Profitable businesses can support demanding prices—but not cost-free growth

The answer

The accumulated evidence supports a real, investable commercial expansion more strongly than an economy of empty revenues and exclusively paper gains. It does not justify treating every profitable business, financed facility or successful exit as a fairly priced investment. The two new ownership analyses do not establish that either selected security is demonstrably a bubble. They do identify materially different conditions on which its price depends.

NVIDIA is the stronger observed profit franchise. Its ownership price does not require assuming that every downstream AI customer earns a satisfactory return. Under a costed continuation that allows manufacturing margins to narrow, funds research, compensates employees, finances growth and pays for renewal, the $237.47 reference can be matched by approximately 22.6% annual revenue growth for five model years, followed by a five-year slowdown to 3% nominal continuing growth. That entails about $1.07 trillion of annual revenue in year five and $1.77 trillion in year ten. These are solved requirements, not company forecasts. They are demanding at that scale, but not contradicted merely by the fact that the future has not happened.¹²

Freshworks' $13.60 reference is a different proposition. Recurring customers and high gross margin support a continuing application franchise. The present all-cost operating margin does not support simply treating its cash-flow addback for stock compensation as money free to ordinary shareholders. A paid-renewal path with 15% initial revenue growth and a mature 20% economic operating margin falls short at the reference 11% required return. The same cost package needs about 20.6% initial growth to match the price. But a stronger distribution/compensation outcome supports the price at about 11.8% initial growth, and the less optimistic cost package supports it with approximately 13.7% growth at a 9% required return. A chosen hurdle therefore cannot turn this into a demonstrated overpricing verdict.³⁴

The new distinction is not “hardware good, software bad.” NVIDIA's principal burden is the duration and scale of paid demand at high retained margins. Freshworks' is conversion of a real customer franchise into a larger share for continuing owners after acquisition, development and compensation. Both have plausible favorable paths. Neither path is established by the corresponding promotional shorthand: more tokens do not automatically mean more paid revenue, and a high adjusted cash margin does not identify the ordinary shareholder's economic margin.

These are stated-inventory ownership references, not exact October fully diluted capitalizations. The option interval does not bound unmeasured employee-purchase/performance rights. Section 2 identifies the residual claims, dates and sensitivities; sections 6–8 give the complete model. No scenario has a probability weight.

The evidence-weighted whole answer is selective value creation with a supported vulnerability to excessive or incompatible ownership expectations—not a demonstrated, uniformly overpriced AI sector. the earlier investigations strengthened productive completion, buying, efficiency and actual realization. this investigation shows that profitable continuing operations deserve substantial value even after paying to remain competitive. It also shows where optimistic explanations actually fail an account: ignoring a procurement cash requirement, treating compensation as free, or assigning the same cash to both repurchases and remaining assets. Those are identifiable analytical failures. They are not estimates of how many market participants make them.

Build the claim before deciding what the business is worth

2.1 The price date is not the balance-sheet date

The inspectable price source gives NVDA $237.47 and FRSH $13.60 at 4 p.m. Eastern on October 7. The model pairs each price with a matched statement-date share, asset and liability inventory: July 26 for NVIDIA and June 30 for Freshworks. Multiplying a recent quote by an older disclosed inventory is transparent; calling the result an exact current market capitalization would not be.¹³

NVIDIA's July equity roll-forward ends at 24,147 million issued shares. Its later cover gives 24.1 billion as of August 21, rounded too coarsely to infer an exact intervening repurchase. Freshworks has 238.848283 million Class A plus 23.961336 million Class B shares at June 30. The July 30 cover gives fewer Class A shares, but that net change does not disclose every intervening award, withholding and cash repurchase. Weighted-average diluted earnings shares are neither company's ending ownership inventory.²⁴

The analysis therefore does not take the later, smaller count while retaining all of the older cash. The relevant identity is useful beyond these cases. For a repurchase of q shares at transaction price a, with an observed reference price P, the required operating value changes by:

Δ required operating value = (a − P) × q.

Both shares and cash fall. At a = P, the required operating value is unchanged before fees. A lower share count alone creates a fictitious improvement if the cash cost is omitted. For example, a hypothetical $100m repurchase at $8.50, valued against a $13.60 reference, lowers the implied operating requirement by $60m, not by the full $160m marked value of the retired shares. The example uses a disclosed historical price scale, not a claim that this post-June transaction occurred. Primary issuance at a fair identical-claim price has the corresponding cash-and-claim conservation property. the realized-return investigation established why capital recycling must follow both sides of the transfer.⁵

At fixed dated basic share counts, the September 24-to-October 7 price changes imply about $311.255bn more NVIDIA equity value and $207.620m more Freshworks equity value. These are repricing calculations, not company cash receipts, new investment in equipment or realized aggregate shareholder gains. No new cash is inserted in the balance-sheet bridge because a share traded higher.¹³

2.2 NVIDIA: cash, financial investments and operating assets are not one pot

The principal bridge below is in USD billions. Each reported stock is July 26; the price is October 7. The assumed adjustments are deliberately separate from reported amounts.²

The 80% investment credit is not a probability that the holdings survive. A 50–100% credit moves the paid-renewal growth requirement only from about 22.69% to 22.52%. That narrow change is informative: the quoted operating franchise, not the existing investment portfolio, does most of the valuation work. It does not imply that a particular investment is correctly marked or can be sold when cash is needed.

The filing identifies roughly $36.9bn of public securities subject to short-term sale restrictions, and longer-restricted holdings as well. Its separate $824m warrant asset carries associated deferred-benefit considerations and is not automatically added as free net wealth. Investment gains in the income statement are not forecast as operating income and then capitalized a second time through the holdings. Trade receivables, inventory, goodwill and operating property are not separately added to the value of the business that uses them.²

The filing also identifies about $1.7bn of foreign-held liquid resources for which repatriation could trigger foreign or state taxes not yet accrued. That is an additional realization qualification, not a claim that all $1.7bn would be lost. The cash buffer is an operating-resource assumption, not a precise provision for that tax. Any separately estimated net tax belongs in the bridge. There is no inference that the excluded portion of cash or investments is worthless. Required liquidity supports operation; illiquidity and tax timing affect what can be distributed. Conversely, the full face of the deferred tax asset is not added while also assuming favorable future cash taxes. A model using a different realization assumption must change the credit, rather than silently treating every accounting asset as cash.

2.3 Freshworks: a different residual claim

The Freshworks bridge is in USD millions, using June 30 inventories.⁴

The option inventory has a weighted-average exercise price of $11.33, but no complete strike distribution sufficient for an exact October option valuation. 2.328m × max(13.60 − 11.33, 0) = $5.28456m is a lower bound on aggregate intrinsic value under the stated inventory, not a fair option value: averaging strikes loses information and remaining time can have value. 2.328m × 13.60 = $31.6608m is the gross underlying-stock bound. The main table uses this conservative option treatment without pretending all exercise proceeds are received immediately. The difference between the two endpoints is only $26.37624m; it does not resolve the more material uncertainty about future operating costs.

Freshworks' $174.047m deferred tax asset is not cash on hand. Its future use may improve owner cash, but must be valued through the taxes actually avoided, once. Crediting the full face immediately is shown only as a generous boundary: it lowers required initial growth in the reference from 20.64% to about 19.65%. That is not an assertion that every tax benefit is available, immediately usable or additional to a lower-tax scenario.

The $19.1m long-term accrued compensation is also visible. It belongs to the payroll and payment-timing perimeter, not a newly invented funded note. The modeled ongoing compensation account and retained liquidity must support it; the bridge does not deduct every operating payable as debt while charging the same ongoing costs again. Unpaid capital purchases of about $1.019m similarly matter to cash timing and must fit the investment account. A current liquidity certificate or complete post-June cash roll-forward was not obtained.

2.4 Existing awards, new compensation and the unpriced residual

The principal convention is opening award inventory plus a future new-grant budget. The opening stated RSUs/PSUs are counted in the ownership claim. The forecast's other operating costs exclude recognition expense on that old award inventory. The forecast then charges the economic value of new grants, or the cash equivalent needed to compensate employees without transferring further value from the remaining owners. It does not also grow the share denominator for those same future grants.

This distinguishes three things that are often blended: previously granted rights, the accounting expense recognized as they vest, and future awards needed to retain and recruit staff. NVIDIA's approximately $19.4bn and Freshworks' approximately $352.621m of unrecognized stock-compensation expense are not extra principal claims to deduct again after counting the opening awards. Nor are Freshworks' 116.487m shares reserved for possible grants already outstanding. Cash taxes withheld on net share settlement do not justify charging both the full gross stock claim and a second identical employee benefit.²⁴

The convention has limits. Gross stated awards ignore some forfeiture and discounting before service is performed; performance awards can settle differently from the reported inventory, and existing employee-purchase rights require terms beyond a simple share count. Freshworks identifies up to 173.6% for specified performance awards; this is not applied to every RSU. Exact current option values, maximum performance dilution and employee-purchase option values were not established. This is a full map of material claim categories with a bounded calculation, not a claim to an exact fully diluted current share price.

Let U be an additional net value required for those unresolved rights or other post-statement claim changes after matching any related cash. It must be added to the required operating value, not presumed absent because disclosure is incomplete. For NVIDIA, adding $1bn or $5bn changes the reference growth requirement to 22.590% or 22.604%. Separately varying retained liquidity from zero to $40bn moves it from 22.52% to 22.66%; zero is a permissive mathematical comparison, not a recommendation or a finding that operating liquidity is unnecessary. For Freshworks, adding $25m, $100m or $350m changes it to 20.77%, 21.18% or 22.49%. These are sensitivity amounts, not estimates or legal maxima for U. A materially larger newly disclosed claim must be modeled explicitly.

This treatment makes the direction of a missing claim inspectable. It does not use an arbitrary allowance to certify complete dilution, and it does not let a small accounting EPS denominator erase economically relevant compensation.

NVIDIA: a cash-producing franchise with a paid supply and renewal requirement

3.1 The observed business can finance much of its own growth

The July quarter reports $96.221bn revenue and $63.734bn operating income. The half-year cash statement reconciles $118.010bn net income to $74.421bn operating cash, notably removing $23.707bn net equity-security gains and including a net $27.164bn operating-asset/liability cash outflow. Thus both propositions matter: supplier profits are substantial, and rapidly expanding booked sales can consume cash before collection.²

The complete cash-stock check is:

$10.605bn opening cash + $74.421bn operating − $35.124bn investing − $27.459bn financing = $22.443bn ending cash.

The $24.896bn debt issuance proceeds are already inside that account. They cannot be added again to July liquidity to finance another repurchase or support call. Share repurchases of $39.044bn and dividends of $6.290bn are real financing outflows, not deductions from operating revenue and not new customer demand.

Operating cash less the $4.434bn cash property/intangible investment line is $69.987bn. Extending the cash boundary to $92m financing-classified asset principal, the $2.944bn Groq purchase payment, and $3.954bn recognized stock compensation leaves $62.997bn. This last number is a deliberately defined historical cash-plus-compensation diagnostic, not reported free cash flow or the forward model's margin. It still cannot determine every future replacement, acquisition or support cost. It does show that acknowledging real investment and compensation does not erase the operating franchise.

No historical repurchase is subtracted again from the forward operating valuation. Those transactions already changed the opening cash and share inventory. The new-grant budget in the forward model is a different prospective cost, replacing—not supplementing—future dilution on the same awards.

3.2 The talent and generation-renewal budget has an empirical anchor

The half-year granted 49m awards at a $183.74 weighted-average grant-date value, approximately $9.00326bn of gross grant-date value. Doubling that amount and dividing by four times the latest quarter's revenue gives 4.678%. This is not a forecast of annual grants: gross awards are lumpy, cancellations matter, and the calculation does not independently price every employee-purchase right. It nevertheless provides a more relevant anchor for future ownership transfers than assuming the much lower old-award recognition expense measures all new compensation.²

The paid-renewal reference allocates 5% of future revenue to new equity-compensation value, alongside 8% to research costs excluding old equity recognition, 2% to other operating overhead and a manufacturing/delivery cost share that rises from 26% to 30%. The assumption is demanding on future cash rather than artificially generous: research dollars grow with the business, manufacturing margins narrow, and fixed and working capital must grow.

A supplier can sustain a powerful product and software ecosystem by spending on the next generation. That favorable mechanism is consistent with the actual research expense and current order response. It does not establish a permanent margin. Rival accelerators, customer-designed systems, changing software compatibility, export limits, supplier bargaining and a slower replacement cycle can change the share retained by NVIDIA. These are mechanisms and disclosed operating risks, not assigned probabilities or proof that a particular competing design wins.²

3.3 A concrete cash-timing condition: the supply bill cannot disappear

The filing's future-obligation table includes $279bn of supply commitments, of which $92bn falls in the remainder of fiscal 2027. These are not an extra $279bn operating loss. Some spending produces inventory or prepayments rather than same-period cost of sales, and cancellation/rescheduling conditions matter. Equally, a model cannot use a low recognized cost ratio to assume that the entire cash bill is equally small.²

Consider a deliberately matched July-origin, six-month condition, not an October cash forecast. Two quarters at the latest revenue scale total $192.442bn. Recognizing a 25% delivery cost against that amount is $48.1105bn. If the $92bn bucket is paid in cash in that interval, the $43.8895bn difference must be funded as additional inventory/prepayment, other timing changes or a documented change in payment terms. “Pay the full bucket, retain the 25% recognized cost, make no corresponding asset or financing change” is an inconsistent account. The inconsistency is conditional on the payment assumption; it is not an observation that the issuer missed a payment.

A second calculation asks whether such a payment necessarily prevents useful capital allocation. Begin with July's $56.586bn cash/marketable-debt resource; assume $192.442bn customer collections; pay the $92bn supply bucket, twice Q2 non-equity-recognition R&D/SG&A ($12.904bn), a $25.4936bn cash-tax proxy, $7bn capital commitments, $18bn investment subscriptions, a $12.094bn dividend continuation, and a $9.00326bn cash-equivalent new-grant allowance; retain $20bn. The remaining amount is $52.53314bn before debt service, further collection delays, other working capital, fees and other unallocated uses.

This is positive financing room under stated receipts and costs, not an unrestricted distribution permission. A stipulated $60bn additional payout would exceed that envelope by $7.46686bn even before those remaining uses. Removing a purely hypothetical payout, improving collections, changing a documented procurement schedule or realizing an available asset would change the result. The calculation does not assert that NVIDIA planned a $60bn payout in this interval. It shows why a payout must be reconciled to cash rather than inferred from an authorization.

The grant allowance here describes a cash-equivalent old-owner-preservation path, not a forecast that all awards are settled in cash. Issuing equity instead can relieve immediate liquidity while transferring an equivalent economic claim. Own R&D cloud commitments are already inside the operating-cost perimeter; adding them again would overcharge. Previously acquired-but-unpaid capital may overlap the $7bn bucket, so it is not silently stacked on top. The prospective investments consume cash but create investment claims: their cost is not automatically a loss. This gate and the ten-year valuation answer different questions and are not combined by subtracting the same procurement bill twice.

3.4 Support exposure changes liquidity and risk, not just a headline enterprise value

NVIDIA's disclosed third-party commitments include AI-cloud capacity arrangements, intended lease reassignments and phase-dependent guarantees. Its August agreement's $105bn cap over nine phases / 4.25GW is not a July immediately payable liability. Completion, lease and triggering conditions matter; the first phase was expected to begin in fiscal 2029. The earlier approximately $3.5bn guarantee and counterparty escrow are different components. The escrow is not NVIDIA cash.²⁶

Likewise, the $36bn AI-cloud arrangement schedule, $29bn of own cloud/R&D commitments and $20bn of intended third-party lease reassignments are not automatically independent losses to sum. Third-party sales, own usage, assignment and contractual offsets can change what remains. Future equity subscriptions exchange cash for a claim; a fair-price purchase has zero incremental net present value before costs, not zero cash requirement. An overpriced or strategically necessary low-return purchase reduces owner value; repeated support costs must reduce the retained margin or enter a separate net-loss path.

The main valuation credits existing investments separately and does not grant extra value for future subscriptions. The favorable reference assumes future strategic exchanges do not destroy net value while the cash gate still finances them. That assumption is unverified; the adverse sensitivity increases continuing renewal/support costs rather than treating all announced commitment faces as current losses. A guarantee's gross funding call and its net loss after recoveries remain different quantities. None of this establishes receipt of the separately qualified NVIDIA–Energy Global prepaid-forward payment: the accepted finding remains obligation established, receipt unverified, with no inference of nonpayment.⁷

3.5 Test the strongest contemporary favorable argument

The October 1 NVIDIA explanation emphasizes productive, durable and reusable AI capacity, argues that power efficiency can increase earning capacity, and invokes continued older-design use. It is a relevant original supplier argument with a commercial incentive, not an independent forecast of customer cash.⁸

the older-capacity investigation supports selective continued use; the efficiency investigation supports genuine technical cost improvements. Those are reasons to give a continuing franchise value rather than assume it becomes worthless after a fixed short window. But the efficiency investigation also demonstrates that preserving revenue after a price cut can require another paid capacity step, and that extra review can absorb a seemingly large component saving. The missing link in “more tokens means more revenue” is paid, adequate work at the realized price. Better throughput supplies an opportunity, not a measured demand elasticity.

NVIDIA's September 28 additional $150bn repurchase authorization, leaving $235bn authorized with execution expected through fiscal 2028, is similarly meaningful but limited. It expresses a capital-allocation intention and confidence in resources. It is neither a cash receipt nor a compulsory immediate bill. The supply/collection gate above demonstrates what would have to finance an actual payout. A successful repurchase can redistribute the claim efficiently; it cannot create the same cash a second time.⁹

Freshworks: valuable recurrence, a narrower present claim on profit

4.1 The buyer relationship is real; the AI-only margin is unidentified

The principal empirical basis remains the accepted the customer-capture investigation company-level case, freshly matched to the original June filing for this ownership analysis. The existing-customer revenue bridge shows expansion rather than all growth coming from replacement logos. Reported retention and constant-currency retention move differently, so neither is silently chosen as the sole demand trend. Paid AI products and enterprise attachment strengthen a commercial mechanism; they do not identify an AI-only cash cohort.¹⁰⁴¹¹

The latest quarterly statement reports $237.377m revenue, $201.281m gross profit and $6.060m operating income. That is a profitable quarter in a business with substantial recurring sales, not a permanently established 85% owner margin. H1 operating income is −$2m. Purchased hosting, research, sales, administration and compensation explain why the gross and residual claims differ so greatly.

The preceding full year gives a useful additional check. Freshworks reported $13.205m operating income in 2025, but the release also identifies an approximately $41m stock-compensation reversal on the executive chairman's departure. The same annual accounts include a large tax benefit. It would be misleading to perpetuate that reversal or tax benefit as normal annual profitability. It would be equally misleading to deny the $242.370m actual operating cash merely because accrual profit contains these adjustments.¹²

Management's August revenue outlook is $963.5–966.5m for 2026, about 15% growth, and its non-GAAP operating forecast is stronger than the current GAAP margin. The revenue guidance is evidence of management's near-term expectation, not an independent five-year forecast. A forecast that excludes future compensation cannot substitute for the new-grant budget required to value the ordinary claim.¹¹

4.2 A complete cash distinction, including capitalized compensation

H1 operating cash of $120.912m, less $8.942m cash property investment and $4.850m software cash, leaves $107.120m before acquisition and shareholder uses. the customer-capture investigation already established this genuine cash result and the correct −$4.059m signed working-capital change. The source's contrary prose label is not revived as evidence of an inflow or failed payment.¹⁰⁴

For the ownership diagnostic, deduct the $81.770m operating cash-flow stock-compensation adjustment and $1.165m capitalized stock compensation. The result is $24.185m for the half-year. The additional capitalized amount matters: it was not included in the cash software line or the operating compensation adjustment. The one-thousand-dollar difference from the compensation note's total is rounding, not a missing payment. This diagnostic is not also deducted in the award-inventory valuation; doing so would charge old awards twice.

The matched stock-of-cash check remains:

$632.250m opening cash including restricted + $120.912m operating − $27.626m investing − $221.334m financing − $8.383m exchange effect = $495.819m.

Cash repurchases of $207.411m and the $56.913m acquisition cash-flow line are therefore already reflected. The $204.5m repurchase-cost description excluding excise tax and the $8.50 rounded average price have different precision/timing from the cash statement; no extra residual is invented as shareholder profit. Past repurchases are not deducted a second time from forward cash or added as a separate asset.

4.3 The new calculation: grants, not just recognition expense

Freshworks granted 20.155m RSU/PRSU awards at an $11.00 weighted-average grant-date value in H1, a calculated $221.705m gross grant-date amount. That amount is already more than 23% of the latest quarter's annualized revenue scale. It must not simply be doubled into a perpetual annual cost: annual refreshes, acquisition employee grants, forfeitures and performance conditions can make the half-year unusually large. FireHydrant's replacement employee awards are explicitly post-combination compensation rather than purchase consideration.⁴

Nevertheless, a favorable ownership argument must explain how the ongoing grant policy compares with this observation. Lower recognized expense can reflect old awards, cancellations and the prior executive arrangement; it is not direct evidence that new ownership transfers have become small. The paid-renewal model's 15% initial / 12% mature new-grant budget and the stronger path's 10% mature budget are therefore conditional management/talent outcomes—not measured normalized rates.

At 15% initial revenue growth, the reference's other costs and capital assumptions need a mature new-grant share of only 6.647% to match $13.60 at 11% / 3%. That corresponds to approximately 25.35% economic operating margin, instead of the reference's 20%. Alternatively, more growth, a larger reduction in selling costs, a lower required return, more favorable taxes or another supported combination can do the work. The solved 6.647% is not an instruction to cut compensation or a claim that such a cut is feasible.

A disciplined adverse comparison is to charge one H1 gross-grant amount per year at the current revenue scale, or 23.349% of revenue, while retaining the reference's other cost improvements. That policy produces only about $575m operating value under the same 15% growth and 11% / 3% continuation assumptions. This is not an estimated annual policy or expected price; it shows that calling compensation “noncash” cannot rescue an otherwise incompatible owner-cash account. The decisive observation is the future grant/forfeiture/retention policy and service performance—not the label attached to adjusted free cash flow.

4.4 Acquiring and maintaining the workflow is not free

FireHydrant's stated January purchase consideration is $88.737m, including $4.349m acquired cash. The June account records $9.043m unpaid acquisition-related liabilities and the half-year cash statement reports $56.913m net acquisition cash. Those disclosed figures leave $18.432m of purchase-payment/timing allocation unreconciled in the inspected material. The difference is not assigned an invented earlier payment date or an additional current creditor. The observed holdback is deducted; the incomplete historical join remains explicit.⁴

The acquired customer relationships, technology and goodwill are part of the operating franchise. They are not added to the value of the revenue they help produce. Future renewal spending is separately funded in the model: research continues, software and customer/workflow capabilities must be maintained, and growth consumes additional capital. An acquisition can be a productive way to obtain capability; repeatedly buying indispensable capability still costs money. Treating the old purchase as sunk for a new buyer does not make future acquisitions free.

Freshworks also discloses $199.6m of cloud/service/sponsorship commitments through 2029, operating-lease liabilities of $34.594m, and additional uncommenced leases. These commitments constrain the cost/timing path. They are not mechanically added as debt after the forecast has already paid hosting, distribution and rent. Conversely, a small owned-capital balance does not mean the underlying compute is capital-free: much of it is purchased as service from another owner.⁴

The favorable mechanism is a lasting workflow and distribution advantage: customers may keep paying for integration, reliability and accountability while underlying model service gets cheaper. the buyer-budget investigation shows why substitution can be economically real without creating equivalent net new sector spending. The adverse mechanism is that customers or competitors capture the savings, or that serving and selling costs do not decline enough. Neither a software-replacement slogan nor a list of paid accounts identifies which outcome dominates at the required ownership price.

What the accumulated economics permits—and what it does not supply

These ownership tests use the earlier findings as economic constraints. They do not transfer a measured margin or study population between unrelated companies.

Completion and available resources. the completion investigation established legitimate financing before final acceptance as well as milestone, reimbursement and reserve restrictions. It did not obtain a complete current cost-to-complete certificate or make every restricted dollar available. For NVIDIA, this means a customer's announced capacity is not automatically installed, billable demand; procurement and support can precede final customer cash. For Freshworks, buying cloud service transfers the physical investment to a supplier rather than eliminating it. Neither company's valuation inherits Core Scientific's reserve formula or IREN's acceptance terms. the completion investigation supports asking who bridges the timing, not a universal delay probability.¹³

Buying and substitution. the buyer-budget investigation obtained a signed public-sector extension and buyer-originated accounts of moving marketing production in-house. The preferred private-buyer route did not yield a complete AI-only renewal/cost ledger; the public procurement alternative was substantive, not an imagined private invoice. The care case's throughput observation did not establish causal payroll savings, and the purchased allowance did not establish consumption. Klarna's attributed savings did not isolate all internal costs. These limitations matter: neither result calibrates NVIDIA's trillion-dollar demand nor Freshworks' product renewal. The supported mechanism is that a buyer can rationally purchase more useful output or replace another bill, while the winning supplier receives only part of the benefit.¹⁴

Efficiency and the owner of the saving. the efficiency investigation could not match the care or creative-production cases to a complete public technical-cost comparison and therefore used its documented OCR/service alternative. It retained the original paper's inconsistent cost denominator rather than laundering it into a precise all-in margin. The resulting contribution and review-work thresholds establish conditional possibilities, not an audited model-provider profit. A second agent-workflow comparison reinforces the importance of accepted output rather than counting calls. The transferable lesson is that quality, review, distribution, utilization and continuing development determine who captures an efficiency gain. The numerical OCR tariff, throughput and overhead budget are not substituted for either principal company's costs.¹⁵

Who receives the financial return. the realized-return investigation documented issuer funding, secondary-holder proceeds, earlier tender receipts, a historical protected-share allocation and a debt-funded parent transfer/bridge repayment. It did not obtain complete private LP cash/basis records; it used the completed public-transaction alternative. The historical Series C-derived put terminated in September 2025 and is not a current liability. Actual seller receipts are favorable realization evidence, but neither those receipts nor a financing-derived parent transfer demonstrates the return on a later common-stock purchase. The lesson for these ownership tests is to match rights, cash and subsequent funding—not to remove all new finance from a viable growth business.⁵

The earlier favorable cases remain constraints too. Strong paying parents can protect particular creditors; older equipment and facilities can continue to earn; an investor can realize cash before every company is self-financing. These are reasons not to infer failure from construction, refinancing or uncertainty alone. They do not supply a free terminal value. Microsoft's payer resources and IREN's project conditions remain contextual evidence, not a third and fourth newly valued security in this report.^{1617181920621}

The strongest updated proposition is therefore: useful work, recurrent buying, productive financing and investor realization are all documented, but their distribution across claims remains uneven. The missing joint measurement is economically consequential. A customer may retain the efficiency gain, an application may retain it through workflow ownership, a model may lose pricing power, and a hardware supplier may sell more capacity. None of these allocations is established by summing the parties' revenues.

The cash path and the price required to own it

6.1 One annual cash account, not independently favorable sliders

The model begins at an annual revenue scale of $384.884bn for NVIDIA and $949.508m for Freshworks, four times each latest reported quarter. This is a transparent starting scale, not an observed full year, trailing-twelve-month result or forecast. The Freshworks scale is below management's $963.5–966.5m annual outlook; using that outlook would not justify assuming the same growth for five further years. The NVIDIA scale does not automatically incorporate its $108bn next-quarter revenue guidance. A sensitivity uses twice each reported half-year instead.²⁴²²¹¹

The future annual periods start at the analytical reference date, not at the beginning of either issuer's already partly completed fiscal year. Cash is received at model year-end. Consequently, year five and year ten are approximately five and ten years after the reference, not relabeled fiscal-year guidance. The latest operating scale is combined with the explicitly dated opening inventory; unresolved intervening cash and claim changes remain in the bridge sensitivity.

For each company:

Economic operating income = revenue − product/hosting cost − continuing research − distribution/administration − economic cost of new compensation.

Retained operating cash = economic operating income − cash taxes + depreciation/amortization − gross maintenance/renewal − growth fixed investment − growth working/customer-acquisition capital.

The expense categories include non-stock payroll, rent, support, reliability and distribution. Existing award-recognition expense is excluded under section 2.4's claim convention. Future compensation is then paid through the new-grant budget. The cost model is not management's non-GAAP margin: exclusions from a company presentation do not automatically become exclusions from this cash account.

Gross maintenance/renewal is D&A + an additional revenue-linked renewal allowance. Growth fixed investment is k_fixed × positive revenue increase; growth working/customer-acquisition investment is k_working × positive revenue increase. A revenue decline does not automatically liquidate inventories, release prepaid costs or produce a cash refund. The additional capital account funds incremental assets and customer-acquisition resources; it does not deduct an expensed sales payroll twice. Capitalized acquisition costs already embedded in the operating expense envelope are maintained through the replacement allowance; incremental investment is separately charged as the business expands.

The D&A reference is 1.2% of revenue for NVIDIA and 3.4% for Freshworks. It helps expose gross renewal investment, including purchased technology and software, rather than letting an addback create free cash. Because matched maintenance includes the same D&A, changing both by the same amount leaves valuation unchanged. That cancellation is deliberate and tested. The economically decisive quantities are the additional renewal allowance, incremental growth investment and fully paid operating expenses. No accounting depreciation life is asserted to be the actual competitive life of a GPU or workflow.

Lease convention: operating leases remain rented operating inputs. Their payments must fit the expense envelope; neither the right-of-use asset nor its matching liability is separately capitalized into the bridge. The NVIDIA historical diagnostic separately charges financing-classified asset principal; its disclosed finance-lease expense is not material. No future leased fleet is assumed free: future rented service is paid in expenses, while modeled owned capital is paid in investment. A different capitalization convention would need to change expenses, investment and claims together, not add lease debt alone.²⁴

Financing convention: this is operating cash before funded-debt distributions, paired with a debt-at-face opening bridge. Debt interest and principal are not deducted in operating cash and again through the full debt claim. The 20% cash-tax assumption is applied to positive operating income without a separate interest tax shield or automatic tax refund on a loss. This is a simplified operating-value comparison, not a calibrated weighted-average cost-of-capital appraisal. Debt settlement premiums, separate net support costs and changes in outstanding claims remain additional adjustments where material.

6.2 The six complete economic packages

The following packages are research assumptions constrained by the disclosed starting business, not issuer forecasts or probability-weighted outcomes. Percentages are shares of revenue. Arrows show the initial-to-mature expense transition. NVIDIA costs reach the mature setting in year ten; Freshworks in year five. Revenue grows at the stated initial rate for five years, then slows in equal growth-rate steps to its continuing rate in year ten. The NVIDIA erosion case instead begins with a 15% revenue fall, followed by 3% growth in years two–five.

The NVIDIA renewal package allows gross margins to narrow and spends more on research relative to revenue than the latest non-stock quarterly rate. The stronger package requires a durable competitive position and scalable distribution; its lower cost shares still imply much larger absolute expenditure. The erosion package combines weaker sales, lower product margin, heavier competitive expense and greater capital intensity. It remains a profitable continuing business, not a constructed insolvency case.

Freshworks' initial operating cost shares are anchored near the latest disclosed non-stock cost structure, rounded conservatively rather than presented as measured next-year expenses. The latest restructuring is substantially complete according to the filing; the forecast does not perpetuate that one episode, but ongoing personnel change and product renewal must still fit the expense and renewal budgets. The remaining restructuring payable belongs in retained resources and the near-term operating cash account. A repeated material restructuring would require a new cost, not a permanent “one-time” exemption.⁴

The key favorable changes are explicit: renewal assumes sales expense falling from 40% to 31%, general administration from 10% to 8%, and new compensation from 15% to 12%. Stronger execution reduces these further. Those changes cannot be justified merely by naming AI automation: they require retained service quality, successful customer acquisition/retention and a sustainable talent policy. The adverse package does not assume total loss of the installed customer base. It shows that a useful, growing service can remain worth much less when little of the revenue reaches the residual owner.

6.3 Continuing value still pays for the next generation

The reference uses an 11% nominal required return and 3% nominal continuing revenue growth. For years after ten, mature costs, taxes and capital intensity continue; growth still requires additional investment. The terminal calculation is:

Value at year ten = year-eleven retained operating cash / (required return − continuing growth).

Year-eleven cash is calculated from the whole paid account, including growth investment, not by multiplying year-ten revenue by an untouched peak margin. At the mature 3% growth rate, the renewal packages retain approximately 41.83% of revenue at NVIDIA and 13.63% at Freshworks after modeled tax, extra renewal and growth capital. Those are conditional owner-cash margins, not observations. A model that chooses them must defend the continuing manufacturing/workflow advantage and the budgets that preserve it.

A 3% nominal continuing rate is a modest-growth assumption for a mature continuing business, not a forecast that either firm maintains today's market share forever. Zero, 2% and 4% alternatives reveal its importance. Even 3% cannot grow faster than the addressable economy indefinitely without a change in scope or share; the model is an analytical continuation, not a literal infinite corporate history.

The finite comparisons value only the first ten or thirty years of the same fully funded cash path. They answer how much support comes from the chosen window. They do not assume the business closes, every asset is worthless at the cutoff, or uncertainty makes a continuing franchise valueless. Conversely, a large continuing-value contribution does not validate itself merely because the business is profitable today.

6.4 Required return is a condition, not a verdict

The Treasury's October 7 curve reports a 5.28% ten-year nominal yield. The 9%, 11% and 13% operating-return sensitivities correspond to differences of approximately 3.72, 5.72 and 7.72 percentage points. They span materially different compensation for uncertain business cash; they are not estimated equity premia, company betas, default probabilities or a claim that a risky business can be discounted at a government yield.²³

A strong balance sheet and durable franchise can support a lower required return than a fragile financing-dependent claim. Concentration, competition, contingent support and uncertainty about future owner cash can justify requiring more. NVIDIA and Freshworks do not necessarily deserve identical rates. Using the same grid is a controlled comparison of assumptions, not an assertion of equal risk. Neither the completion investigation's utility documents nor the buyer-budget investigation's buyer benefits supplies either company's required return.

An investor's particular hurdle cannot establish that the market is wrong. The question here is narrower and harder: does the price require an economically incompatible combination under a reasonable range, or is it supportable under plausible but unproved conditions? The answer differs between a failed cash identity and an ambitious forecast.

What the prices require, and how much arrives late

7.1 Continuing businesses have value in both favorable and adverse paths

The following table is a calculation of the six packages above, all at the reference 11% required return. The common-equivalent numbers are scenario outputs, not price targets or a forecast range. They use the stated opening award inventory and the Freshworks gross-option bound. No probability is attached to any row. The separate-asset credit is held fixed to isolate the operating mechanism; a correlated investment markdown or additional support loss must also be applied rather than assumed away.

The adverse paths still value useful operations. NVIDIA's erosion package remains a large profitable business; Freshworks' still has positive retained cash and an installed franchise. A lower value does not imply that every product disappears, a bond defaults or the entire difference is a realized investor loss. Conversely, a favorable scenario exceeding the observed reference does not make its assumptions true.

The ordinary-share reference includes no separate premium for control, no investment-specific transaction fees or personal taxes, and no unsupported claim that the entire separate-asset allowance could be distributed immediately. Different claim realization, timing or taxes would change it. To reproduce an investment-specific result, change those assumptions together with the related claims.

7.2 Solving the economically compatible requirement

Here the initial growth rate is solved, holding each package's complete cost, tax and capital account fixed. The revenue scale and mature cost structure change together through the explicit transition, rather than allowing a low-cost steady state on the first day. The cash column is retained operating cash before funded-debt distributions. Money columns are USD billions; the Freshworks amounts should not be mistaken for trillions.

NVIDIA: about 22.6% growth in the paid-renewal package is less dramatic than extending the latest reported expansion rate indefinitely, but produces enormous absolute sales. The stronger package needs less growth because it retains more of each sale after its paid costs. That is a credible economic alternative for a competitively strong supplier; persistence of both scale and rent remains the condition.

Freshworks: the current near-15% growth observation does not settle the price. A plausible 29% mature economic operating margin supports it at a lower growth rate than the 20% margin reference. The 29% margin requires substantial, sustained selling and compensation efficiency, not just lower serving costs. It has more assumptions than an extrapolation of current complete operating profit, but is not contradicted by the fact that a software franchise can scale distribution.

7.3 A finite horizon is a dependence test, not a demand to shut the firm

The first-ten-year comparison excludes a large amount of still-performing cash, so it is not a neutral fair-value rule. The thirty-year comparison is more informative about extreme remoteness: it raises the NVIDIA renewal growth requirement from 22.59% to 25.52%, and Freshworks' from 20.64% to 23.98%. A substantial part of value lies after year ten, but a much smaller part requires cash after year thirty. The blanket claim that these prices are supported only by an infinite, costless future overstates the model's dependence.

This is not evidence that the competitive position survives thirty years. It identifies the observation that matters: whether successive generations of product, customer relationships and reinvestment keep producing retained cash, rather than whether today's specific hardware or product version lasts indefinitely.

7.4 The rate and continuation sensitivity changes the interpretation

Each cell below is required initial five-year revenue growth using the paid-renewal cost/capital package. The rate and continuation settings are alternative assumptions, not confidence intervals.

At 9% / 3%, Freshworks' paid-renewal requirement is below the roughly 15% current revenue-growth reference. At 13% / 3%, its required growth is much higher. That makes required return and durable margin conversion a genuine unresolved valuation question, not a technical footnote under a predetermined negative verdict. The same principle applies to NVIDIA: a demanding price can remain supportable under strong execution and a lower opportunity cost of capital.

A lower hurdle does not excuse inadequate cash. Every cell still pays its operating, talent and investment costs. A higher hurdle does not demonstrate a bubble merely by producing a lower present value. Evidence that would justify a different rate must concern the cash risk, durability and outside alternatives, not a customer-survey response or the investor's preferred conclusion.

7.5 Costs and capital can be more consequential than the opening asset allowance

These comparisons retain all other inputs. They are not assigned simultaneous probability or added as independent losses. In an adverse economic state, weaker revenue, higher support needs and investment markdowns may be correlated; a combined case must change all affected quantities and preserve their payment timing. The complete erosion packages already join growth, margins and capital intensity. Portfolio/guarantee stress is an additional specified claim adjustment, not a second count of the same supplier or customer shortfall.

Freshworks' valuation is particularly sensitive to paid capital and competitive renewal because the residual owner margin is much smaller. NVIDIA's existing investment-credit range makes relatively little difference to its growth requirement compared with enduring operating cash. This is an analytical ranking of these model sensitivities, not an empirical volatility or loss-probability estimate.

Judgment: what has support, what requires execution, and what actually fails

8.1 NVIDIA's demanding path is plausible, not demonstrated

The favorable evidence is unusually substantial: operating earnings, operating cash after investment, paid research, supplier commitments, broad system revenue and documented capital access. A supplier need not wait for each downstream asset's lifetime return before earning a valid sale. Its software, systems and distribution can allow it to capture some of the value of cheaper useful computation. A price-supported path therefore deserves serious weight, rather than setting future value to zero because competitive risks exist.²²²

The renewal package is not a free-growth construction. At its illustrative 25% growth, year-five research expenditure is approximately $93.97bn, distribution/administration $23.49bn, new compensation $58.73bn, and gross fixed investment $49.33bn. The first ten years fund roughly $1.03trn of research in total. Suppliers are paid inside cost of goods and incremental working investment; existing and new owned technology must be renewed. These absolute budgets make explicit what a small percentage means at the required scale.

That does not establish physical feasibility. The same illustrative path has year-ten revenue 5.29 times the starting scale. If the realized price of a comparable unit fell 10% annually, producing those sales would require 15.17 times as many paid comparable units by year ten. The 10% decline is an assumption for this quantity identity, not an observed NVIDIA price series or a demand elasticity. A system sold is not interchangeable with an API token, and a claim of more output per watt cannot be converted into this revenue without price, mix, installation and customer-payment evidence.

The original filing also supplies the serious adverse case in its own terms: current supply constraints, possible architecture-transition delays, customer shortages of land/power/capital, and extended payment terms even for some investment-grade buyers. Management attributes its new partner commitments to those constraints. This is disclosed economic exposure, not evidence that any specified customer failed to pay. There is a documented supply response, but no inspected ten-year wafer, packaging, memory, power, site-acceptance and end-customer budget account that independently verifies this scale. That is the limit of the physical test—not proof that the resources cannot be supplied. the completion investigation shows where unaccepted capacity and conditional funding can break the transition; the efficiency investigation shows why technical capacity and paying demand are different. The required resource/cash path is feasible within the specified cost model, and its unobserved physical completion remains a condition.¹³¹⁵

The adverse case has a concrete mechanism: improved alternatives or concentrated customer retrenchment can lower realized prices or growth while competitive research and ecosystem support remain costly. Securities linked to the same customers may also lose value. NVIDIA's own substantial cash and adjustable distributions contain part of that risk; they do not guarantee enduring rents. A recurring two-point revenue burden for renewal/support materially raises the required growth, even though a one-time markdown of existing investments has a smaller effect on the overall price.

Judgment: the evidence does not justify calling the selected NVIDIA price unsupported merely because it is large, long-duration or connected to financing-intensive customers. A costed, nonzero continuing franchise supports a credible path. The unresolved proposition is very large, durable paid demand at high retained margins—not whether current supplier earnings exist. Evidence of repeated customer payment, generation transitions with maintained realized margins, and supplier spending converting into cash would strengthen the price-supporting interpretation. Growing commitments and receivables without comparable paid deployment, or durable erosion of the retained margin, would weaken it. No crash date or probability follows.

8.2 Freshworks' price is a conversion question, not a verdict that the workflow disappears

Freshworks has a real recurring franchise. The latest positive operating quarter and reported cash strengthen the favorable case; the prior full-year reversal and tax benefit prevent treating every recent earnings improvement as recurring. A workflow business can retain customer value even when the underlying model becomes cheaper, especially where integration, accountability and distribution remain expensive to replace. the buyer-budget investigation's substitution and the efficiency investigation's accepted-work boundary support that mechanism, not its numerical margin.^4121415

The favorable cost path still funds meaningful absolute work. In the renewal package, year-five research is approximately $267.37m, sales/marketing $592.04m, new compensation $229.18m, and gross fixed investment $118.49m. In the stronger package, research is approximately $307.15m and sales/marketing $637.92m, despite their smaller revenue shares. Thus the favorable case does not obtain a 29% mature economic operating margin by eliminating distribution or product work.

The burden is nevertheless less directly established than the recurring sales. The model requires lower selling cost per revenue dollar and a new-grant policy below the one-half-year gross observation. Those improvements can coexist with larger absolute employee and marketing budgets as revenue grows; they cannot be claimed from a drop in old-award expense. Stronger revenue growth accompanied by unchanged high acquisition and compensation intensity need not deliver the same owner value.

The modeled first-year economic funding requirement is only about $15.43m in the renewal path and $18.99m in the stronger path. The retained $150m liquidity assumption can cover those annual deficits; subsequent modeled cash turns positive. This is an economic cash-equivalent budget, not a forecast of actual cash settlement of awards or proof of sufficient intraperiod liquidity. It does show that the favorable paths do not depend on an unspecified enormous new funding round. Issuing compensation shares instead of paying the equivalent in cash changes who owns the future, not whether compensation has economic cost.

Judgment: the observed price needs meaningful conversion of customer relationships into owner cash; current gross margin and adjusted cash alone do not establish it. But the stronger, fully costed business package and the lower-required-return comparison make “demonstrably overvalued” too strong. The narrower supported vulnerability is compensation and acquisition/distribution costs being treated as though they disappear when calculating shareholder returns. Subsequent gross grants, forfeitures and net economic dilution, together with retained customers and complete operating cash, are more decisive than the next isolated adjusted-margin headline.

8.3 Separate contradiction from lack of evidence

A model can reject a particular combination without rejecting all plausible explanations of a price. Conversely, the existence of one favorable set of assumptions does not demonstrate that the market has sensibly weighted it. No calibrated distribution of future outcomes is available here. These six packages are therefore not an “expected value,” an uncertainty band or an estimated chance of loss.

8.4 The connected whole-question assessment

The working bubble question concerns whether ownership expectations can be supported by economically compatible future cash, not whether AI is useful or every investment is riskless. The accumulated work establishes more genuine economic substance than the strongest blanket negative narratives allow. Customers buy useful work, application and supplier businesses retain value, selected assets earn again, protected creditors can be paid, and investors have received real cash. More finance can be productive risk-bearing rather than merely postponing an inevitable collapse.^247102065

There is also a supported vulnerability: some attractive-looking claims require margins, timing, capital recovery or ownership allocation that the headline omits. this investigation makes that concrete at two new profitable layers. the completion investigation's completion resources cannot be spent twice; the buyer-budget investigation's buyer savings are not all new industry revenue; the efficiency investigation's serving contribution does not pay every continuing cost; the realized-return investigation's early realized gain does not validate the later investor's entry price. These are related constraints, not fifteen independent observations of the same crisis.

The evidence presently distinguishes strong and weak arguments better than it measures a sector-wide speculative premium. It rejects the claim that current profits alone settle valuation and also rejects inferring a bubble from every financing dependency. It leaves a substantial class of ambitious ownership prices within rationally contestable uncertainty. That is a narrower finding than “nothing can be known”: actual costs, rights and cash identities eliminate specific unsupported accounts, and observed operating strength makes other paths more credible.

The new weighting is modestly more favorable to a continuing commercial foundation, while more demanding about the ordinary owner's share. The principal unresolved risk is not that every customer dollar is fictitious. It is that future growth, competitive rents and cash realization disappoint a price that capitalizes them, with financing and contingent claims determining where the disappointment is borne. This report supports that vulnerability's existence and mechanism, not its prevalence or aggregate loss size.

8.5 Coverage must remain visible

The repeated CoreWeave-related chain is valuable depth, not multiple independent industry samples. GECC's concentration cannot be scaled into global credit exposure. NVIDIA and Freshworks are two securities with different cash models, not proxies for every asset in their respective layers. Neither prices nor intermediate sales, overlapping commitments and marked holdings are added into an industry “bubble size.” Employment, environmental and fiscal welfare questions also require evidence beyond the financial claims studied here.

Methods, evidence limits and calculations

9.1 What this analysis adds

This investigation adds matched claim inventories, new-grant compensation analysis, complete operating-value requirements, a near-term procurement/liquidity consistency test, paid continuation/finite-horizon comparisons and a connected valuation judgment for NVIDIA and Freshworks. It does not refresh the ownership-price investigation's Microsoft grid and call it new work. the customer-capture investigation's Freshworks operating observations are inherited, freshly checked against the original statement and extended into a new ownership/cost analysis; the recurring-business evidence is not claimed as a rediscovery.

Earlier research remains unchanged and dated. In particular:

The NVIDIA prepaid-forward receipt qualification remains intact; the January CoreWeave ownership block is not an October holding. The later September 22 CoreWeave notes closing remains established, without treating proceeds as current unspent cash. the older-capacity investigation's Austin calculation retains its March 2025 historical anchor. the joint-cash investigation's April 8.25% issuer illustration and May 7.75% executed coupon remain distinct under the completion investigation's already completed status repair. the realized-return investigation's historical put and its September 2025 termination remain distinct. None of these is relabeled a new this investigation correction.^71820621135

9.2 What the companion files do

ownership-economics-data.csv is the single source-linked numerical ledger. Its fields are id, entity, metric, value, unit, period, status, source_id, source_url, locator, note. Money and shares are in millions unless a unit says otherwise; percentages are stored as fractions. Reported observations, market references, issuer forecasts, authorizations and research assumptions have different statuses. All new external inspection was on October 8; the period field preserves each actual observation date. A rounded or incomplete source quantity remains so even when the code reproduces its arithmetic precisely.

ownership-economics-calculations.py requires only the Python standard library. Run it beside ownership-economics-data.csv, with its matching CSV, or with an explicit path:

python ownership-economics-calculations.py --data ownership-economics-data.csv
python ownership-economics-calculations.py --data ownership-economics-data.csv --json

It prints the claim bridges, scenario outputs, price-implied growth, rate/continuation grid, compensation boundary, cash reconciliations and sensitivities. The JSON option supplies each annual cash row for inspection. Numerical precision supports reconciliation; it does not imply comparable precision in future cost or growth assumptions.

Core equations use floating-point arithmetic for annual paths and bisection for price-matching growth; source reconciliations use Decimal. The required-growth solver searches 0–100% initial annual growth and reports a failure when the target is not bracketed. It does not force every possible claim into a plausible positive-growth solution. The principal modeled growth, cost, tax and reference-return assumptions can be changed through the explicit ledger; the documented comparison-grid points and sensitivity increments are in the code. Changing a company source requires retaining its new date and status.

9.3 Limits that affect the result

Current inventory: these are dated-share/award/asset reference claims priced at a subsequent close, not complete October settlement registers. Intervening repurchases, issues, awards, taxes and investments require matched cash adjustments. Gross awards and the option bound are conservative common-equivalent treatments but not exact fair values. The U sensitivity does not estimate or cap all missing rights.

Operating normalization: latest-quarter annualization, future cash taxes, new-grant policy, competitive renewal and incremental capital are assumptions. The source reports do not isolate a complete maintenance-versus-growth bill or all replacement needs. The renewal budgets show what is paid and what would have to change; they do not certify that the budgets are sufficient for every future technological transition.

Funding and timing: an annual cash-equivalent surplus is not a subsidiary liquidity certificate. The procurement gate deliberately leaves debt service and other unidentified uses explicit. It does not reconstruct October cash or certify any permitted payout. Supplier commitments, acquired assets, customer advances, restrictions and investment subscriptions cannot be treated as interchangeable cash flows.

Investments and support: carrying values are not necessarily realizable net proceeds. Future subscriptions are net-value-neutral only under the disclosed fair-exchange assumption; support calls can be correlated with weaker operations and marks. The main operating scenarios do not claim a calibrated distribution of those outcomes. A gross guarantee ceiling is not a debt face or expected loss to insert mechanically.

Coverage and inference: no private bank records, paid data, external contacts, executed trades, independent human source audit or reader-comprehension study was obtained or performed. Public figures were checked against the stated originals and price source; predecessor investigations were used at their original evidence boundaries, not presented as a new audit of every underlying historical document. Some attempted annual-filing/search routes were inaccessible or returned unrelated results; the original annual release and current quarterly filing provide the stated Freshworks evidence, not a claim that an unavailable filing contains no additional information.

These tests do not establish a unique fair price, an expected future return distribution or a systemic loss estimate. The useful result is the documented set of cash/claim requirements and the judgment about which arguments they support or rule out.

Sources and calculations

Evidence boundary: 8 October 2026. The notes identify original documents, reproductions, attributed claims and access limits. Earlier investigations keep their own dates. Source links are deliberate links, not live data feeds.

Source-linked inputs (CSV) · Reproduce the calculations (Python)

The reader companions retain the supplied numerical records and model equations, with public provenance and descriptive filenames. Observations, accounting reconciliations and conditional scenarios remain different result types. Running the code checks arithmetic, not the empirical assumptions or completeness of the claim inventory.