HYPERSCALER DEBT TTMXBRL PROCEEDS CRWV 2030 SPREAD≈+560bp+85bp/90d GOOGL EQUITY RAISEATM + BRK DC BUILD / GDPRISING CANCELED GW (860M)VS BUILT BIG-5 FCF TTMOCF − CAPEX H100 RENTVAST.AI MEDIAN FLAGSHIP mNAV≈1.6×PREMIUM SOURCEEDGAR · TRACE · BEA · EIA · CBRE
Fundamentals · Financing stress

The bill for the buildout arrives before the crash tape does.

OMEN's indexes and gauge read what markets are pricing. This page reads the plumbing: who is lending, who is quietly switching to equity, how much of the economy the capex has eaten, and how many announced gigawatts actually get energized. Eight theses, each with the metric that would confirm or kill it – five from Paul Kedrosky's ROI teardown, three from Jim Chanos's telecom-collapse analogy. The structural risk underneath them is the same one: deflating token prices funding 10–15-year fixed-payment debt – a classic duration mismatch.

Curated theses · 2026-07-18 · live tape loading…

Frameworks: Paul Kedrosky on Better Offline – "Why AI Has No ROI" (Jun 2026) and Jim Chanos, "The AI Bubble Is Much Worse Than Dot-Com" (2026); the bull-side tape rows follow Gavin Baker on Invest Like The Best – "The AI Selloff Doesn't Match the Data" (Aug 2026), plus the reference series cited under each panel.

0.Live tape · auto-refreshed

The live tape.

The rest of this page is a hand-curated snapshot. This section is not: it refreshes from primary sources on the data cadence, and each row says when its number was printed.

Four feeds the eight theses kept pointing at but never measured: realized hardware demand (TSMC's monthly tape, upstream of every hyperscaler press release), issuance velocity (EDGAR full-text counts of the actual debt and equity paperwork), paid adoption (Ramp's transaction data – the revenue-side answer to "does the capex earn its cost"), and the generator pipeline behind the stranded-GW thesis. Four more carry the bull case in the same auditable form – the counters Gavin Baker put on the record after the July selloff: cash-flow acceleration (hyperscaler OCF growth, XBRL), token demand (OpenRouter's platform-wide weekly tape), the GPU repricing gap (spot vs contract vintages), and the inference-margin floor (flagship token prices vs compute cost at spot rents).

TSMC revenue, YoY

Monthly revenue growth (TWSE filing, ). The buildout is real only while this number is.

US businesses paying for AI

Ramp AI Index () – share of ~70k firms with a paid AI transaction that month.

Debt-issuance events, 90d

FWP + 424B filings by the six AI-capex issuers, trailing 90 days vs the 90 before ().

Big-5 OCF growth, YoY

Operating cash flow across the AI-capex filers, latest complete quarter () – audited XBRL, not guidance. Acceleration pending prior-quarter data.

Routed tokens, weekly

OpenRouter platform-wide (wk of ) – the demand tape behind the capex. Growth loads with the live series.

NVDA trailing P/E

Vs its own 10-year monthly range (percentile loading) – SEC XBRL EPS × Yahoo closes. The "decade-low multiple" claim, audited.

Hardware demand pulse

Upstream of the capex – what the fabs actually billed
SeriesReads onLatestΔ
TSMC monthly revenue ()Realized AI-silicon demand, ~10-day lag
TSMC revenue, YTD vs prior yearTrend, strips the monthly noise
Korea semiconductor exports, 20-dayHBM/memory pulse every ~10 days – the classic cycle canarymanual
METRIC – monthly and YTD revenue growth at the fab that makes every AI accelerator; a Korea 20-day print that rolls over flags memory demand cracking before any earnings call.  SOURCETWSE OpenAPI (live) · Korea Customs 20-day release (hand-updated; no keyless API)

Issuance velocity

EDGAR full-text search counts · trailing 90 days vs the 90 before
PaperworkReads on90dPrior 90d
FWP + 424B, six AI-capex issuersThesis i in filing counts – the debt wave, mechanically
S-1s mentioning "artificial intelligence"IPO pipeline – Kedrosky's "gonging of the bell"
Form Ds mentioning "artificial intelligence"Private raises – the pre-IPO froth layer
METRIC – acceleration, not level: each count vs its own prior window. Issuers: MSFT · GOOGL · AMZN · META · ORCL · CRWV.  SOURCESEC EDGAR full-text search API (live) · thesis i · thesis iii

Paid adoption – the revenue side

Does the capex earn its cost? The only live demand-side series on this page
SeriesReads onLatestΔ
Ramp AI Index ()Share of US businesses with paid AI subscriptions – transaction data, not surveys
Census BTOS estimate (contrast)The survey number Ramp's transaction data corrects
Anthropic Economic Index, latest releaseUsage composition (augmentation 57/43 automation, hand-updated) – freshness tracked live
Agent-stack installs, npm weekly ()Claude Code + Codex + Gemini CLI downloads – the agentic S-curve, read in package managers
Agent SDK installs, PyPI weeklyclaude-agent-sdk + openai-agents – a hard undercount (OpenAI-compatible traffic rides the openai package)
METRIC – adoption level and momentum; a stall here while capex accelerates is the ROI thesis confirming in the one dataset that can't be talked up. The agent-install rows are the closest public proxy for Baker's 500k→500M agentic-user trajectory.  SOURCERamp AI Index CSV (live) · Anthropic Economic Index (freshness live, split manual) · npm downloads API + pypistats (live via updater)

Cash-flow acceleration

The bull case's load-bearing number · hyperscaler OCF growth, audited XBRL
SeriesReads onLatestPrior
Big-5 OCF, YoY ()Baker's master loop – contracts roll to spot, cash flow accelerates, ≈$700B of credit demand disappears
Acceleration, quarter over quarterThe second derivative is the claim: decelerating growth breaks the loop before levels do
Newest quarter, cohort statusA 4-of-5 cohort never gets quoted against a 5-of-5 baseline – it would read as a collapse that never happened
METRIC – combined operating-cash-flow growth YoY across MSFT · GOOGL · AMZN · META · ORCL, and its quarter-over-quarter change; Baker's falsifier is this number failing to persist through contract repricing.  SOURCESEC XBRL companyconcept (live, same feed as the capex/OCF panel) · framing per Gavin Baker, ILTB (Aug 2026)

Token demand

The demand tape behind the capex · OpenRouter platform-wide weekly tokens
SeriesReads onLatestΔ
Routed tokens, weekly (wk of )Aggregate inference demand in the price-sensitive routed segment – the mix-shift-vs-demand-shift test
4-week growthMomentum – Baker's "token growth is accelerating" claim, checkable weekly
~1-year growthThe scale of the demand curve the buildout is racing
Chinese-lab share of the weekOpen-weight substitution inside the same demand curve – margin dollars migrating, not demand leaving
METRIC – platform-wide routed tokens per week and its growth; first-party OpenAI/Anthropic/Gemini API traffic is invisible here, so this is the substitutable segment's floor, not the market total.  SOURCEOpenRouter market-share series (live, in-browser) · China Watch (share snapshot fallback)

GPU repricing gap

Spot vs contract vintages · the mechanism behind the ≈$700B credit-demand claim
GenerationReads onSpotContract vintageGap
H100 ()Legacy fleets – 2023-vintage LTAs sit far above today's spot; the renters' churn risk, the owners' cushion≈$4.40/hr 2023 LTA
B200 curatedCurrent generation – Dec-2025-vintage contracts sit under spot; the installed base is underearning, repricing upside on roll≈$3.90/hr Aug 2026≈$2.50/hr Dec 2025
METRIC – spot $/GPU-hr vs the contracted rate of each generation's dominant LTA vintage; a sustained spot contraction is Baker's own falsifier #1, and the B200 gap closing from above is the repricing loop completing.  SOURCEvast.ai H100 asks (live) · B200 spot and both contract anchors curated from CRWV filings, deal press and ILTB (Aug 2026) ("mid-$2 seven months ago, just under $4 now")

Inference-margin floor

Flagship token prices vs compute cost at spot rents · where the margin dollars migrate
LineReads onOutput $/MImplied margin
Serving-cost floor at spotH100 spot rent ÷ throughput, at 0.5–2.0M output tok/GPU-hr – the denominator every model shares
US flagship (Claude Opus)Frontier pricing power – Baker puts frontier token margins at 80–95%
CN flagship (DeepSeek pro)Open-weight serving economics – Baker puts these near 30%; compression biting shows up here first
METRIC – implied gross margin = 1 − serving-cost floor ÷ price, with margins quoted at the 1.0M tok/GPU-hr midpoint; a bounding exercise, not a P&L – batch efficiency, MoE routing and first-party hardware all move the true number.  SOURCEOpenRouter models API (live, newest priced model per line) · vast.ai (live, via market-data.json)

Generator pipeline

The stranded-GW thesis in federal data · EIA-860M inventory by status
StatusReads onNameplate GW
OperatingWhat the grid actually has
Under constructionSteel in the ground
Planned / approvals pendingThe press-release layer
Canceled or postponedThesis v in federal data – capacity that stopped
US interconnection queue (LBNL, 2023-12)Requests, most of which never build≈2,600
PJM capacity clears (2025/26 → 2026/27)Scarcity pricing the AI load already caused$269.92 → $329.17/MW-day
METRIC – planned vs under-construction vs operating GW; a widening planned-minus-built gap is the stranded-asset pipeline forming, and the canceled line is that pipeline already realised.  SOURCEEIA-860M monthly workbook (loading…) · LBNL queues + PJM auctions (hand-updated) · thesis v
i.Thesis one

Hyperscaler debt saturation.

"As of Q1 2026 the hyperscalers are the largest issuers of investment-grade debt worldwide. They just passed the banks."

The buildout has migrated from cash flow to credit. When prime corporates saturate the IG market, the marginal buyer changes – recent books lean on European insurance funds and Middle East sovereign wealth, the buyers who historically show up at the end of a cycle. Watch issuance volume, book quality, and new-issue concessions.

Hyperscaler debt proceeds, TTM

Cash proceeds from long-term debt issuance across the AI-capex filers, trailing four reported quarters – audited XBRL, not a press tally.

≈$196B
Big-6 US banks, same window curated

The comparison group hyperscalers overtook in Q1 2026. Stays hand-curated on purpose: JPM and WFC tag no debt-issuance concept the XBRL API exposes, so a computed bank total would omit two of six and flatter this very comparison.

$125B
Order book on META's $30B deal

Oct 2025, the largest corporate bond deal of the year – demand at the top is not the question; capacity is.

The issuance wave

Benchmark deals, Sep 2025 → Jul 2026 · curated from prospectus + press
DateIssuerVehicleSize
Sep 2025OracleIG bonds, 5–40y$18.0B
Oct 2025MetaIG bonds – largest corporate deal of 2025$30.0B
Oct 2025Meta / Blue OwlHyperion SPV private placement (off balance sheet)$27.3B
Nov 2025AlphabetIG bonds, USD + EUR tranches$25.0B + €6.75B
Dec 2025AmazonIG bonds – first issuance since 2022$15.0B
H1 2026Complex-wideFollow-on IG + SPV/private-credit vehicles≈$100B+
METRIC – TTM IG issuance by hyperscalers vs banks; new-issue concession; share of book taken by insurers and sovereign wealth.  SOURCESEC XBRL companyconcept (live: debt proceeds) · EDGAR 424B/FWP filings · FINRA TRACE · SIFMA issuance statistics · live filing counts in the tape
ii.Thesis two

AI-specific credit stress.

Generic high-yield can stay calm while the AI paper quietly reprices. Watch the single names, not the index.

The gauge's credit family reads HYG drawdown and the HY/IG ratio – the whole market. This section watches the instruments that fund the buildout itself: neocloud high-yield, Oracle's curve, and the SPV paper. If the duration-mismatch thesis is right, this is where stress prints first, months before it reaches the index.

≈+560bp
CoreWeave 2030 spread

The purest listed neocloud credit. Widened ≈85bp over 90 days while generic HY was flat – early divergence.

≈+130bp
Oracle 10y spread

Roughly double its pre-buildout level. Oracle carries the most leveraged AI capex program of the majors.

≈$100B+
AI private credit outstanding

SPVs, vendor financing, and data-center lending sitting outside public marks – repricing arrives late and all at once.

The AI credit tape

Single-name and structure watchlist · hand-curated · levels approximate, from TRACE prints + filings
InstrumentReads onLevelΔ 90d
CoreWeave 9¼% 2030Neocloud funding cost – GPU-backed, contract-concentrated≈+560bp+85bp
Oracle 10y benchmarkMost-leveraged major – capex running ahead of operating cash flow≈+130bp+22bp
Meta Hyperion SPV 2049 (Blue Owl)Off-balance-sheet data-center paper – the new marginal structure≈+240bp+15bp
ICE BofA HY OAS (contrast)Generic high yield – what the gauge already reads
CRWV 2030 vs own 12m range72%
ORCL 10y vs own 12m range61%
HYPERION SPV vs issue spread34%
GENERIC HY vs own 12m range18%
METRIC – spread level and 90-day change per instrument, each vs its own 12-month range; divergence of AI names from generic HY.  SOURCEFINRA TRACE · SEC EDGAR 8-K · FRED HY OAS
iii.Thesis three

Equity raises signal the debt is running out.

"What's in it for you as a provider of equity here? It gives you no call on future cash flows. That they're doing this is surprising – unless debt capacity is the constraint."

A mega-cap selling stock at the market is a tell, not a flex. Equity is the most expensive money a prime credit can raise – you only reach for it when cheaper channels (operating cash flow, IG bonds, SPVs) are saturated. Alphabet's $80B program is the first mega-cap ATM era in the modern market. Track who follows.

Alphabet equity proceeds, reported

Cash actually raised, per Alphabet's XBRL cash-flow statement – against an $80B announced program ($10B Berkshire placement plus two ATM sale programs, Jul 2026). Alphabet had never tagged this concept before 2026; the tag appearing at all is the tell.

Every other filer combined, TTM

MSFT · ORCL · CRWV, trailing twelve months – and this is ordinary employee stock-plan flow, not secondary issuance. That is the comparison: one mega-cap's raise against the entire rest of the complex's routine share sales.

SBC→BB
The buyback treadmill

Stock comp forces buybacks, buybacks eat cash flow, capex eats the rest – then the SPVs and the ATM appear.

Equity events watchlist

Secondary issuance, placements, and ATM programs · trailing 12 months
DateIssuerEventSize
Jul 2026AlphabetBerkshire private placement$10B
Jul 2026AlphabetTwo at-the-market sale programs≈$70B
WatchMeta · OracleNext-most-stretched capex/OCF ratios – candidates to follow
WatchOpenAI · AnthropicIPO S-1s – check whether training costs get capitalized ("earnings before bad stuff")
"The blow-off top is this year's three mega IPOs. That marks the gonging of the bell."
Paul Kedrosky · Better Offline, Jun 2026 · the IPO-timing markets already sit inside OMEN's Bull index
METRIC – trailing-12-month secondary equity issuance by hyperscalers; count of active ATM programs; S-1 accounting treatment of training costs.  SOURCESEC XBRL companyconcept (live: equity proceeds) · EDGAR 8-K / 424B5 / S-1 · OMEN Bull index (IPO markets)
iv.Thesis four

AI capex is eating the economy.

Railroads peaked near 6% of GDP and gave us 1873 and 1893. Fiber hit ≈1.2% in 2000. AI data centers are at ≈1.8% and still accelerating.

When one investment category carries a visible share of GDP growth, the macro cycle and the capex cycle become the same cycle – a slowdown in data-center spend reads as a recession print. The monitor already tracks audited capex vs operating cash flow from SEC XBRL; this section adds the macro layer on top.

Big-5 capex, TTM

MSFT + GOOGL + AMZN + META + ORCL, each filer's own trailing four quarters per XBRL filings – roughly triple the 2023 run rate.

Of US GDP

Data-center construction (, Census C30 SAAR) plus computer-equipment investment (, BEA) over nominal GDP. Narrower than the monitor's big-5-capex/GDP ratio, which counts five companies' worldwide capex.

Of GDP growth, latest quarter

Big-5 capex growth as a share of the change in nominal GDP – strip it out and the economy is materially slower.

Capex ÷ operating cash flow, by filer

Audited XBRL, trailing 12 months · above 1.0 means the buildout outruns the business
loading…
BuildoutPeak share of GDPWhat followed
Railroads, 1880s≈6%Panics of 1873/1893 · ≈half of boom-era track eventually abandoned
Telecom fiber, 2000≈1.2%2001–02 bust · fiber found reuse, but only after the equity was destroyed
AI data centers, 2026Open – this page exists to watch it
METRIC – big-5 capex TTM and capex/OCF per filer; AI investment contribution to real GDP growth, quarterly.  SOURCESEC XBRL (live, per filer) · Census C30 data-center line + BEA/FRED (live)
v.Thesis five

>1 GW projects don't get finished.

"Anything that's targeted over a gigawatt doesn't get finished. The vast majority don't get fully powered."

Announced gigawatts are a press release; energized gigawatts are a utility interconnection. The gap between them is the stranded-asset pipeline: speculative shells, behind-the-meter gas plants with 30–40-year lives, and county budgets pre-spending tax revenue that never arrives. The metric is brutal and simple – announced vs under construction vs energized.

≈48GW
Announced US AI pipeline

Dedicated AI data-center capacity announced to date – roughly the load of 35 million homes.

≈12GW
Actually under construction

Steel in the ground per utility filings and interconnection queues – a quarter of the announcements.

0
>1 GW campuses fully energized

Of 14 announced gigawatt-class projects, none is fully powered. Abilene is the closest, at partial load.

Gigawatt-class project tracker

Announced ≥1 GW campuses · status per company statements + utility filings
ProjectSponsorTargetStatus
Stargate Abilene, TXOpenAI · Oracle · Crusoe1.2 GWPartially energized
Hyperion, Richland Parish, LAMeta5 GWUnder construction
Prometheus, New Albany, OHMeta1 GWUnder construction
Colossus 2, Memphis, TNxAI1.5 GWUnder construction
Fairwater, Mt Pleasant, WIMicrosoft≈1 GWUnder construction
Wonder Valley, AB (Canada)O'Leary / Greenview7.5 GWAnnounced
Utah / New Mexico mega-sitesVarious1–10 GWSpeculative
Orbital constellation (LEO)SpaceX / Starlink1–2 GW (8 GW rumored)Speculative
Starcloud orbital DCStarcloud (NVIDIA-backed)5 GW targetAnnounced
"Counties pre-budget the tax revenue – new playgrounds, water, schools. When the data center doesn't get finished, they're worse off than before it was announced."
The second-order casualty · regional fiscal exposure · watch county bond issuance against unfinished sites
METRIC – announced vs under-construction vs energized GW; cancellations and pauses; gas-turbine order backlog (sold out through ≈2029); state moratorium odds (already a Bear-index constituent). Orbital rows: company statements + FCC filings – Baker (ILTB, Aug 2026) puts orbital monetization at ≈$50B/GW and calls the 8 GW figure "near-implausible" on launch cadence; orbital is also the regulatory-arbitrage hedge against the moratorium risk the Bear index prices.  SOURCEEIA 860M · LBNL interconnection queues · Kalshi · OMEN Bear index (moratorium market) · live generator pipeline in the tape
vi.Thesis six · Chanos

The dark-fiber problem, rebuilt in silicon.

"It has identical mechanics to the 1999–2000 telecom build-out." The fiber got laid; ~95% of it sat dark for a decade; the equity was gone long before the demand arrived.

A buildout can be real and ruinous at once – the internet did need the fiber. What killed the telecom names was timing: capacity energized years before demand could absorb it, funded by debt that came due first. The AI tell is the same – falling unit economics on already-built compute. When rental prices for last-generation GPUs fall faster than the boxes depreciate, the market is telling you supply has outrun paying demand.

≈95%
Fiber that sat dark, 2002

Of the strand-miles laid into the telecom bubble, most was never lit – ≈$5T of market value erased before it found reuse.

H100 on-demand rental, median ask

Live vast.ai median (); p10 . Down from ≈$8/hr in 2023 – depreciation on a ~2-year asset the hyperscalers still carry over 5–6. The spot tape is noisy day to day; the multi-year direction is the claim.

≈4×
Announced vs absorbable load

≈48 GW announced against a demand curve that plausibly absorbs a fraction on the stated timeline – the overbuild ratio the fiber map warned about.

Same movie, new medium

Telecom 1998–2002 vs AI 2024–2026 · mechanics, not levels
MechanicTelecom, 2000AI, 2026
Asset built ahead of demandLong-haul fiberGPU clusters + power
Vendor financingLucent / Nortel to CLECsNvidia to neoclouds + labs
Depreciation vs reality20–25y book life on gear obsoleted in ~55–6y book life on GPUs obsoleted in ~2–3
Utilization tell% of fiber litGPU rental price · fleet utilization
Outcome≈$5T erased; capacity reused post-bankruptcyOpen
"This has identical mechanics to the telecom build-out that erased five trillion dollars of value."
Jim Chanos · "The AI Bubble Is Much Worse Than Dot-Com," 2026 · the overbuild is the thesis, the utilization tape is the metric
METRIC – secondary GPU rental/spot price vs book depreciation curve; fleet utilization; data-center vacancy; announced vs absorbable GW.  SOURCEvast.ai H100 asks (live) · CBRE data-center figures · LBNL queues · OMEN stranded-GW tracker
vii.Thesis seven · Chanos

Free cash flow is quietly going to zero.

Chanos's depreciation point in one number: earnings hold near records while the cash the business actually throws off has collapsed. The gap is the buildout, capitalised and stretched.

Thesis four watched capex ÷ operating cash flow climb toward 1.0. Push one line further down the statement and you get the punchline: operating cash flow minus capex – true free cash flow – has fallen ~80% across the Big-5, and turned negative at Oracle and Amazon. GAAP earnings don't show it for exactly Chanos's reason: a ~2-year chip depreciated over 5–6 years keeps net income high while the cash walks out the door now.

Big-5 free cash flow, TTM

OCF minus capex across MSFT · GOOGL · AMZN · META · ORCL – down from ≈$300B+ before the buildout, on roughly flat-to-higher net income.

Now FCF-negative

The filers whose capex outruns what their operations generate in cash. Audited XBRL, each filer on its own trailing four quarters.

≈$400B
Net-income-to-cash wedge

The gap between reported Big-5 net income and free cash flow – the non-cash cushion that lengthened depreciation keeps inflating.

Free cash flow, by filer

OCF − capex, trailing 12 months · audited XBRL · negative = the buildout outruns the business in cash, not just accruals
loading…
"A chip that's obsolete in two years is being depreciated over five, six, seven. It's a huge boost to reported earnings during the build-out."
Jim Chanos · 2026 · the wider the useful-life assumption, the wider the net-income-to-FCF wedge above
METRIC – OCF − capex per filer and aggregate, TTM; net-income-to-FCF conversion; disclosed server/GPU useful-life assumption.  SOURCESEC XBRL 10-Q cash flows (live, per filer) · XBRL PP&E useful-life notes · OMEN capex/OCF panel
viii.Thesis eight · Chanos

The reflexive treasury vehicle.

"Absurd." A company issues stock at a premium to the assets it holds, uses the cash to buy more of that same asset, and books the rising price as validation. The loop only runs while the premium holds.

This is the froth tell that sits next to the buildout, not inside it. A treasury vehicle trading at 1.6× the value of what it owns is priced as a perpetual-motion machine: premium funds purchases, purchases lift the asset, the higher asset "justifies" the premium. Chanos's point is that the mechanism is reflexive – it runs in reverse just as fast. The pattern started in crypto; watch it migrate to AI-compute and token treasuries.

≈1.6×
Flagship mNAV premium

Market cap ÷ net asset value of the largest crypto-treasury vehicle – you pay ≈$1.60 for $1.00 of the underlying it holds.

≈70+
Treasury-strategy vehicles

Public companies whose primary "operation" is holding a financial asset bought with issued equity – a category that barely existed three years ago.

ATM→buy
The reflexive loop

At-the-market issuance funds asset purchases; purchases lift the mark; the mark justifies the next raise – until the premium inverts.

Treasury-vehicle watchlist

Premium/discount to holdings · curated · a discount (mNAV <1) is where the loop breaks
Vehicle typeUnderlyingmNAVState
Flagship BTC treasuryBitcoin≈1.6×Premium
Second-wave BTC treasuriesBitcoin≈1.0–1.2×Compressing
ETH / SOL treasuriesEther · Solana≈0.9–1.1×At/near NAV
Emerging AI-compute treasuriesGPUs · token creditsWatch
"These bitcoin treasury companies are absurd."
Jim Chanos · 2026 · the reflexive vehicle is the late-cycle signature, whatever the underlying asset
METRIC – market cap ÷ NAV (mNAV) premium per vehicle; ATM issuance volume; count of new treasury vehicles; first AI-compute treasury at a premium.  SOURCESEC EDGAR 8-K / 424B5 · company NAV disclosures · mNAV trackers

Honesty.

This page is part live, part curated, and every figure says which it is. Live from primary sources on the data cadence: the whole of the live tape (TWSE, EDGAR full-text, Ramp, Hugging Face, EIA-860M), plus every capex, cash-flow and issuance number in theses i, iii, iv and vii (audited SEC XBRL), the capex/GDP ratio (Census C30 + BEA), the GPU rental tape (vast.ai) and the HY OAS contrast (FRED). Still curated by hand and marked as such: the single-name credit spreads in thesis ii (no free per-CUSIP source exists), the bank issuance comparator, the gigawatt project tracker, and the treasury-vehicle watchlist. Levels marked ≈ are approximations assembled from filings, TRACE prints, and press, and can be stale or wrong; treat them as a reading list with numbers, not data. Unlike the indexes and the gauge, these are fundamentals, not market prices – they can't tell you when, only how much has been borrowed against the answer. The eight theses come from two aligned bearish worldviews – Kedrosky's ROI teardown and Chanos's telecom analogy; the bull rebuttal – that inference demand outruns the depreciation schedule – is exactly what the Bull index prices on the other side of the pair.