packet.ai review

Facts from packet.ai’s own pricing and docs pages, September 2026 — 102 read from a quoted sentence on the page we archived.

Summary

packet.ai rents GPUs from $0.39 per GPU-hour on demand, bills hourly, and asks no minimum commitment. Egress is free.

SOC 2 Type IIGDPR · DPANo training on your data99% uptime SLASpend limitsRefundable balanceVolume discountsFree egressSSOOpenAI-compatible API

From

$0.39/GPU-hr

Billing

Hourly, per-second metered

Minimum commitment

None on hourly

Egress

Free

Uptime SLA

99% monthly (Dedicated)

Registered in

🇺🇸

Visit packet.ai ↗

On-demand pricing

NVIDIA L40S 48GB

×1, on demand · L40S 48GB

$0.92/hr

Deploy

NVIDIA A100 80GB

×1, on demand · A100 80GB

$1.43/hr

Deploy

NVIDIA RTX 4090 24GBLowest

×1, on demand · RTX 4090 24GB

$0.39/hr

Deploy

See all 7 packet.ai offers →

What’s good

Lowest published on-demand rate is $0.39/GPU-hr (NVIDIA RTX 4090 24GB), with no minimum commitment

Egress is free — no per-GB charge on data leaving

Billing is hourly

No KYC or approval needed to start

Publishes a 99% monthly (Dedicated) uptime commitment, with service credits when it is missed

Free credits available: $10 + 10k tokens

Plans

	Dynamic	Dedicated	Bare metal	Clusters
From	$0.66/GPU-hr (RTX 6000 Pro)	$0.39/GPU-hr (RTX 4090)	$5.90/GPU-hr (B200)	~30% below retail
Minimum term	None	None; 6/12-mo optional	Fixed prepaid term	2–3 yr commit
Billing	Hourly	Hourly or monthly	Fixed price per term	By quote
Uptime SLA	Best effort	99% monthly	99% monthly (Dedicated Services)	Dedicated SLAs
Tenancy	Shared, scheduler-isolated	Single-tenant GPU	Whole 8-GPU node	Dedicated multi-node, InfiniBand
Provisioning	Under 5 min	5–10 min	—	2–6 weeks
Payment	Prepaid wallet	Wallet or card subscription	Invoice or bank wire	Order Form

Details

Billing & access

How you pay

Prepaid wallet · subscription · invoice

Prepaid wallet (usage-based); monthly subscription; prepaid term for Bare Metal; commitment billing via Order Form

Currency

USD

Commitment discounts

Up to 20% monthly; more on 6/12-mo

Monthly up to 20% off hourly; 6/12-month additional; rollover credits for unused hours

Free credits

$10 + 10k tokens

$10 wallet credit for new accounts; 10,000 free Token Factory tokens

Approval or KYC

None (sanctions screening only)

None for self-serve; sanctions screening in terms

Auto-renewal

Monthly

Commitment terms auto-renew monthly; subscriptions renew monthly

When funds run out

Instances paused

Instances may be paused/suspended at zero balance

Late payment

1.5%/month

1.5% per month interest on overdue amounts

Billing dispute window

30 days

30 days after month end

Provisioning guarantee

Not guaranteed

Not guaranteed unless in a written agreement

What you get

GPUs listed

RTX 4090, RTX 5090, RTX 6000 Pro, L40S, A100 80GB, H100 SXM, H200, B200

Regions

California, Virginia, Texas, Oregon, Frankfurt, Amsterdam, Paris, London, Dublin

Node configuration

8 GPUs · 2 TB RAM · NVMe · 100 Gbps

8 GPUs per node; Dedicated nodes: 2 TB RAM minimum, local NVMe, 100 Gbps

Interconnect

clusters: InfiniBand, 400 Gb/s per node, NVLink 1.8 TB/s

dedicated: 100 Gbps per node

Datacenter tier

Tier-3 (Dedicated tier)

Local storage

NVMe, up to 2 TB/node, free

Local NVMe included, up to 2 TB per node; ephemeral, cleared on terminate

Persistent storage

NFS, $0.10/GB-mo

kind: NFS network volume; price: $0.10/GB/month; billed_while_stopped: True

Ingress

Free

Public endpoints

Public IP:port per service

Expose a port to a public IP:port (LoadBalancer/NodePort); SSH port-forwarding alternative

Images

Ubuntu 22.04, CUDA 12, PyTorch, vLLM, Jupyter

Ubuntu 22.04 LTS with CUDA 12.x/cuDNN, PyTorch, vLLM, Jupyter preinstalled

Access methods

SSH, web terminal, dashboard, CLI, REST API

Model deployment

One-click HuggingFace (vLLM)

One-click HuggingFace model deployment with vLLM

Managed inference

Token Factory (OpenAI-compatible)

Token Factory — OpenAI-compatible API, pay per token

FAQ

How much does packet.ai cost?

As of September 2026, the lowest published on-demand rate is $0.39 per GPU-hour for the NVIDIA RTX 4090 24GB. packet.ai publishes an on-demand rate for 3 GPU models. Billing is hourly.

Does packet.ai have an uptime SLA?

packet.ai states 99% monthly (Dedicated). Service credits: 10–30% of monthly fee.

Is there a minimum commitment on packet.ai?

No. packet.ai states none on hourly.

Does packet.ai charge for egress?

No. packet.ai states that egress is free.

Can I sign up to packet.ai without approval?

packet.ai states that signup is self-serve. On approval and identity checks it states: none (sanctions screening only).

Are packet.ai payments refundable?

packet.ai states: unused wallet funds refundable.

Where are packet.ai's data centres?

packet.ai lists california, Virginia, Texas, Oregon, Frankfurt, Amsterdam, Paris, London, Dublin.

Who do you contract with at packet.ai?

The contracting entity is Hosted AI Inc (Delaware).

Reviewed September 2026, from packet.ai’s own site. 102 facts, each quoted from the page it was read on.