reinvestment the business permanently requires.

Cost of equity 8.95% = 4.67% ten-year Treasury (July 29) + 0.90 beta (computed, one year daily vs S&P 500) × 4.75% equity risk premium. Terminal growth 4.0%. Diluted shares held flat at 7,453m: buybacks — $22.3bn of repurchases in FY2026, a figure that includes $5.6bn of tax-withholding settlements on vesting stock awards — have offset stock-comp issuance to −0.2% a year, and we assume that continues; SBC ($12.4bn, 3.7% of revenue) stays fully expensed in net income — no add-back anywhere in this note. The margin path takes ~1 point of pressure in FY27–28 as the model's own reported-clock server depreciation climbs from USD 18bn to USD 33bn (est.), partly offset by operating leverage and the buildings-life benefit; conversion starts at 48% because guidance points to another record capex year.

• Reported-clock terminal (90% conversion): $424 per share. Exit P/E ~19×.
• House-clock terminal (84% conversion): $400 per share. Same explicit years — only the perpetual reinvestment assumption changes.

Blending 60% DCF / 40% comps on each basis: $427 on reported earnings; $403 on the house basis — a spread of about $24 a share (~5%), which is what the clock is worth in this valuation. Both land below $451.10.

Source: TON618 model (model_out.json, July 31, 2026). All fair-value estimates; price line = July 30, 2026 close, Massive Market Data.

Scenarios and expected value (probabilities are our judgment, stated so they can be argued with):

Probability-weighted expected value: $398 — 11.8% below the price. The payoff is asymmetric the wrong way: +14% to the bull case, −40% to the bear.

The embedded-expectations test. Run backwards, the market price is informative: to justify $451.10 on the house basis, our model needs revenue growth of roughly 20% next year fading to 12% by FY2031 — a five-year path (~16.3%/yr) that would take Microsoft to ~$705bn of revenue by FY2031, against our base path's $581bn. With commercial RPO at $678bn that is not absurd — it is simply priced: the market is paying today for the bull case arriving on schedule, on earnings counted on the friendliest clock in the group's range.

§7 The case against this note

The steelmanned bull — the strongest version, argued as its holders would argue it.

One: the backlog is real money. Commercial RPO grew 84% to $678bn with ~30% converting inside a year; "demand exceeds available supply" (Hood). Our 15.5% base growth could simply be wrong, and every point of sustained growth is worth roughly $11 of fair value. Two: the clock may be closer to right than we allow. Microsoft holds fleets of prior-generation GPUs at high utilization years after deployment; if blended server economic life is five-plus years — inference workloads running on depreciated hardware are among the most profitable compute sold — the re-mark shrinks toward zero, and our 45% server-share assumption (unverifiable from the filing) overstates the base. Three: the balance sheet and annuity mix are the group's best: $76.8bn of cash and investments against $40.3bn of debt, 60% of revenue subscription-like, operating margins near 47% while absorbing a 56% depreciation jump — the machine has absorbed every prior investment cycle and re-rated afterward. Four: the de-rate already happened — the stock fell ~37% peak-to-trough on exactly these fears and still printed +18% revenue growth; paying 25× for that is not obviously a bubble. Five: our bear case double-counts — if capex slows, conversion recovers; if it doesn't, growth stays high; the state where growth fades and capex stays at $150bn+ requires management to burn money into falling demand, which this management has not historically done.

Our answer, briefly: points one and four are why our bull case carries a full 25% weight and why our fair value sits below the price on the multiple, not on the franchise. Point two is testable and is on the falsifier list (§8). Point five understates how much of the spend is committed — $194bn of purchase obligations, $169bn due within a year — spend that arrives whether or not demand does.

Risk register (likelihood × magnitude on a 12-month view; magnitude = plausible fair-value impact, our estimates):

Market risk numbers (computed from one year of daily prices): realized volatility 31.7% (1-year), 41.4% (90-day), 55.4% (30-day, inflated by the earnings gap); one-day parametric VaR at 95/99%: 4.3% / 6.1% of position value; the print itself moved the stock 15.5% in a day — treat the tails as fat. Liquidity is not a constraint: ~$50bn traded on July 30 alone.

§8 Catalysts, falsifiers, and the ledger we will keep

Dated catalysts.

What would change our mind — posted as falsifiers, same standard we hold the July note to: (1) disclosed server-fleet life/utilization data supporting five-plus-year economics — this would cut the re-mark directly; (2) capex-to-revenue turning down while Azure growth holds above ~35% — conversion would recover ahead of our path and the base case migrates toward the bull; (3) two consecutive quarters of depreciation growth decelerating below revenue growth — the wedge would be closing, not opening; (4) evidence our 45% server share is materially high (e.g., a buildings/land split showing structures dominate the recent vintages). Any of these posts to the series scoreboard, which re-scores each earnings season.

The ledger. Per house practice this note's fair value is logged to the Signal Ledger and scored out-of-sample.

Where Microsoft sits in the series. Five companies, one invoice, five ways of carrying it. Microsoft is the name that held its server clock while the bill arrived fastest — and stretched a building clock the same afternoon. Alphabet is the benchmark the group is measured against; Meta is compounding up the dose curve fastest; Amazon is the one company that shortened a clock, and pays for the honesty in cash; Apple is the control — the income statement the buildout never touched. The scoreboard that holds them together, and re-scores every season, is The First Invoice.

§9 Sources & Method

Every figure in this note traces to a source below with an as-of date; the Phase-0 capture (raw pulls, normalized tables, content hashes) is archived with a manifest at ton618-tech-valuation/data/MSFT/2026-07-31-052516/. No figure is retyped between the model and the charts — all exhibits render from the archived data and model_out.json.

Issuer filings (primary). Microsoft FY2026 Form 10-K, accession 0001193125-26-323660, filed July 29, 2026, pulled via SEC EDGAR (edgartools) July 31, 2026 — income statement, balance sheet, cash flow, segment note (Note 18), PP&E note (depreciation expense, servers gross book, capex-in-accounts-payable, $34.6bn construction commitments), useful-life policy ("servers and network equipment, two to six years"), RPO disclosures, commitments table, effective tax rate, share counts (cover: 7,425,545,491 shares at July 23, 2026). Prior-year figures (FY2015–FY2025) from Microsoft Forms 10-K via the July note's archived pull. Peer trailing EPS: SEC XBRL company facts via edgartools, July 31, 2026; missing fiscal fourth quarters derived as fiscal-year minus three reported quarters; four-quarter windows verified consecutive (Meta's window ends March 31, 2026).

Market data. Closing prices (MSFT, AAPL, GOOGL, AMZN, META, ORCL, NVDA, CRM), July 30, 2026, adjusted, via Massive Market Data (/v2/aggs), pulled July 31, 2026; 52-week range, realized volatility and the 0.90 beta computed from the same daily series against SPY. Risk-free rate: FRED series DGS10, 4.67% as of July 29, 2026, pulled July 31, 2026.

Model (estimates). TON618 v3 vintage depreciation model, published July 17, 2026, unchanged since — MSFT server share 45% (band 40–50%), three-year straight-line house clock; FY2026 server-depreciation delta USD 9.14bn pre-tax (band 8.12–10.15); FY2026–28 cumulative USD 37.7bn (band 33.5–41.9); model FY2027–28 capex assumption USD 115bn/yr, now below guidance direction. Re-marked EPS = reported EPS − delta × (1 − 0.194 effective tax rate) ÷ 7,453m diluted shares. Ex-gains adjustment: USD 11.047bn pre-tax net recognized gains on investments and derivatives (10-K cash-flow add-back), tax-effected at 19.4%; peer ex-gain P/Es tax-effect disclosed gains at the 21% statutory rate per the series convention. FCFE DCF: five-year explicit + Gordon terminal; all assumptions in §6; cost of equity 8.95% = 4.67% + 0.90 × 4.75% ERP (ERP is a house assumption). Beta 0.90 is a one-year daily regression against SPY (251 paired observations, adjusted closes, independently recomputed at 0.904); trackers quoting 5-year monthly windows show materially higher figures (~1.1–1.3) — the window choice is disclosed rather than harmonized, and a reader preferring the longer window should mark every fair value here lower. Enterprise value = market capitalization ($3,349.7bn at $451.10) + total debt ($40.3bn) + finance-lease liabilities ($66.6bn, treated as debt) − cash and short-term investments ($76.8bn) = $3,380bn; excluding finance leases from the bridge, EV would be ~$3,313bn. Accrual FCF estimate = cash FCF − year-over-year change in capex-in-accounts-payable. Scenario probabilities are analyst judgment.

Press-sourced (labelled). FY2027 capex guidance ("grow year-over-year"; Q1 incl. finance leases >$50bn), the buildings useful-life extension (15→25 years, effective FY2027), the finance-to-operating lease reclassification, and the ~$175bn vs ~$190bn calendar-2026 capex-incl-leases figures are earnings-call statements sourced from three named carriers pulled July 31, 2026: Benzinga (two articles, July 29–30) and CFO Dive (July 30). We do not cite third-party transcripts; the only forward capex figure in an issuer document is the $34.6bn construction commitment. All press-sourced claims are flagged inline and will be checked against the FQ1-2027 10-Q.

Known gaps, stated. Azure revenue is not disclosed in absolute dollars (growth rates only); the server share of capex is a model assumption, not a disclosure; Meta's comp window trails one quarter pending its 10-Q in the facts feed; scenario probabilities and the justified 24× multiple are judgment, argued in text. This note carries a target price and will be scored against it once published.

§10 Disclosures

Information only. TON618 Capital. This report is for information purposes only. Nothing here is an offer to sell or a solicitation of an offer to buy any security, fund interest, or digital asset, and nothing here is personalized investment advice or a recommendation regarding any instrument.

Publisher's exclusion. All research is published solely as general, impersonal information of regular circulation. It is not tailored to the objectives or circumstances of any individual and is not issued in connection with compensation from any client. The Fund has no clients and distributes all research free of charge. On that basis it publishes in reliance on the publisher's exclusion from the definition of "investment adviser" under the Investment Advisers Act of 1940 (§202(a)(11)(D); cf. Lowe v. SEC, 472 U.S. 181 (1985)).

Registration & conflicts. TON618 Capital is not registered as an investment adviser or broker-dealer in any capacity. The Fund is a Bitcoin fund and may hold or transact in the securities or digital assets it discusses. This report discusses Microsoft (MSFT) and, for comparison, Apple (AAPL), Alphabet (GOOGL), Amazon (AMZN), Meta Platforms (META), Oracle (ORCL), NVIDIA (NVDA) and Salesforce (CRM); the Fund holds no position, long or short, in any of them, and has no economic interest in the price of any security named here. The Fund receives no compensation from any party in connection with its research.

Use of AI. Artificial intelligence is used in the creation of this research. All methodology and data integrity are reviewed and approved before publication by TON618 Capital's Chief Investment Officer, Keyth Beck; errors may nonetheless occur, and readers should verify independently.

CFA. This report was prepared to align with CFA Institute analytical standards (methodology only). CFA® and Chartered Financial Analyst® are registered trademarks owned by CFA Institute. That reference describes the analytical framework applied; it does not imply the report was prepared, reviewed, or authored by a CFA charterholder, and the report is not issued, reviewed, endorsed, certified, or approved by — nor affiliated with — CFA Institute.

Risk & feedback. Past performance is not indicative of future results. Digital assets and equities are volatile and may result in total loss of capital. Corrections and feedback are welcome — please direct them to CIO Keyth Beck at keyth@ton618capital.com.