NyquistNyquist
PILOT PROGRAM · BY APPLICATIONStart a 60-day pilot →
★ Nyquist Research · No. 15 · Credit Model

What equity implies about the credit of AI-exposed issuers.

Twenty listed companies whose revenue AI can substitute, one structural credit model, and the spread it implies against US high yield. How AIDI-20 is built from public prices and point-in-time balance sheets, where it stood on 28 October 2026, and what it cannot tell you.

Published
27 Oct 2026 · revised 29 Oct 2026
Reading time
8 min
Author
Nyquist Research
Topic
Credit · Structural Models · AI Exposure
Revised 29 October 2026. The first version of this note, published on 27 October, rested on hand-entered market values and debt, a calibrated spread premium and an illustrative history of the index. It has been withdrawn. Every number below is computed from public prices, published balance sheets and FRED series by the code that now recomputes the index each weekday at index.nyquist.pro.
01 — What it measures

A credit reading where there is no credit market.

AIDI-20 reads the credit risk that the equity market implies for 20 listed companies whose revenue is exposed to substitution by AI: customer-experience outsourcing, content and translation, per-seat software, IT services and freelance marketplaces. Few of them have liquid CDS. All of them have a share price and a balance sheet, and a structural model turns the two into a one-year default probability and a spread.

The index is the equal-weighted average of those spreads. The headline number subtracts the ICE BofA US High Yield option-adjusted spread, so a positive value means the basket prices wider than the high-yield market as a whole, and a negative one means narrower.

02 — The basket

Twenty issuers, four clusters, fixed.

TickerIssuerCluster
TEP.PATeleperformance SEBPO & Customer Experience
CNXCConcentrix Corp.BPO & Customer Experience
TTECTTEC HoldingsBPO & Customer Experience
CNDTConduent Inc.BPO & Customer Experience
GGenpact LimitedBPO & Customer Experience
CHGGChegg, Inc.Content, EdTech & Translation
GETYGetty Images HoldingsContent, EdTech & Translation
SSTKShutterstock, Inc.Content, EdTech & Translation
RWS.LRWS Holdings plcContent, EdTech & Translation
PSON.LPearson plcContent, EdTech & Translation
ALITAlight, Inc.Legacy-SaaS (Per-Seat)
GTMZoomInfo TechnologiesLegacy-SaaS (Per-Seat)
RNGRingCentral, Inc.Legacy-SaaS (Per-Seat)
DOCUDocuSign, Inc.Legacy-SaaS (Per-Seat)
DBXDropbox, Inc.Legacy-SaaS (Per-Seat)
DXCDXC TechnologyIT Outsourcing, Freelance & Agencies
KDKyndryl HoldingsIT Outsourcing, Freelance & Agencies
UPWKUpwork Inc.IT Outsourcing, Freelance & Agencies
FVRRFiverr InternationalIT Outsourcing, Freelance & Agencies
WPP.LWPP plcIT Outsourcing, Freelance & Agencies

The list was chosen for exposure and leverage and is not rebalanced. Getty Images is in the basket but not yet priced: its quote history is shorter than the 252-day volatility window, so on 28 October the index averaged 19 issuers.

03 — The model

Merton, solved rather than approximated.

Equity is treated as a call option on the firm’s assets V, struck at a default barrier D, over a horizon of one year. For each issuer and each trading day:

  • Barrier. D = current debt + 0.5 × long-term debt, from the latest balance sheet assumed to be public on that day.
  • Equity volatility. σE is the annualised standard deviation of daily log returns over the last 252 trading days.
  • Assets. V and σV are solved jointly from E = V·N(d1) − D·e−rT·N(d2) and σE·E = N(d1)·σV·V, not taken from the shortcut V ≈ E + D.
  • Distance to default. DD = d2 = [ln(V/D) + (r − σV²/2)·T] / (σV·√T), and PD = N(−DD).
  • Spread. S = −ln(1 − PD·(1 − R)) / T with recovery R = 25% and T = 1, capped at 2,500 bps.
  • Rate. r is the one-year Treasury yield (FRED DGS1) on the day.
  • Index. AIDI = (1/n)·ΣSᵢ over the issuers priced that day, published only when n ≥ 16; the net spread is AIDI − HY OAS.

There is no empirical EDF mapping, premium or multiplier on top. The spread is what the normal model gives, which understates tail risk for issuers far from default and is one reason seven of the nineteen priced issuers contribute almost nothing.

04 — Data, point in time

What was knowable on the day, and what is assumed.

  • Prices are daily closes in the quote currency.
  • Balance sheets are vendor-reported quarterly, half-year and annual statements. The vendor gives no filing dates, so a report is assumed public 45, 60 or 90 days after its period end. Before the first report is assumed public, the earliest one is used and the day is flagged as look-ahead; the index page shades those stretches.
  • Missing debt fields. When current or long-term debt is blank but total debt and lease liabilities are reported, the field is rebuilt as total debt less leases. A field that is not reported at all is taken as zero. Both are flagged per issuer.
  • High yield. FRED BAMLH0A0HYM2 carries only the trailing three years, so the net series starts on 30 October 2023 while the basket series starts on 30 November 2022.
  • Nothing is filled in. If a source fails or fewer than 16 issuers price, the previous files stay as they are and the reason is written to status.json.
05 — Where it stood on 28 Oct

Wider than high yield, and concentrated.

At the 28 October 2026 close the basket spread was 384.5 bps, the highest close in its series and 214.5 bps wider than at the end of 2025. Against a high-yield OAS of 318.0 bps (27 October), the net spread was +66.5 bps. The basket started the series at about 21 bps on 30 November 2022; the net spread started at −420.4 bps on 30 October 2023 and peaked at +93.7 bps on 18 September 2026.

IssuerDDPD, 1ySpread, bpsContribution, bps
Alight (ALIT)0.6027.6%2,315.4121.9
TTEC Holdings (TTEC)0.9317.8%1,429.375.2
Kyndryl (KD)1.1113.3%1,048.255.2
Chegg (CHGG)1.2211.1%867.645.7

Four issuers carry 297.9 bps, or 77% of the level; seven price under 1 bps. By cluster, per-seat software averaged 490.7 bps, BPO and customer experience 478.9, content and translation 321.9, and IT outsourcing and freelance 234.0. Over the last month the level rose 13.3 bps, mostly from the first two clusters.

06 — What it cannot tell you

Model output, with the usual structural-model caveats.

  • It is not a market quote. No CDS or bond prices go in; the spread is what equity prices and reported debt imply under the model.
  • Equal weights concentrate it. A few issuers close to the barrier set most of the level, and one of them crossing the 2,500 bps cap would flatten its contribution.
  • A year of volatility reacts slowly. A crash stays in the window for twelve months. The window was moved from 90 to 252 days on 29 October 2026: with 90 days the index jumped by up to 135 bps on the day a crash left the window; with 252 days the largest daily move in the history is 75.5 bps.
  • Balance sheets are second-hand. They come from a data vendor, the publication lag is assumed, and some debt fields are rebuilt.
  • The net spread is relative. High yield has a different composition, duration and rating mix; the difference is a gauge of how the basket prices against it, not an isolated AI factor.

Not investment advice and not a tradable benchmark.

07 — Where to read it

The index, issuer by issuer.

index.nyquist.pro shows the level, an interactive history of the basket against high yield, the change over standard periods, each issuer’s contribution to the level and to its one-month move, and the full calculation base: DD, PD, asset volatility, leverage, spread, weight, contribution and the balance sheet each number rests on. The same data is published as /data/aidi/latest.json, history.json and constituents.csv, recomputed Monday to Friday at 22:30 UTC.

★   About Nyquist

Model output you can check — line by line.

Nyquist is a quantitative financial operating platform: pricing, factor risk, stress tests and pre-trade decision support on one typed data model. AIDI-20 follows the same rule as the platform: every number on the index page traces to the issuer, the balance sheet and the date it came from.