inference/research
draftmedium · 0.75updated 2026-10-04

Peer-to-peer rental index: consumer GeForce (4090 / 3090 / 5090 per-GPU-hour)

nvidia · global (16-17 host countries observed per SKU)
This record is a draft. Not yet verified. Provisional. Treat every number on this page as provisional.

Fields

id
supply/rental-consumer-4090-3090-5090-vast-index
Kind
cloud
Vendor
nvidia
Region
global (16-17 host countries observed per SKU)
Channels
rentalmarketplace
Price (USD)
0.4289
Price basis
USD per GPU-HOUR for the GPU component, verified peer-to-peer hosts, 1-GPU bundles, observed 2026-10-03. Asking rental rate, NOT a realised rental. GeForce cards, not datacenter parts - kept as a separate record so consumer and datacenter bands never mix PERISHABLE-FIGURE STAMP (added 2026-10-04, uniform across data/supply/ so it is greppable). PRICE CLASS: ASKING RENTAL - an advertised rate a host lists on a peer-to-peer marketplace, NOT a realised rental transaction, NOT a broker SLA price and NOT a provider list rate. A host can and does undercut this within hours, so the figure is the least durable in this directory. OBSERVED: 2026-10-03 RE-VERIFY BY: 2027-01-04 - the repo's quarterly re-verification bar. After that date this figure is presumed drifted and must be re-checked against its source before it is quoted again. A price without its observation date is noise; this stamp exists so a reader can tell a fresh figure from a stale one without reading prose. UNIT: USD per GPU-HOUR for the GPU component, verified peer-to-peer hosts, 1-GPU bundles, observed 2026-10-03 READING RULE: No further perishable flag is published for this figure.
Availability
in-stock
Lead time source
structural-na
Lead time basis
lead_time_weeks is null because the quantity is structurally not applicable: on-demand cloud capacity starts on request, so there is no order-to-delivery lead time. The constraint is the price per GPU-hour and the availability snapshot, not an interval. The one caveat worth keeping: scarce tiers gate capacity behind a reservation or Capacity Block commitment (AWS caps the booking horizon at 8 weeks, source spl-aws-capacity-block-booking-horizon), and that is a commitment window, not a lead time.
Export controlled
no
Export control regime
EAR 3A001 or EAR99 (below the 3A090.a TPP / performance-density thresholds) + EAR 744.23 (Entity List) + EAR 744.6 (Unverified / MEU lists)
Export control note
MECHANISM: the part falls BELOW the ECCN 3A090.a/.b TPP and performance-density thresholds, so the advanced-computing licence regime, its case-by-case China/Macau carve-out and its D:5 screening do NOT attach. What remains is the general EAR plus destination-country import rules and screening of the consignee and its ultimate parent against the Entity, Unverified and MEU lists. NO LICENCE REQUIRED for the sale itself; the screening is a party-listing check, not a performance-threshold check. The binding ECCN determination for any specific part is the exporter's own classification, not this record.
notes
OBSERVATION DATE 2026-10-03. Source sup-price-vast-rental-api. THREE DISTINCT BANDS, all observed 2026-10-03, all per-GPU-hour on 1-GPU verified hosts: RTX 4090 n=43, range $0.3024-$0.9356, median $0.4289; RTX 5090 n=41, range $0.4015-$0.7356, median $0.4896; RTX 3090 n=32, range $0.1213-$0.2523, median $0.1498. price_usd records the RTX 4090 median only because the schema has one price slot and 4090 has both the largest sample (n=43) and the tightest spread; the other two medians are carried in notes. THE SEGMENT BOUNDARY, stated as a number: the RTX 3090 rents at ~0.35x the RTX 4090 rate ($0.1498/$0.4289 = 0.35) while its asking PURCHASE price is ~0.46x ($1,514/$3,299 medians, sup-price-picclick-ebay-asking). Rental and purchase price do not move together across these SKUs, which is the practical reason neither number can stand in for the other. Sample sizes here (32-43 per SKU) are large enough that these medians are the most statistically solid figures in this entire slice. EXPORT CONTROL (2026-10-04, mechanism-level; source spl-eccn-3a090-ccl): consumer gaming cards of this generation generally do NOT meet the 3A090.a TPP/performance-density parameters and fall under ECCN 3A001 or are designated EAR99, i.e. the advanced-computing licence regime with its case-by-case and D:5 screening does not attach to them. This is exactly what makes export status a PER-PART property rather than a per-vendor one: a GeForce RTX 5090 and a datacenter/professional part from the same vendor and the same process generation can carry entirely different control status. Recorded as false for that reason, NOT because the part is unregulated - it remains subject to the EAR generally, to destination-country import rules, and to screening of the consignee and its ultimate parent against the Entity, Unverified and MEU lists (15 CFR part 744). The binding ECCN determination for any specific SKU is the exporter's own classification, which no public vendor page states. LEAD TIME: null - and specifically because a PEER-TO-PEER marketplace has no lead time at all. Host nodes appear and disappear, so the availability a buyer sees is a snapshot of what some host posted on the observation date, not a schedulable delivery date. This is the widest availability uncertainty in the whole supply set and it is structural rather than a missing datum: it argues directly for the renting-over-buying decision this record feeds. Probe log in source spl-negative-no-oem-publishes-lead-time.

Notes

OBSERVATION DATE 2026-10-03. Source sup-price-vast-rental-api. THREE DISTINCT BANDS, all observed 2026-10-03, all per-GPU-hour on 1-GPU verified hosts: RTX 4090 n=43, range $0.3024-$0.9356, median $0.4289; RTX 5090 n=41, range $0.4015-$0.7356, median $0.4896; RTX 3090 n=32, range $0.1213-$0.2523, median $0.1498. price_usd records the RTX 4090 median only because the schema has one price slot and 4090 has both the largest sample (n=43) and the tightest spread; the other two medians are carried in notes. THE SEGMENT BOUNDARY, stated as a number: the RTX 3090 rents at ~0.35x the RTX 4090 rate ($0.1498/$0.4289 = 0.35) while its asking PURCHASE price is ~0.46x ($1,514/$3,299 medians, sup-price-picclick-ebay-asking). Rental and purchase price do not move together across these SKUs, which is the practical reason neither number can stand in for the other. Sample sizes here (32-43 per SKU) are large enough that these medians are the most statistically solid figures in this entire slice.

EXPORT CONTROL (2026-10-04, mechanism-level; source spl-eccn-3a090-ccl): consumer gaming cards of this generation generally do NOT meet the 3A090.a TPP/performance-density parameters and fall under ECCN 3A001 or are designated EAR99, i.e. the advanced-computing licence regime with its case-by-case and D:5 screening does not attach to them. This is exactly what makes export status a PER-PART property rather than a per-vendor one: a GeForce RTX 5090 and a datacenter/professional part from the same vendor and the same process generation can carry entirely different control status. Recorded as false for that reason, NOT because the part is unregulated - it remains subject to the EAR generally, to destination-country import rules, and to screening of the consignee and its ultimate parent against the Entity, Unverified and MEU lists (15 CFR part 744). The binding ECCN determination for any specific SKU is the exporter's own classification, which no public vendor page states.

LEAD TIME: null - and specifically because a PEER-TO-PEER marketplace has no lead time at all. Host nodes appear and disappear, so the availability a buyer sees is a snapshot of what some host posted on the observation date, not a schedulable delivery date. This is the widest availability uncertainty in the whole supply set and it is structural rather than a missing datum: it argues directly for the renting-over-buying decision this record feeds. Probe log in source spl-negative-no-oem-publishes-lead-time.

Cross-references