gpurank
esc
  • Nothing named “”. Press Enter to search the offers instead.
↑↓ to move · ↵ to open · esc to close0 entries
Operator · United States

Voltage Park

US owner-operator of NVIDIA HGX H100 fleets (now merged into Lightning AI) selling self-serve H100 VMs and 8-GPU bare-metal nodes by the hour, plus Blackwell systems on long-term contracts only.

GPU types1with a current offer
Current offers1515 readings on file · last 45 days
H100 from—per GPU-hour, on-demand
Entry price—cheapest GPU-hour listed
Countries12 named regions
Trust Score v0.1
89/100
  • Company & track record73
  • Infrastructure control100
  • Pricing & billing transparency100
  • Reliability signals75
  • Security & compliance100
  • Customer signal—

Scored Sep 29, 2026 from the provider's own pages. Dimensions not yet measured are excluded from the denominator. Criteria ↓

Trust Score · 5 of 6 dimensions measured

The score, criterion by criterion

Each point comes from something the provider publishes, with the page it was read on and the date. Anything not disclosed earns zero.

Company & track record
weight 15%
73 / 100
Founding year publishednot disclosedvoltagepark.com · Sep 29, 20260 / 4
Headquarters publishedVoltage Park, Inc.voltagepark.com · Sep 29, 20263 / 3
Legal entity namedVoltage Park, Inc.voltagepark.com · Sep 29, 20264 / 4
Funding or ownership disclosedPrivate. About page cites a '$1 billion long-term capital foundation'. Merger with Lightning AI completed January 2026; the combined company operates under the Lightning AI name.voltagepark.com · Sep 29, 20264 / 4
Infrastructure control
weight 15%
100 / 100
Capacity model publishedoperates or rents dedicated capacityvoltagepark.com · Sep 29, 202610 / 10
Regions published2 regions or countriesvoltagepark.com · Sep 29, 20265 / 5
Pricing & billing transparency
weight 20%
100 / 100
Public prices without an accountyes (read on the pricing page)voltagepark.com · Sep 29, 20265 / 5
Billing granularity publishedper hourvoltagepark.com · Sep 29, 20264 / 4
Egress policy publishedincludedvoltagepark.com · Sep 29, 20264 / 4
Stopped-instance billing publishedStopped instances are charged at the full running rate; restart takes about 20 minutes. Only termination stops billing.voltagepark.com · Sep 29, 20264 / 4
Public IPv4 policy publishedincluded for bare metal; VMs sit behind a NAT firewall with selected ports forwardedvoltagepark.com · Sep 29, 20263 / 3
Reliability signals
weight 20%
75 / 100
Status pagePer-data-centervoltagepark.com · Sep 29, 20266 / 6
SLA published99.5% monthly uptime per VM (VM SLA) and per node (Node SLA); credits 10% (95-99.5%), 25% (90-95%), 100% (below 90%)voltagepark.com · Sep 29, 20266 / 6
SLA with a numeric uptime commitment99.5% monthly uptime per VM (VM SLA) and per node (Node SLA); credits 10% (95-99.5%), 25% (90-95%), 100% (below 90%)voltagepark.com · Sep 29, 20263 / 3
Uptime history publishednot disclosedvoltagepark.com · Sep 29, 20260 / 5
Observed availability (30 days)not measured yet—
Security & compliance
weight 15%
100 / 100
SOC 2Type IIvoltagepark.com · Sep 29, 20268 / 8
ISO 27001ISO 27001:2022 certified (Trust Center)voltagepark.com · Sep 29, 20264 / 4
HIPAA or GDPR DPAStated compliant for eligible workloadsvoltagepark.com · Sep 29, 20263 / 3
Customer signal
weight 15%
not measured
Aggregated public reviewsnot collected yet—
provisional v0.1: evidence from the provider's own pages; dimensions not yet measured are excluded from the denominatorFull method →
Market report · Oct 25, 2026 · Complementary actors (report v2) · prices read on the provider's page

What the market report notes about Voltage Park

Our internal report lists 248 actors with their zone, segment, status and the prices it could read, dated Oct 25, 2026. Its notes are reproduced as written, in French, with their sources.

Listed as
Voltage Park
Zone
United States
Segmentfr
GPU cloud / clusters
Priority in the reportfr
A → A
Statusfr
Actif — self-service H100 + clusters Blackwell sur contrat
Evidence levelfr
Corroboré (computeprices 05/10/2026, Spheron juil. 2026, page comparatif officielle)
News notedfr
Fusion avec Lightning AI finalisée le 21/01/2026 (marque combinée Lightning AI, page tarifaire Voltage Park maintenue). Flotte annoncée 35 000–36 000 GPU (H100, B200, GB300). B200/B300/GB200/GB300 et InfiniBand uniquement sur contrat.
Prices notedfr
H100 SXM : 1,99 $/GPU-h à la demande, sans contrat, 1 à 1 016 GPU en self-service (relevé 05/10/2026) ; variantes 1,89–2,49 $ selon agrégateur ; blocs multi-mois remisés (3, 6, 12 mois). SOC 2, HIPAA. 2 régions US.
Prices by GPU · 1 GPU types with a current offer

What Voltage Park charges, against the market

Cheapest current on-demand true cost here, next to the market median for the same GPU. Negative means cheaper than the market.

GPUCheapest hereMarket medianvs marketCurrent offersModes
H100$1.89$3.17−40%6on-demandH100 prices
All offers · 15 readings on file

Every Voltage Park offer, with its receipt

Current offers (last 45 days) ranked by true cost, all rental modes, each with its read date. Older readings stay listed below, in grey.

GPUConfigurationModeRegionStockListTrue costRead
H100SXM8× H100 SXM208 vCPU · 928 GB RAM · 19,200 GB diskOn-demandUSUnavailable$1.89$1.8921:10 UTCfreshReceipt
H100SXM1× H100 SXM26 vCPU · 116 GB RAM · 2,400 GB diskOn-demandUSUnavailable$1.89$1.8921:10 UTCfreshReceipt
H100SXM4× H100 SXM104 vCPU · 464 GB RAM · 9,600 GB diskOn-demandUSUnavailable$1.89$1.8921:10 UTCfreshReceipt
H100SXM2× H100 SXM52 vCPU · 232 GB RAM · 4,800 GB diskOn-demandUSUnavailable$1.89$1.8921:10 UTCfreshReceipt
H100SXM8× H100 SXM104 vCPU · 1,024 GB RAM · 18,000 GB diskOn-demandUSUnavailable$1.99$1.9921:10 UTCfreshReceipt
H100SXM8× H100 SXM104 vCPU · 1,024 GB RAM · 18,000 GB diskOn-demandUSUnavailable$2.49$2.4921:10 UTCfreshReceipt
6 fresh offers · prices per GPU-hour · true cost includes the add-ons the provider publishesOpen in the comparison tool →
True cost & billing · read Sep 29, 2026

What the hourly price does not say

Billing granularity, minimums, what a stopped machine costs, traffic, addresses, deposits and identity checks, as published by the provider.

Billing

Billing granularity
per hour
Minimum charge
not published
Stopped instance
Stopped instances are charged at the full running rate; restart takes about 20 minutes. Only termination stops billing.
Egress
included
Public IPv4
included for bare metal; VMs sit behind a NAT firewall with selected ports forwarded
Billing model
prepaid credits
Minimum deposit
not published
Free credits
not published

Account

Payment
card · invoice · crypto (not published)
Sign-up
email, password
Identity check
not published
Support
email, slack, help center chat
Enterprise support
not published
SLA
99.5% monthly uptime per VM (VM SLA) and per node (Node SLA); credits 10% (95-99.5%), 25% (90-95%), 100% (below 90%)
Published uptime
not published

Infrastructure

Regions
2 named regions in 1 countriesUS data centers (six Tier 3+ sites · not named) (United States)
Countries
United States
Platform
api, jupyter, ssh, persistent_volumes, bare_metal · SDK : — · Terraform : — · Kubernetes : managed · interconnexion : 3,200 Gbps InfiniBand (bare metal), 100 Gbps Ethernet (VMs and Ethernet bare metal), NVLink within HGX · clusters : Self-serve InfiniBand bare metal from 8 to 1,016 H100 GPUs · long-term reserve from 32 to 8,000+ GPUs
Storage
volume · 0.10 USD/hour per 2,400 GB bundled with each GPU of a VM (about 0.03 USD/GB-month at 720 h) · Charged as storage_rate_hourly with VM presetsnetwork · not published · VAST-backed shared storage volumes, rate shown only in the dashboard/authenticated API
Migration out
Instances run Ubuntu 22.04 with SSH access; customers copy data out over SSH/rsync or from VAST NFS shares before terminating. Terminated instance data cannot be recovered, and Voltage Park has no backup obligation under the Compute Services Agreement.

Security & compliance

SOC 2
Type II
ISO 27001
ISO 27001:2022 certified (Trust Center)
HIPAA
Stated compliant for eligible workloads
GDPR / DPA
not published
Other
SOC 2 Type 1 report also listed; Trust Center reports available on request
Legal entity
Voltage Park, Inc.
Funding
Private. About page cites a '$1 billion long-term capital foundation'. Merger with Lightning AI completed January 2026; the combined company operates under the Lightning AI name.
GPUBank market study, read on the provider's own pages · 24 pages read · facts marked "not published" were looked for and not foundPages read ↓
From the study · assessment

Voltage Park, in short

What stood out when we read the documents: strengths, weaknesses, who it suits, and what we could not find.

Strengths

  • H100 SXM from 1.99 USD per GPU-hour (VM or Ethernet bare metal) and 2.49 with 3.2 Tbps InfiniBand, read from the official API
  • Self-serve InfiniBand bare metal from 8 up to 1,016 GPUs without a contract
  • Ingress, egress and support stated as included
  • 99.5% SLA for both VMs and bare-metal nodes with up to 100% credit
  • ISO 27001:2022, SOC 2 Type II and HIPAA via a public Trust Center
  • Managed Kubernetes add-on with GPU Operator, MPI Operator, Prometheus/Grafana and VAST CSI
  • Blackwell systems (B200, GB200, B300, GB300) available on long-term contracts

Weaknesses

  • Pricing page itself says 'Contact for pricing' for every tier; per-GPU prices are visible only through the API or the dashboard
  • Official API showed zero available GPUs for bare metal and no VM location with availability at check time
  • Stopped instances bill at the full running rate
  • Single GPU model for self-serve (H100); Blackwell only under long-term contract
  • API keys cannot be generated self-service according to the help center (must be requested)
  • US-only data centers
  • No spot tier, no serverless product, no published crypto payment

Ideal for

  • Teams needing multi-node H100 InfiniBand clusters without a contract
  • US buyers who need ISO 27001 / SOC 2 / HIPAA documentation
  • Enterprises planning 6-12+ month Blackwell or H100 reservations

Not published

  • Shared (VAST) storage price per GB (dashboard only)
  • Whether bare-metal nodes carry a storage charge in addition to the GPU price
  • Exact billing increment (Agreement says charges are continuous and may be more frequent than hourly)
  • Minimum deposit (not published)
  • Crypto acceptance (not published
  • deposits by card or bank)
  • Identity verification: the Agreement says account information is 'subject to authentication' without detail
  • Names and cities of the six US data centers
  • Reserved and Blackwell prices (contact sales)
  • Founding year and headquarters city (not on pages read)

Observations

Pricing page shows three H100 tiers each labelled 'Contact for pricing'; the only visible figure is 'starting at $1.99/hr'

Banner states Voltage Park has merged with Lightning AI; help center is branded Lightning AI Help Center

Dashboard sign-up asks for email address and password

A 'Compare' page lists Lambda, Nebius, Crusoe and Runpod as named competitors

Pages read

Questions about Voltage Park

What GPUs does Voltage Park offer?

Voltage Park has 1 GPU types with a fresh offer today, for example H100 from $1.89 per GPU-hour on-demand. The full list with prices and market medians is on this page.

How does Voltage Park bill?

Billing granularity: per hour. Stopped instance: Stopped instances are charged at the full running rate; restart takes about 20 minutes. Only termination stops billing.

How trustworthy is Voltage Park?

Its Trust Score is 89/100 (v0.1), read from its own documents. Strong on infrastructure control, pricing & billing transparency, security & compliance. The score measures transparency and published guarantees, not customer satisfaction; every criterion and its source is listed above.

Where are Voltage Park's data centres?

2 named regions in 1 countries: United States.

How do you pay Voltage Park, and are there credits?

Payment by card, invoice, crypto. No sign-up credit is published.