Reference

Methodology

SpotWire publishes infrastructure intelligence from a mix of observed prices, quoted or listed prices, and proxy signals. The payload for each collection snapshot discloses which inputs are present rather than treating every component as a transaction.

Current collection snapshot: September 4, 2026 · immutable snapshot JSON

Principles

Collection snapshot, not issue identity. as_of describes the observation date. snapshot_id identifies the immutable content-derived payload; editorial cadence is separate.
Index-specific windows. Each index exposes its own rebase_window_days, data_through, and comparison fields when available. The site does not substitute a universal lookback.
Configured versus effective weights. Configured weights are the formula inputs. Effective weights are shown only when explicitly reported by the index engine; the website does not infer them from raw component values.
Source-aware measurement. The table preserves each component's reported primitive, including transaction prices, quotes, published series, and proxies. These are not interchangeable evidence types.
Missing values stay missing. A null component, count, or comparison remains an em dash in the public display. Sparse or single-source conditions are disclosed through payload diagnostics where emitted.
Formula preservation. Composite formulas and weights remain defined by the index engine. This page reports the current payload; it does not infer or alter a formula when an input is unavailable.

Data source tiers

Tiers describe source authority and provenance, not a guarantee that all inputs share one price primitive. The per-component tier, primitive, source structure, and source mix are carried in each index record.

Tier	Definition	Examples
T1	Authoritative primary or published series	OEM/channel feeds, BLS PPI, hyperscaler APIs, industry research and filings
T2	Integrator or channel observation	Integrator surveys and distributor or channel list prices
T3	Regulatory and company disclosure	EDGAR backlog, RPO, delivery language, and cohort signals
T4	Secondary-market observation	Observed marketplace prices where the component uses a secondary source
T5	Public infrastructure dataset	ISO/LMP, queue, and other public infrastructure observations

Current payload disclosures

The tables below are rendered from the current v2 snapshot. Counts are source-specific observations, not a count of all transactions. A primitive tells you whether a value is observed, quoted, or proxy-based; null values are shown as missing.

SWCICompute

snapshot as of 2026-09-04 · data through 2026-09-04 · rebase window 84 days

Legacy GPUs · Servers · Enterprise SSD · Server DDR4 RDIMM · Hyperscaler GPU rental · HBM stress · Frontier GPU OEM

Component	Value	Observed n	Configured	Effective	Primitive	Tier
gpu	—	0	6.0%	—	ebay transacted	tier 4 ebay secondary
server	—	0	6.0%	—	ebay transacted	tier 4 ebay secondary
ssd	—	0	4.0%	—	ebay transacted	tier 4 ebay secondary
ram	—	0	4.0%	—	ebay transacted	tier 4 ebay secondary
hyperscaler	4.00	105	20.0%	—	cloud quoted	tier 1 hyperscaler api
hbm	100.00	7	25.0%	—	stress proxy	tier 1 industry research and filings
frontier gpu	315618.77	131	35.0%	—	oem list	tier 2 integrator channel survey

Active observations used: 243 · index flag: not marked indicative · source mix is disclosed in the index record.

SWIIInterconnect

snapshot as of 2026-09-04 · data through 2026-09-04 · rebase window 28 days

Networking eBay · Transceiver eBay · Ethernet optics survey · InfiniBand survey · DAC/AOC survey · NVSwitch OEM

Component	Value	Observed n	Configured	Effective	Primitive	Tier
networking ebay	100.00	0	15.0%	—	ebay transacted	tier 4 ebay secondary
transceiver ebay	70.75	0	5.0%	—	ebay transacted	tier 4 ebay secondary
fs new list	100.00	7	15.0%	—	channel list	tier 2 channel partner
infiniband fs	—	0	30.0%	—	channel list	tier 2 channel partner
dac aoc fs	100.00	10	15.0%	—	channel list	tier 2 channel partner
nvswitch oem	100.00	1	20.0%	—	oem list	tier 2 integrator channel survey

Active observations used: 18 · index flag: Indicative · source mix is disclosed in the index record.

SWPIPower

snapshot as of 2026-09-04 · data through 2026-09-04 · rebase window 730 days

Nationwide LMP · Queue depth · Backlog growth · Lead-time · BLS PPI

Component	Value	Observed n	Configured	Effective	Primitive	Tier
lmp	101.75	1	10.0%	10.0%	iso settlement price	tier 5 public infrastructure
queue depth	98.86	2,630	30.0%	30.0%	infrastructure dataset	tier 5 public infrastructure
backlog growth	153.66	5	25.0%	25.0%	edgar filing signal	tier 3 edgar filings
lead time	66.67	14	20.0%	20.0%	edgar filing signal	tier 3 edgar filings
ppi	106.69	4	15.0%	15.0%	gov price index	tier 1 bls ppi

Active observations used: 2,654 · index flag: not marked indicative · source mix is disclosed in the index record.

SWFIFacility

snapshot as of 2026-09-04 · data through 2026-09-04 · rebase window 730 days

T&M · PLC · PPI process instruments · PPI HVAC · EDGAR cohort sentiment · CDW cooling channel

Component	Value	Observed n	Configured	Effective	Primitive	Tier
tm ebay	—	17	10.0%	—	ebay transacted	tier 4 ebay secondary
plc ebay	250.00	253	10.0%	14.3%	ebay transacted	tier 4 ebay secondary
ppi process	103.55	24	20.0%	28.6%	gov price index	tier 1 bls ppi
ppi hvac	104.28	24	20.0%	28.6%	gov price index	tier 1 bls ppi
edgar sentiment	98.08	4	20.0%	28.6%	edgar filing signal	tier 3 edgar filings
cooling	—	0	20.0%	—	channel list	tier 2 channel partner

Active observations used: 322 · index flag: not marked indicative · source mix is disclosed in the index record.

Period comparisons

MoM and YoY values are displayed only when the current payload supplies them. They describe comparisons calculated by the engine; the website does not relabel a missing comparison as zero or assume that collection timing equals an editorial schedule.

2026-09-04 snapshot: interpretation limits

These are dated disclosures for the September 4, 2026 snapshot, documented in methodology v0.7.1 on 2026-09-06. They are not assumptions imposed on every future snapshot. Publication schema v2 is separate from the methodology and engine software versions; it does not clear the upstream reconciliation hold recorded in the dated correction history.

Weighted composition and missing components. Configured weights were unequal. Drop-out was keyed to no value, not to zero new comparables; surviving weights were renormalized. The renormalized configured share was SWCI 20%, SWII 30%, SWFI 30%. A component could retain a carried value on zero new observations, or report observations yet yield no value. SWPI and SWFI reported component_weights_used; SWCI and SWII did not. The current tables show only explicitly reported effective weights.
Indicative gate. The production compositor used a weight-adjusted total of component observation counts against thresholds SWCI 20, SWII 20, SWPI 10, SWFI 30. The pre-RT-20 per-component gate was real and was replaced on 2026-05-29; it was not a rule that had never existed.
What a MoM delta does not tell you. In this snapshot, 80% of SWII by weight was mechanically carried between refreshes. Its series had eleven identical points followed by a single step, with zero new comparables in both volume-varying components. That delta cannot be attributed to current transacted-price movement. Separately, SWII was Indicative because its weight-adjusted comp count was 2.75, below 20.
No retroactive math change. The September 6 entry corrected disclosure and citation history; no index level, weight, or threshold changed. Frozen archives are not rewritten. The historical issue-number crosswalk remains in the changelog, but current v2 snapshots do not acquire an issue_no.

Versioning & corrections

SpotWire publishes a dated methodology changelog and corrections policy. Historical source files and collection snapshots remain identifiable at their own URLs; a later snapshot does not rewrite the identity of an earlier one. See the full methodology changelog for version history.

Outlier handling

Where a component uses observed marketplace prices, the source pipeline applies its documented outlier and eligibility rules before the value enters the component basket. Quoted and proxy components follow their own source-specific processing; no single transaction rule is applied to every component.

Questions, corrections, or methodology suggestions: [email protected]