Shaduf.Research preview
Is AI a Bubble?/Research — all reports
Skip to content

Is AI a Bubble?

Research — all reports

The complete catalogue: current investigations, earlier reports, evidence dates and source-linked data.

All dated investigations and their downloads are retained. Recurring-business evidence: ; annual progress, matched half-year cash and conditional owner allocation retain distinct meanings. Latest substitution evidence: ; customer-funding evidence: ; ownership economics: , with October 7 price references. Reporting periods, market dates, transactions and retrieval boundaries remain distinct.

Updated as new research is ready

All current and historical investigations are collected here, with their dates and downloads. The whole-system foundation connects the question; the dated chapters deepen and test it. Earlier current investigations do not become History merely because a later chapter arrives.

Use Spending & obligations, Investment returns and Evidence & tests for subject-led explanations. This catalogue takes you directly to the complete reports and their data.

Find the evidence for your next question

Find the evidence for your next questionA question-led map of eighteen current investigations. It groups related research without treating connected cases as independent population samples. Four historical reports remain listed below.FIND AN ANSWER, THEN FOLLOW THE COMPLETE EVIDENCEWhat is the whole assessment?Whole-system foundationWhy does the buyer pay—and who retains the benefit?Demand, customer capture, buyer budgets and full-cost efficiencyWhat has to be completed and funded?Physical recovery, completion, joint cash and payer flexibilityWhat survives, and which creditor or holder is exposed?Continued use, loan protections and holder fundingWho already received cash, and what remains owed?Completed ownership transactions and capital recyclingWhat future economics does the ownership price require?Earlier ownership models; NVIDIA and Freshworks costed pathsWho can fund the next purchase?Customer accounts, matched quarterly budgets and contract rightsCan the workload move before the bill falls?Deployed alternatives, contract timing and supplier returnsEighteen current investigations; four historical reports remain below.Can operating progress reach the continuing owner?Recurring customers, paid renewal, repurchases and compensation claims

On a narrow screen, scroll the figure horizontally to inspect the labels.

This map organizes questions, not a predicted sequence or an industry sample. The catalogue below preserves each full investigation, its evidence boundary, sources and numerical companions.

The Customer demand and Capacity & recovery guides have distinct, maintained routes. The investigations below retain their dated evidence and original addresses; older current research is not automatically History.

Current research

Current investigation ·

Recurring customers can finance better owner conversion—but not every buyback

Annual operating progress, paid selling and renewal, and actual ownership contraction strengthen a recurring-franchise case. The joined cash and gross-claim account shows why the latest half-year does not establish a sustainable continuing-owner result.

October 17 retrieval boundary; Freshworks annual and matched half-year windows remain distinct. Gross units are not fair-valued fully diluted capitalization. Cash requirements use historical prices and explicit acquisition, portfolio-interest, tax and reserve boundaries. Three Salesforce interim observations are indexed-only, not the annual main contrast or a full current account.

Read the full report → · 551-record source-linked ledger · Calculations and local tests

Current investigation ·

The workload can move before the bill falls

Deployed alternatives make useful competition concrete. Contract timing, waiting and whole paid replicas show when a buyer can retain a saving; supplier investment distinguishes the cloud receipt from the next chip sale.

October 16 retrieval boundary; December 2024 and May 2026 deployments retain their dates. The contract horizon and replica fleet are conditional. Private invoices, acceptance and migration costs remain unobserved. The AWS residual is unallocated, not a loss; cloud price reductions do not measure NVIDIA margins.

Read the full report → · 86-record source-linked ledger · Calculations and local tests

Current investigation ·

Who funds the next purchase?

Cash is arriving and the current build was funded. NVIDIA’s customer-account bridge, Meta’s matched quarterly uses and CoreWeave’s financing and working-capital account show what another purchase would need.

October 15 retrieval boundary; financial and contract dates remain separate. The NVIDIA adjustment J is unidentified. Meta’s fixed-use repetition is conditional, and CoreWeave’s facility availability is not all cash. No whole-sector funding total or default forecast is inferred.

Read the full report → · 241-record source-linked ledger · Calculations and checks

Current investigation ·

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

NVIDIA’s scale and duration requirements contrast with Freshworks’ cost-conversion task. Dated claims, paid renewal and future compensation produce plausible paths without establishing either selected security as a bubble.

October 8 retrieval boundary; October 7 regular-session prices; July NVIDIA and June Freshworks inventories. Conditional common-equivalent values and required growth are not targets, probabilities or evidence of market error. Existing awards, future grants and unmeasured claims remain distinct.

Read the full report → · 269-record source-linked ledger · Calculations and checks

Current investigation ·

Cash can reach investors before the business pays for itself

Company funding, seller consideration, a historical put and a repaid bridge show how genuine realization can coexist with remaining investment risk. The completed CoreWeave transactions and Core Scientific refinancing distinguish productive capital recycling from proof of self-financing operations or a fair later entry price.

October 8 retrieval-day boundary. The IPO, tenders, capital snapshot, actual September 2025 rights termination and May/June 2026 financing records keep their dates. Gross receipts are not complete gains. Historical allocation is conditional; the parent-resource reference uses unresolved closing proxies and is not a payout, profit or measured bound.

Read the full report → · 99-record source-linked ledger · Calculations and checks

Current investigation ·

Cheap extraction is not the same as cheap accepted work

Who pays for checking and integration, and who keeps the gain? A documented document-conversion pipeline, managed-service offers and metered coding rights support selective efficiency while a finite cost budget tests what remains for the business.

Retrieval-day boundary: October 6. Historical timing plus newly inspected tariffs is a hybrid calculation, not a rerun. Additional review is not total review; the remaining-cost figure is conditional room for every unpaid cost and any surplus, not an audited margin. Paid volume, collection, reliability and continuing-cost limits remain explicit.

Read the full report → · 55-record ledger · Calculations · Remaining-cost figure

Current investigation ·

Buyer demand and budget substitution

What do buyers keep paying for? NNC’s signed post-pilot purchase, service observations and Klarna’s marketing-production substitution strengthen the case for economically motivated demand without counting every benefit as additional supplier revenue.

Retrieval-day cutoff: October 5. Signed obligation, purchased allowance, service outcomes and unverified payment remain separate. The two 14% benefit tests are conditional alternatives, not payroll savings. Klarna’s attributed reduction and expense reconciliations retain their periods, rounding and unmeasured internal costs.

Read the full report → · 48-record ledger · Calculations · Marketing expense figure

Current investigation ·

Completion and cash before stabilization

The cash has to be usable before the milestone. A matched customer-funded construction account, the reserve’s full and minimum branches, and legitimate pre-acceptance IREN funding show both productive bridging and conditions that can keep sponsor funds committed.

Retrieval-day cutoff: October 4. June’s actual $383.758m construction match differs from the conditional $191.879m pay-first bridge. The reserve determination and Denton phase allocation remain unresolved; no compliance, payment-delay or campus-wide-delay finding is implied.

95-record ledger · Calculations · Three-branch reserve figure

Current investigation ·

Can the cash claims all hold together?

A cash-generating chain can still need another financing. The connected CoreWeave–Core Scientific case reconciles tenant credits, project costs, executed debt and owner distributions, with both feasible and adverse finite refinancing paths.

Retrieval-day cutoff: October 3. April operating illustration, May executed financing and June assumed amortization start are distinct. Additional funding is a required source under assumptions, not an approved facility or expected loss.

198-record ledger · Calculations · Owner-payout and maturity figure

Current investigation ·

Where an investment loss becomes a funding problem

Named GECC holdings, realized cash distributions and the holder’s own creditors make loss absorption and funding constraints inspectable. NVIDIA supplies a different equity/support mechanism. Neither yields a complete lender register or a crisis probability.

Retrieval-day cutoff: 2 October 2026. GECC snapshot: June 30; NVIDIA ownership: January 23; resources: July 26. The one-covenant figure is not complete compliance or cash capacity.

113-record ledger · Calculations · Holder sensitivity drawing

Current investigation · 29 September 2026

Older capacity can earn again—but it does not become free

An older-design contract and reused facilities strengthen selective continuation. A matched historical Austin budget constrains extra costs without claiming a current appraisal, machine-age record or complete second-cycle return.

Retrieval-day cutoff: 29 September 2026. Austin model anchor: 31 March 2025, using amended historical disclosures inspected later.

Source-linked data · Reproducible calculations · Finite-budget drawing

Current investigation · 29 September 2026

Paid persistence is real; the division of the gain is not uniform

Freshworks’ company-wide retention, complete operating expenses and reconciled cash, together with Meta’s provider-reported embedded-monetization experiment, strengthen the favorable commercial case. Neither supplies an AI-only renewal or cash-margin estimate.

Retrieval-day boundary, not an exact finish time. Constant-currency retention, compensation, the signed working-capital reconciliation and experiment limits remain explicit.

194-record input ledger · Calculations · Accrual-profit figure.

Current investigation ·

Strong customers, less effortless funding

Microsoft’s full lease-payment and restricted-resource perimeter; Meta’s seasonally matched and continuing funding tests; Oracle’s customer advances and equity; and a contract-specific limit on spending reversibility.

Balances and events retain their dates. Cash-growth, buffer, financing and policy assumptions stay beside the results; no insolvency or legally executable minimum-spending programme is inferred.

386-record ledger · Calculations · Funding-flexibility figure.

Current investigation ·

What today’s ownership prices require

IREN’s project-to-common-owner bridge, Microsoft’s revenue and reinvestment requirements, frontier funding/dilution and the revised whole answer. September 24 price references and older financial/share dates remain explicit.

Includes the later CoreWeave closing; preserves required residuals versus appraisals and the NVIDIA receipt correction. Model: GPT-6 Pro. 118-record ledger · Calculations · Expectation figure.

Current physical-capacity and recovery investigation

Capacity that earns revenue is not yet capital that earns its cost

Evidence checked: . Retrieval-day boundary, not a publication date or same-day utilization reading.

Accepted capacity, IREN’s equipment and facility budgets, customer advances and extended service terms, finite replacement cycles and conditional continuation-value thresholds. Preserves the CoreWeave repayment case and distinguishes facility-only hosting economics.

Read the full report → · Inspect the 83 inputs and assumptions · Recovery sensitivity figure

Current financing investigation

Who has to pay when AI returns disappoint?

Evidence checked: . Not a publication date. June/July balances and the original pricing record remain dated; a visible later note gives the confirmed closing.

Meta-backed contracted repayment, CoreWeave cash and reserve thresholds, conditional collateral recovery, NVIDIA residual-value support and parent resilience. Explains creditor containment, equity loss and the missing links to wider transmission.

Read the full report → · Inspect source-linked data and assumptions

Current whole-system assessment

Real demand, demanding prices, and growing repayment risk

Whole-system evidence: 18 September 2026; demand 19 September, selected financing 20 September, and physical recovery 22 September. These are inspection boundaries, not publication dates.

Customer value, spending, financing, recovery, valuation and loss transmission in one connected assessment. Conditional models retain their assumptions and the NVIDIA receipt correction.

Read the full report → · Inspect source-linked data

Earlier research — the complete preserved record

History: earlier pages and claim-specific context →

What changed, with the reason attached

Substitution · October 16: actual deployments support feasible alternative service. The new contract and finite-capacity tests explain when that can reduce a bill and who may retain the gain. Historical source measurements, current terms and conditional examples remain separate; the earlier ownership and customer-funding calculations are not rerun. Full interpretation and limits.

Customer-funding evidence · October 15: a selected customer-account bridge retains its missing adjustment, and matched quarters separate operating cash from investment, repayment and completed finance. The current build was funded; repeating the selected uses is a further conditional test. The revised answer.

Ownership economics · October 8: two different profitable layers receive full claim-and-cost tests. The result supports continuing-business value while distinguishing NVIDIA’s scale burden from Freshworks’ cost conversion. It does not identify either selected price as demonstrably a bubble or turn scenario prices into targets. Judgment, countercases and coverage.

Completed transactions and surviving rights: the October 8 investigation distinguishes primary funding from earlier seller receipts, reports the put’s actual September 2025 termination and applies the distinct Issue Date parent-distribution permission. The option, dividend-timing and tender-row conflicts remain open. Earlier report bytes and numerical models are retained. The connected answer and its limits.

Efficiency and full costs · October 6: the new study grounds cheap serving in a documented workload and separates it from accepted-work cost and full-business capture. It retains the printed ACL denominator conflict beside its runtime reconstruction, then applies a separately dated tariff. This is not an author-confirmed correction, a measured provider margin or a replacement for the earlier buyer cases. What changes in the whole answer.

Buyer budgets · October 5: the new evidence is a priced buyer decision, distinct service and cash-saving tests, and the reconciliation of a claimed production saving with broader expense. It strengthens purposeful demand while narrowing the claim on additional spending. Earlier completion, payment, financing and ownership findings remain dated and unchanged. The contribution and its limits.

Holder exposure and funding: an actual portfolio schedule and creditor terms now replace a purely illustrative holder link. Reported distributions support investor realization; frozen-balance loss tests show why capital, borrowing permission and cash differ. No loss event, complete registry or synchronized October NVIDIA position is claimed. What this new evidence adds.

Older service and facility reuse: a seller-reported A100 contract and matched historical Austin accounts now give selective continuation more concrete support. The finite expense-based budget is not a newly discovered asset price; the amended disclosure, operating period and calculation anchor are different dates. What changes in the whole answer.

Persistent payment and retained value: the latest accounts strengthen the favorable commercial case while preserving full-expense, cash and currency boundaries. The working-capital sign is a reconciliation of a dated disclosure, not a change in the reported cash total. Meta’s experiment adds a provider-reported route to monetization; it does not allocate all advertising profit to AI. Inspect the new finding and its limits.

Payer cash and spending flexibility: complete funding accounts and full lease/resource perimeters test the assumption that strong customers imply effortless funding. Meta’s new seasonal/continuation analysis differs from the earlier half-on-half sensitivity; Oracle’s full statement explains advances and equity. The IREN contract finding is a fuller reading of dated terms, not a new termination event. Inspect the distinctions and unresolved mapping

Ownership-price analysis: the new IREN test values remaining project cash against the whole-company claim; Microsoft now separates required revenue growth from cash capture and reinvestment; private expectations include funding and dilution. These are fuller, different models—not observed appraisals or updated versions of every prior financial quantity. What changed in the price and claim boundary.

Full-cost recovery: the earlier general cohort model is now complemented by a disclosed IREN customer contract and separate equipment/facility budgets. The narrower equipment perimeter can clear a hurdle while the combined investment requires continuing value. This is a better-defined calculation, not a newly observed company loss. The cost and cash boundary.

Contractual timing and capacity: the delay analysis retains extended service terms instead of automatically deleting fees; the replacement-cycle test charges new capital and shows a fixed-capacity limit to volume growth. Future prices, utilization and net asset value remain assumptions. Timing protection; continuation thresholds.

Stronger contractual containment: April DBRS analysis identifies Meta’s take-or-pay contract and expected full amortization. This is newly incorporated underwriting evidence about a dated arrangement—not a new September customer contract or verified compliance. Repayment proposition and limits.

Real-obligation cash and recovery tests: the investigation now reconciles actual principal and interest payments and derives collection, reserve and collateral thresholds. Unknown project schedules remain variables; assumed sale values are not observed prices. Definitions and numerical boundaries.

Financing stage advanced: the proposal became priced notes by the September 20 inspection. A later filing confirms $4.2bn issued September 22 and $3.5708bn after discounts and capped calls, before other expenses. This is new closing evidence, not a correction of the earlier cutoff. Chronology and completed terms.

Deeper demand evidence: positive coding field evidence, office time savings and recurring payment strengthen commercial substance. Distinct populations and outcome measures remain separate. The findings do not establish complete cash margins or annual paid renewal. Demand investigation.

Measurement window: the $510.703bn July–June parent cash-investment total contains January–June’s $294.800bn. These are overlapping periods, not amounts to add. Nested-window visual and exact basis.

Later financial period: the June debt snapshot replaces March in the current assessment. A smaller through-2027 share is not pure deleveraging. What the two snapshots do and do not say.

Attributed profitability evidence: the favorable Anthropic adjusted-profit reporting remains visible, with its stock-compensation exclusion and limited access to underlying documents. Interpretation and source.

Explicit conditional models: capital recovery and ownership prices now have inspectable assumptions. These do not turn incomplete market or private-account information into fair-value estimates. Investment returns.

Previously corrected evidence coverage: Cash, commitments and capacity had already corrected the first baseline’s missed Amazon filing. The disclosure was available before the omission; the current assessment does not claim its discovery.

NVIDIA receipt characterization: the stronger assertion about what the prepaid-forward contract proves is withdrawn. Obligation is established; receipt remains unverified, not disproved.

NVIDIA entered into a $1.5bn prepaid-forward contract with Energy Global, LP. The agreement requires payment within three business days after its date. The documents inspected for this investigation do not verify receipt.

Source, exact documentary issue and numerical consequence →

Forecast vintage: the current whole-system account uses the IEA’s April 2026 update; the earlier report retains its older outlook. Both concern global data centers, not measured AI-only cash demand. Dated forecast context.

Connected cash · October 3: the new investigation joins actual funding accounts, an earlier operating illustration and executed debt terms. It tests retained cash versus parent payouts and a finite refinancing, without rewriting the previous chapters or extending GECC’s holdings to a different issuer.

Completion-to-cash investigation · October 4: customer reimbursement, conditional vendor timing, milestone-gated rights and eligible early equipment finance deepen the current answer. The new ledger also corrects the April illustrative coupon’s status without changing earlier debt calculations. No old chapter or companion is replaced. Changes and preserved findings.

Search published pools, pages, reports, and evidence.