Where the money comes from: —
Use the tabs below to switch between ten pages. Charts by tab: Overview has the spread ladder, option skew and daily trends; "Spreads & CDS" has the new-issue spread curves; "Option smile" has per-name smiles and implied densities.
Each line is a change versus the previous trading day, not a level: short-term compute price, who pays for 2030 compute, oversupply odds before 2035, financing conditions, regulatory structure, plus upcoming events.
| Lens | Question | Today's view | Marginal change | Evidence |
|---|
| Name | Strategy | Today | Today's trades | Logic |
|---|
Overall tilt = sum of position dollar deltas ÷ sum of their absolute values: −1 is all short, +1 all long; deltas use Cboe option prices and Black–Scholes. Personal record, not investment advice.
Count of indicators in the Spreads & CDS table by threshold status. Red = red line breached, yellow = yellow line breached, improving = crossed a threshold in the good direction.
Each issuer's 5y CDS (or cash spread), bp on a log axis, one line per issuer; solid lines and filled points are CDS quotes, red rings are jumps flagged by the model below. Dashed lines with hollow points are 5y CDS implied from secondary-market bond prices: they start from the latest real CDS print and only plot days when the bond price changed (method in the table below). Line ends show the latest value.
| Issuer | Note | Date | Price | YTM % | Treasury % | G-spread bp | Basis bp (anchor · CDS) | Implied CDS bp | vs prior update | Price |
|---|
One ~5-year senior USD note per issuer. Price is the Business Insider daily close (latest price for today); yield to maturity is solved from the clean price plus accrued interest, 30/360, semi-annual; the matching Treasury yield is linearly interpolated from FRED constant-maturity yields (CNBC closes for the latest session FRED has not published, CNBC intraday for today). G-spread = yield − matching Treasury; basis = real CDS on the anchor date − that day's G-spread, absorbing the swap spread, the CDS-bond basis and the tenor gap; implied CDS = G-spread + basis, assuming the basis has not changed since the anchor date. A price equal to the prior session's is marked 'not updated' and is not plotted. Nvidia has no dated real CDS print, so only its G-spread is recorded.
For each issuer, the log change between consecutive observations divided by a robust scale (1.4826 × MAD of that issuer's changes; pooled across issuers while a series has fewer than 5 changes; floor 2%). A jump is |z| ≥ 3 and |change| ≥ 8%. Observations are not equally spaced, so z is per observation, not per day, and thresholds are lenient while the sample is small.
| Date | Issuer | From → to (bp) | Δbp | Δ% | z | Verdict |
|---|
One row per CDS or spread (bp, log scale). Hollow dot = table baseline, solid dot = latest print; the longer the link, the bigger the move. Short ticks mark the yellow and red thresholds. CoreWeave above 1,000 gets its own scale.
Implied vol of the 12/18-expiry 10% OTM put (left bar) and call (right bar) minus ATM vol, in vol points. Below 0 (green) = cheaper than the lognormal priced at ATM vol; above 0 (red) = richer. Per-name smiles and implied densities are on the "Option smile vs lognormal" tab.
Latest snapshot price and change versus prior close; the static page does not pull live quotes. Oracle new-share issue price $116, Meta forward sale price $625, METAM issue price $50.
CDS is the latest available print that day (bp); the hike probability is CME FedWatch for Dec 9.
| Date | ORCL CDS | AVGO CDS | META CDS | CRWV CDS | NBIS price | IG index | HY OAS | Dec hike % | H100 index | ORCL price | Status | Events |
|---|
X-axis is tenor (years), y-axis is spread over Treasuries (bp). Oracle's five tranches came 145–175bp wider than the same tenors in February; Meta is the 11/3 USD initial price talk, and final pricing usually tightens 15–30bp from talk.
The Today column is the latest available print; "carried over" means no new quote today. Thresholds are the monitor's settings. Sources are in the Sources panel and the last column.
| Group | Indicator | Today | Prior / baseline | Change | Status | Threshold | Source |
|---|
For the December hike probability the yellow line is a one-day move of more than 15 points; red >90%; <60% improving; <40% green.
| Metric | Today | Baseline | Status | Driver / note | Source |
|---|
In-flight and new events from the last two weeks, newest first. The status column is the event's read-through for "where the money comes from".
| Date | Category | Event | Impact | Status | Source |
|---|
Annual value per MW is in $ millions per MW per year; new contracts of ≥3 years and ≥$5bn below $16M are yellow, below $13M (Oracle's locked price) red.
| Kind | Item | Buyer → seller | Total | Term | MW | $M/MW-yr | When | Status | Note |
|---|
Ceasefire = the arms race cooling (bullish for spenders, bearish for upstream suppliers); Winner = a capability gap opening (bullish for the winner and its supply chain, bearish for losers and the neoclouds that depend on them). Windows: Oracle mid-December, Microsoft and Meta late January, Alphabet and Amazon early February.
| Group | Signal | Triggered | Today's evidence | Direction | Strength |
|---|
The "temporary turbine" loophole: xAI powers up fast with portable gas turbines treated as permit-exempt mobile sources, its quickest route to expansion; grid-connected rivals (Microsoft, Meta, Oracle and others) are little affected. If federal courts or the EPA ultimately close the loophole (stationary-source permits, emissions offsets written into permits), xAI's self-generation build-out slows. The risk has little to do with the election, but a Democratic House can apply pressure, and a change of administration in 2029 could flip the EPA's stance.
Thresholds: a court ruling that the turbines need permits with a shutdown order or fines, a final EPA rule requiring permits or emissions offsets, or legislation → red; a case proceeding on the merits, EPA rule delays, states or counties withdrawing temporary-turbine exemptions, House hearings → yellow; the EPA nonroad-turbine exemption taking effect or a court upholding the mobile-source exemption → improving (for the self-generation route).
| Date | Topic | Item | Status | Direction | Next trigger | Next date | Source |
|---|
Price is split into forward EPS × forward P/E and each side is watched: EPS for the earnings gain, P/E for the discount rate (10-year real yield DFII10 and the equity risk premium).
Δln(price) = Δln(NTM EPS) + Δln(forward P/E). Blue = EPS contribution, orange = P/E contribution, tick = the basket's actual return; the gap between the sum and the return is the residual (coverage gap or ETF tracking), red beyond ±0.5%. The start-point EPS is Yahoo's 90-days-ago consensus (about 63 trading days back). The long-duration basket is ETFs with no EPS, so only its return is shown.
Left axis DFII10 (%); right axis ERP = forward earnings yield − DFII10 (%). ERP points come from weekly snapshots and Yahoo's 7/30/60/90-days-ago consensus; nothing is interpolated between points.
Per stock, the sign of (up − down revisions) to the +1y consensus over 30 days; breadth = (net-up names − net-down names) / covered names. Blue = equal-weight, purple = cap-weighted.
A positive, widening gap means AI-driven upgrades are concentrated in the suppliers and have not reached the users. History adds one point a week from 2026-11-04.
A 63-day DFII10 change within ±15bp counts as flat. Real rate up with EPS contribution ≥ |P/E drag| is growth-led, otherwise funding-competition-led; real rate down with EPS up is a tailwind, with EPS down recessionary. One cell per week.
Forward P/E = Σ market cap / Σ(shares × NTM EPS); NTM EPS blends 0y and +1y by the share of the fiscal year left. Implied g = (DFII10 + ERP assumption) − earnings yield, with an ERP assumption of 4%: an assumption, not an estimate. Rate β is the 126-day regression of daily basket returns on daily DFII10 changes, shown as the return per +10bp.
| Basket | Coverage | Fwd P/E | Earnings yield | ERP | Implied g (assumed) | Rate β / +10bp | Breadth eq | Breadth cap |
|---|
Quarterly. Capex growth needs five quarters and TTM free cash flow needs four; anything short shows N/A. The source column says whether SEC EDGAR or the Yahoo fallback was used.
| Company | Quarter end | Revenue $bn | Capex $bn | Revenue YoY | Capex YoY | Capex / OCF | FCF TTM $bn | Source |
|---|
| Series | Latest | Observed | Change | Source |
|---|
The top half turns forward P/E into an earnings yield and compares it with the 10-year Treasury: how many points equities pay over Treasuries. Then it asks what P/E would be if equity only demanded "Treasury + these companies' debt-weighted CDS". The bottom half watches leverage on and off exchange: margin debt, the Chicago Fed leverage subindex, the VIX term structure, private-credit BDC discounts, corporate leverage, and marker events of off-exchange deleveraging. Anything missing shows N/A.
Black dot = current forward P/E; the three marks = fair P/E at perpetual nominal growth g of 0%, 2% and 4% (capped at 60). A dot left of g=0 prices in no growth at all, cheap even on this lenient hurdle; a dot right of g=4% needs perpetual growth above 4% or a lower risk premium to justify it. Loss-making companies have no P/E and are not drawn.
Earnings yield = NTM EPS ÷ price; EPS from the equity module's weekly snapshot ({eps}), prices quoted at run time. Premium over Treasuries = earnings yield − 10-year Treasury (nominal, {ust}); premium over real = earnings yield − DFII10. Fair P/E = 1 ÷ (Treasury + spread − g). Equity is paid after creditors, so its premium should exceed CDS: the "fair P/E" is a lenient ceiling, not a target price. Implied g = Treasury + spread − earnings yield: the perpetual growth today's price needs on that lenient hurdle. Basket spreads are weighted by Yahoo total debt; AI users and SPY have no per-name CDS and use the IG index OAS as a proxy.
| Basket | Coverage | Fwd P/E | Earnings yield | Prem. over UST | Prem. over real | Wtd spread bp | Debt with CDS | Fair P/E g=0 | g=2% | g=4% | Implied g |
|---|
Spread "CDS" = the 5y CDS tracked in the monitor (may be carried over, see the Spreads & CDS page); "proxy" = no tracked CDS, an ICE BofA index OAS (IG / BB / HY) stands in, not the company's own spread. Debt is Yahoo latest-quarter total debt (TSMC converted from TWD).
| Company | Fwd P/E | Earnings yield | Prem. over UST | Debt $bn | Spread bp | Spread source | Fair P/E g=0 | g=2% | Implied g |
|---|
On exchange: FINRA margin debt (monthly, about a month behind). Financial system: Chicago Fed NFCI and subindexes (weekly; >0 means tighter or more levered than the long-run average). Forced-deleveraging regime: VIX ÷ VIX3M (>1 inverted). Off-exchange private credit: listed BDC price ÷ latest reported NAV (Yahoo).
| Indicator | Latest | Observed | Change | Status | Threshold | Source |
|---|
Debit balances in customer margin accounts, last 10 years. Dashed = YoY (right axis); the +30% line marks leverage build-up.
Yahoo latest-quarter total debt and cash, TTM EBITDA. Net debt/EBITDA > 3 yellow, > 5 or EBITDA ≤ 0 red. Excludes off-balance-sheet lease commitments and SPVs (true leverage at Oracle, Meta and others is higher; see the SPV / Financing page).
| Company | Debt $bn | Cash $bn | EBITDA $bn | Debt/EBITDA | Net debt/EBITDA | Status |
|---|
The latest marker and date for each event type. "yes" = happened recently; "watch" = near a threshold; "no" = none found. Numeric signals also appear in the leverage table above.
| Signal | Happened | Status | Detail | Date | Source |
|---|
For each name, take the mid-price implied vol of OTM options expiring 2026-12-18 (45 days) and fit a quadratic smile in ln(K/S); the lognormal at ATM vol (i.e. flat Black–Scholes vol) is the benchmark. Bars = implied vol of the 10% OTM option minus ATM vol (vol points): below 0 means that OTM option is cheaper than lognormal pricing (green), above 0 richer (red). Left bar is the 10% OTM put (K = 0.9S), right bar the 10% OTM call (K = 1.1S). Data are Cboe delayed-quote snapshots; implied vols are solved from the bid/ask mid (r = 4.0%, no dividends).
Dots = mid-price implied vol at each strike; solid line = quadratic fit; dashed = ATM vol (lognormal benchmark). Green dots below the ATM line imply thinner tails than lognormal.
The market-implied density comes from the fitted smile via the Breeden–Litzenberger second difference in strike; the lognormal uses ATM vol. X-axis is terminal price / spot.
Verdicts are on a price basis: IV more than 1 point below ATM = "cheaper than lognormal", more than 1 point above = "richer". Tail probabilities are on a probability basis: for puts, the probability the terminal price ends below K; for calls, above K. The market-implied value comes from the fitted smile via the digital-option identity (with the skew correction); a ratio < 1 means the market assigns less probability beyond that strike than the lognormal does. A smile makes the density more peaked near ATM, thinner in the shoulders and fatter far out, so the two bases can disagree.
| Type | Strike | K/S | IV % | IV − ATM (pts) | Mid $ | Delta | Market tail % | Lognormal % | Ratio | Verdict |
|---|