SpotWire composes four flagship indices from transaction-derived data
and public infrastructure signals. The principles below govern all four.
Last data print:
· 53 sources
· 3,237 weekly comps
Principles
Base period = trailing 52-week median = 100. Where less than 52 weeks of history exist, we rebase to whatever window is available and disclose the actual window in each index payload (the rebase_window_days field).
Indicative until thresholds are met. An index displays the "Indicative" badge when its weight-adjusted total comparable count falls below a per-index threshold: SWCI 20, SWII 20, SWPI 10, SWFI 30. The gate sums each component's comp count multiplied by its published weight — it is not a per-component minimum, and a well-supplied heavy component can carry a thin light one. (This is the current-value comp gate — distinct from the 28-day MoM comparison window.) Per-component counts are published as component_n_comps in indices.json.
Outlier trimming. Within each SKU basket, we drop the 1st and 99th percentile per pull before computing medians.
Geometric mean composition. Each composite index is a weighted geometric mean of its component values, normalized to its base-period median. Components carry explicit, unequal weights (SWCI, for example, weights frontier GPU 0.35 and HBM 0.25 against 0.04 for server DDR4 RDIMM). Every weight is published per index as component_weights in indices.json. A component drops out of the composite when it yields no value for the window, and the surviving weights are renormalized. Producing no value is not the same as reporting no comparables: a component can carry a prior value forward with zero new comps and keep its full weight, and a component can log comps yet still yield no value. As of the 2026-09-04 print the renormalized share is SWCI 20%, SWII 30%, SWFI 30%. SWPI and SWFI publish the applied weights directly as component_weights_used; SWCI and SWII do not yet carry that field, so their effective weights must be derived by dropping the null-valued components from component_weights and renormalizing the remainder. Extending component_weights_used to every index is tracked as a payload change, not made here.
Diffusion vs. price. SWPI in its final form is a diffusion index (stack stress), not a $ index. Components include queue depth and backlog growth alongside spot LMP.
Period comparison = month-over-month (28-day lookback). The displayed delta compares the current rolling 7-day window to the same 7-day window 28 days prior. See Period comparison below.
Disclosure of partial methodology. Indices in transition (currently SWPI and SWFI) carry secondary badges indicating their methodology stage.
Period comparison — MoM
SpotWire publishes index levels monthly. The period-over-period
delta shown on each index card is month-over-month
(MoM) — the current rolling 7-day window median vs. the same 7-day
window 28 days prior.
What a MoM delta does not tell you.
A 28-day lookback measures a change in the composite, which is not the same
as a change in transacted price. Where components are carried at a mechanically
fixed level rather than re-measured each window — currently 80% of SWII by
weight — a MoM figure can move without any underlying transaction. In the
2026-09-04 print SWII's series is eleven identical points followed by a single
step, and both of its volume-varying components reported zero new comparables
that window. That delta cannot be attributed to current transacted-price
movement and should not be read as one. Separately, SWII carries the
Indicative badge because its
weight-adjusted comp count is 2.75, below the threshold of 20. Per-component
counts, values and weights are published in
indices.json.
Why not week-over-week?
Physical infrastructure pricing does not move on a 7-day rhythm.
Secondary GPU markets can go 8+ days without a credible completed
transaction; EDGAR backlog figures update quarterly; BLS PPI releases
monthly. A 7-day comparison window amplifies collection-density noise
rather than genuine price signal, producing erratic swings that obscure
structural moves. This is the same reason ISM and S&P Global publish
their PMI indices on a monthly cadence despite surveying weekly
underlying data — the comparison window should match the natural
frequency of the signal, not the publication cadence.
The 28-day window gives each index component enough observations to
produce a stable median, smooths weekend and holiday collection gaps,
and aligns with how infrastructure operators and procurement teams
actually think about price movements. The first MoM delta will appear
from Issue 2 onward.
Field name in indices.json: mom_pct
· Effective from Issue 2 (2026-07-13)
Data source tiers
Every component on every index is classified by data-source tier.
These are about authority, not quality — eBay can be a
perfectly accurate signal for the secondary market it covers; it
just doesn’t represent the supply chain as a whole. The
per-component classification and an aggregate source_mix
block are published in the exported indices.json.
Tier
Definition
Examples
T1
OEM direct list pricing · authoritative survey series · hyperscaler public APIs
Dell.com, HPE channel, Supermicro configurator, BLS PPI, AWS Price List
T2
Channel partners with vendor MSRP/discount disclosure
RT-1 (red-team finding #1): the V3 weighting still gave eBay
30–50% of the supply-chain indices. V4 caps eBay at 20% on
every index and publishes the trust-tier mix so a reader can see
exactly how much of each value comes from authoritative sources
versus secondary-market spot tape. Full derivation in
docs/research/swpi_swfi_weights_methodology.md
.
Versioning & corrections
SpotWire publishes a dated methodology changelog and a corrections policy. Every methodology
revision is versioned so historical index values remain reproducible. See the full
methodology changelog for version history and the corrections policy.
By index
SWCICompute
V4 weighted · RT-1 channel reweight
Seven-component basket, weighted geometric mean rebased to the
trailing 168-day median. V3 (May 2026) added HBM stress and
frontier GPU OEM pricing so the basket reflects the real 2026 AI
buildout (B200 / GB200 / MI300X) instead of EOL A100/H100 gear alone.
Rationale (V4, May 2026 — RT-1):
V3 still gave eBay (Tier 4 secondary market) 50% of SWCI. V4
caps eBay at 20% and redistributes the freed weight to Tier 1
sources — frontier GPU OEM (Dell/HPE/Supermicro direct: 35%),
HBM stress (TrendForce + Micron filings: 25%), and hyperscaler
public APIs (20%). The secondary market is still represented
but no longer dominates a supply-chain index.
Source mix: 67.5% Tier 1 · 12.5% Tier 3 · 20% Tier 4 (eBay)
YoY for SWCI is backfilled from public-source research prior to 2027-Q2. See
YoY backfill disclosure below.
Methodology consistency (RT-21, May 2026):
Both current-week and historical SWCI composites use the same V4
methodology: each component's basket value is rebased against its
own trailing window median = 100 BEFORE the weighted geometric mean.
HBM and frontier-GPU OEM components return None in historical
months (no public backfill series) — historical composites use 5 of 7
components honestly, with weight renormalization. Current composite
uses up to 7 components as the live data flows.
SWIIInterconnect
Six-component weighted geometric mean. InfiniBand carries the largest tied
weight (25%) because Dell’Oro tracks InfiniBand at roughly 80% of AI training
fabric share — SWII is intended to reflect AI-infrastructure interconnect
specifically. RT-15 adds direct-attach copper and active optical cables (DAC/AOC),
which often account for 30–40% of AI cluster interconnect spend. RT-16 adds
an NVSwitch/NVLink fabric component sourced from OEM rack-config channels (the
GB200 NVL72 bundles 18 NVSwitches per rack); standalone NVLink Switch System
pricing is largely RFQ-only, so this component remains thin until OEM channels
expose more pricing.
RT-1 (May 2026):
eBay (Tier 4 secondary) reduced from 35% to 20%; weight
redistributed to channel partners and OEM direct.
Source mix: 20% Tier 1 (OEM direct) · 60% Tier 2 (channel partners) · 20% Tier 4 (eBay)
YoY for SWII is not displayed until 52 weeks of SpotWire's own index snapshots accumulate.
SWPIPower
V2 diffusion · weights research-derived
SWPI is a five-component diffusion index. Each component is normalized to its
own trailing 24-month median (= 100) and then composited by the weights below.
Values above 100 mean the grid stack is more stressed than its trailing baseline.
Component
Source
Weight
Queue depth
LBNL queue + EIA-860M
30%
Backlog growth
EDGAR cohort (GEV, ETN, PWR, et al.)
25%
Lead-time signal
EDGAR text-mining
20%
PPI electrical
BLS 4-series blend
15%
LMP
ERCOT + 6 RTOs (ISO-weighted)
10%
Rationale (V2, May 2026):
Queue depth, backlog growth, and lead-time signal together carry 75% of the
weight because Sightline Climate, JLL, and CBRE all identify the physical
electrical-supply layer — transformers, switchgear, batteries — as the binding
constraint on 2026 AI datacenter buildout (only 5 GW under construction against
16 GW announced). LMP is downweighted from V1's 25% to 10% because it is
volatile daily noise dominated by weather and gas, not a structural read on
AI stack stress. ISO-weighted LMP within the LMP component is ERCOT 30, PJM 25,
SPP 15, MISO 10, CAISO 10, NYISO 5, ISO-NE 5 (renormalized when an ISO is
unavailable). The dominant components (EDGAR backlog, BLS PPI electrical,
LBNL queue) update monthly or quarterly — the MoM comparison cadence is
intentional: it matches the natural frequency of the data rather than
amplifying daily LMP noise or weekly collection gaps.
GEV equipment vs RPO (RT-19, May 2026):
GE Vernova's headline backlog ($150-163B in earnings press
releases) is Remaining Performance Obligations (RPO) — includes
multi-year service contracts. SWPI's backlog component wants
equipment-only (transformers, switchgear, turbines in the
pipeline). The pipeline extracts Power + Electrification
Equipment RPO directly from GEV's 10-K segment disclosure
($55.2B for FY2025) and applies a documented 0.40 haircut to
press-release RPO figures as a proxy when only the headline is
available. Calibration: 10-K equipment ($55B) ÷ Power +
Electrification total RPO ($129B) ≈ 0.43, rounded down to 0.40
conservatively.
SWFI is a five-component hybrid index: two spot eBay baskets (T&M, PLC), two
BLS PPI series (process instruments, HVAC), and one EDGAR cohort diffusion
signal proxying cooling adoption from the otherwise RFQ-only liquid-cooling
market. Each component is normalized to its own trailing-window median.
Rationale (V3, May 2026 — RT-1):
eBay (Tier 4 secondary) reduced from 30% to 20%; PPI HVAC and
CDW cooling channel each lifted to 20%. The four Tier 1/2/3
disclosure-grade components now carry 80% of the index. JLL
projects AI fit-out costs exceeding $25M/MW, dominated by
cooling complexity — so cooling-channel pricing (Vertiv,
Liebert, InRow on CDW) and the BLS HVAC PPI carry the
cooling-stack signal alongside EDGAR sentiment. BLS PPI and
EDGAR update monthly or quarterly; the MoM comparison cadence
reflects this — a 7-day delta on monthly data is noise, not signal.
Source mix: 40% Tier 1 (BLS) · 20% Tier 2 (CDW) · 20% Tier 3 (EDGAR) · 20% Tier 4 (eBay)
EDGAR sub-cohort split (RT-13, May 2026):
The 7-name cooling/automation cohort blends structurally
different businesses. To avoid smearing Vertiv’s direct
AI-datacenter cooling adoption signal with general industrial
noise, the EDGAR component is now a weighted blend across four
sub-cohorts:
SpotWire's own weekly scrape history begins in 2026-Q2. To enable honest YoY
computation before 52 weeks of native data accumulate, SWCI uses approximately
twelve months of source-attributed historical monthly price points researched from
public secondary-market reports, vendor announcements, and resale tracker articles.
Each backfill entry includes a source URL and a confidence level: high
(dated transactions or vendor effective dates), medium (industry roundups
citing prices for a month/quarter), or low (aggregator estimate). Coverage
is on the sparse end — roughly four monthly points per SKU on average across 17
basket SKUs — and gaps are honest. The full source-attributed disclosure dataset is available on request
([email protected]).
From 2027-Q2 forward, YoY for SWCI is computed exclusively from SpotWire's own scraped
snapshots and the backfill is retired.
Update schedule
Index figures on this site bake at build from the most recently committed print
and update when that print is published and deployed. The figure displayed on
the landing page reflects the most recent successful committed print; the date appears next to
"Source: SpotWire" in the index strip footer. Each monthly issue publishes the most
recent index print alongside a free deep dive. Index levels are meaningful
across collection windows. The MoM delta is meaningful issue-to-issue — most underlying
components (BLS PPI, EDGAR, LBNL queue) refresh monthly or quarterly, so the delta
reflects genuine structural movement rather than collection-window noise.
Outlier handling
Within each underlying SKU basket, we trim the 1st and 99th percentile of
completed-transaction prices before computing medians. Listings flagged as
"parts only," "for repair," or visibly bulk-lot are excluded by the scraper
pipeline at ingest time, before they reach the index basket.
Volume disclosure
Each index card displays per-component n
values — the number of completed-transaction comps (or upstream
observations, in the case of PPI / hyperscaler series) that contributed
to the current week's basket median for that component. A low
n means the component is
thinly evidenced that window. There is no per-component comp minimum:
each component's n enters the
Indicative gate multiplied
by that component's weight, and the badge is set from the weight-adjusted
total against the per-index threshold described above.
Volume is published alongside price for transparency. Components carry
fixed, unequal weights set by methodology — not by trailing-window
comp volume. A component that still yields a value holds its full assigned
weight however few comps back it — in the 2026-09-04 print SWII's networking
and transceiver components carry their full 0.15 and 0.05 on zero new
comparables. A component that yields no value at all is dropped and its
weight is renormalized across the survivors. In V2, weighting by
trailing-window liquidity is under consideration. Until then the
per-component n readout and the
published component_weights let
readers judge concentration for themselves.
Questions, corrections, or methodology suggestions:
[email protected]