Preamble
These Terms & Conditions ("Terms") govern access to and use of the Packet.ai platform and services (the "Services").
Packet.ai is a product of Hosted AI Inc, a Delaware corporation ("Hosted AI," "we," "us," or "our"). References to the "Packet.ai platform" or "Services" refer to the product operated by Hosted AI Inc.
By creating an account, accessing the Services, or using any Packet.ai functionality, you ("Customer," "you") agree to be bound by these Terms, the Acceptable Use Policy, Billing, Credits & Refund Policy, Privacy Policy, and any applicable order form or commitment (collectively, the "Agreement"), all of which are incorporated by reference.
1. Introduction
1.1. User Agreement
You agree that by registering, accessing or using our Services (described below), you are agreeing to enter into a legally binding contract with Packet.ai ("we", "us", "our", "Packet.ai") (even if you are using our Services on behalf of a company). If you do not agree to these terms of service ("Agreement"), do not access or use our Services in any way. You can terminate this User Agreement at any time by closing your account and no longer accessing or using our Services.
1.2. Changes
We may modify the User Agreement and Privacy Policy from time to time and will provide you with at least thirty (30) days' notice to review material changes to them before they become effective. We agree that changes cannot be retroactive. If you do not agree to the updated terms, you may terminate this Agreement by providing written notice before the effective date of the changes. Your continued use of the Services after the effective date of any changes will be deemed acceptance of the updated terms.
If you do not agree, do not use the Services.
1.3. Definitions
Bare Metal Server. A single-tenant physical GPU server provisioned exclusively for Customer use, delivered with Management Components installed as described in Section 2.3.
Commitment Term. A fixed period (e.g., monthly, quarterly, or annual) for which Customer commits to specific usage fees, as defined in the applicable Order Form.
Dedicated GPU Cluster. A single-tenant cluster of GPU-equipped nodes provisioned exclusively for Customer use.
Dedicated GPU. A single GPU attached to a Kubernetes pod or KVM Virtual Machine and not shared with other customers.
Infrastructure Partner. A third-party data centre, colocation, or infrastructure operator engaged by Packet.ai, or by a supplier of Packet.ai, to operate, maintain, or provide support in respect of the physical infrastructure underlying the Services.
Management Components. Packet.ai telemetry, monitoring, and support agents installed on the host operating system of a Bare Metal Server, which collect host, hardware, and infrastructure-level operational data and enable support access by Packet.ai and its Infrastructure Partners. Management Components are not designed to collect Customer content, datasets, models, workload code, or outputs.
Shared GPU / Overcommitted GPU. A GPU resource shared among multiple customers with no guaranteed performance isolation.
GPU Hour. One GPU consumed for one hour (or fraction thereof), as measured by the platform.
Wallet Balance. Prepaid funds or credits applied toward usage charges.
Usage Metrics. Measured consumption of GPUs, compute, storage, networking, or other billable resources.
Services. The Packet.ai platform, APIs, dashboards, and underlying GPU infrastructure.
Customer. The Customer and such employees, contractors, agents, and authorized sublicensees to whom Customer provides access to the GPU Services (as defined below).
2. Services
Packet.ai provides GPU-based infrastructure services (the "GPU Services"), which may include the following service configurations, as selected by Customer through the Platform or an applicable Order Form:
- Dedicated GPU Clusters, consisting of GPU compute resources allocated exclusively for use by a single Customer;
- Dedicated GPUs, consisting of individual GPU instances allocated exclusively for use by a single Customer;
- Bare Metal Servers, consisting of single-tenant physical GPU servers allocated exclusively for use by a single Customer, delivered and operated in accordance with Section 2.3; and
- Shared or overcommitted GPU resources, including GPU capacity attached to Kubernetes pods or similar orchestration environments, which are shared among multiple customers.
2.2. Shared GPU Characteristics
Shared GPU resources are provided on a best-effort, non-exclusive basis. Customer acknowledges that Shared GPU resources may be subject to performance variability, including but not limited to resource contention, throttling, scheduling delays, preemption, rebalancing, or other infrastructure-level resource management activities. Packet.ai does not guarantee exclusive access, consistent performance, or uninterrupted availability for Shared GPU resources unless expressly stated in an applicable Service Level Agreement or Order Form.
2.3. Bare Metal Server Characteristics
Bare Metal Servers are delivered with Management Components preinstalled and support access enabled by default. Customer acknowledges and agrees that:
- Management Components operate continuously for the purposes of hardware health monitoring, fault detection, capacity management, service level measurement, billing, and platform security;
- authorised personnel of Packet.ai and its Infrastructure Partners may access the host for monitoring, maintenance, fault diagnosis, and support purposes; and
- Management Components form part of the Packet.ai Technology and must not be removed, disabled, or interfered with. Where Customer identifies a conflict between Management Components and Customer's own tooling, including a port or resource conflict, Customer should contact Packet.ai support, who may agree an alternative configuration. Removing, disabling, or interfering with Management Components may result in loss of service level eligibility or support, and may constitute grounds for suspension under Section 19.
Access under this Section is directed to the host and infrastructure layers and is not for the purpose of inspecting Customer content, datasets, models, workload code, or outputs, save to the extent incidental to such access or where expressly requested by Customer. Packet.ai requires each Infrastructure Partner with host access to be bound by written confidentiality obligations.
Information regarding Infrastructure Partners with host access rights is available on request at security@packet.ai.
This Section applies solely to Bare Metal Servers ordered by Customer and does not modify the terms applicable to any other GPU Services.
3. Provisioning, Capacity & Maintenance
Immediate or on demand provisioning of Services is not guaranteed unless Packet.ai has expressly agreed to specific provisioning timelines in a written agreement or applicable Order Form.
Packet.ai may perform scheduled maintenance or emergency maintenance on the Services or underlying infrastructure from time to time. Such maintenance may result in temporary service interruptions or performance degradation. Where commercially reasonable, Packet.ai will use reasonable efforts to provide advance notice of scheduled maintenance.
Packet.ai may substitute hardware components with functionally equivalent hardware as necessary for operational, capacity, maintenance, or supply-chain reasons, provided that such substitution does not materially and adversely affect the functionality of the Services.
The Services may be supported by infrastructure located in multiple geographic regions. Packet.ai may change the location of infrastructure used to provide the Services from time to time, subject to applicable law and any express geographic commitments set forth in an Order Form or applicable agreement.
We will use commercially reasonable efforts to maintain the availability of the Services. We may change, suspend, or discontinue any feature of the Services; however, for paid Subscribers, except as otherwise provided for termination herein or as required by law, we will provide at least thirty (30) days' prior notice of any material reduction in functionality and will continue to provide the contracted level of service through the end of the then-current Subscription period.
Packet.ai will use commercially reasonable efforts to make the paid, Dedicated Services available with a Monthly Uptime Percentage of 99.9%, excluding Excluded Downtime.
For the avoidance of doubt, Shared GPU Services are provided on a best-effort basis only and are expressly excluded from this uptime commitment and any related service level obligations or credits.
Monthly Uptime Percentage = (Total Minutes in Month − Downtime) / Total Minutes in Month × 100
3.8. Service Credits
If Packet.ai fails to meet the Monthly Uptime Percentage for a given calendar month, Customer may be eligible to receive service credits as follows:
| Monthly Uptime Percentage | Service Credit |
|---|---|
| Less than 99.9% but equal to or greater than 99% | 10% |
| Less than 99% but equal to or greater than 95% | 25% |
| Less than 95% | 50% |
3.9. Service Credit Terms
Service credits:
- Apply only to future invoices for the affected Service;
- Are calculated based on the monthly fees paid for the affected Service;
- In no event shall exceed an aggregate of thirty percent (30%) of the monthly fee payable for the affected Service for the relevant month; and
- Constitute the Customer's sole and exclusive remedy for any failure to meet the uptime commitments under this Agreement.
3.10. Credit Claim Procedure
To receive any service credits, Customer must:
- Submit a written request within thirty (30) days following the end of the calendar month in which the alleged Downtime occurred;
- Submit the request to help@packet.ai; and
- Provide reasonable documentation or evidence supporting the claimed Downtime, including the relevant dates, times, and affected Services.
Failure to comply with this procedure shall result in forfeiture of the applicable service credits.
3.11. Excluded Downtime
Excluded Downtime means any period of unavailability, interruption, or degradation of the Services resulting from or attributable to:
- Scheduled maintenance performed with reasonable prior notice;
- Emergency or unscheduled maintenance required to address security, stability, or operational risks;
- Failures, outages, or performance issues of third‑party infrastructure or upstream service providers, including but not limited to Gcore, Hyperfusion, Tata Communications, Cloudflare, or their respective subcontractors;
- Customer's systems, software, configurations, integrations, or misuse of the Services;
- Internet delays or failures outside Packet.ai's reasonable control;
- Force majeure events or circumstances beyond Packet.ai's reasonable control, including changes of applicable law or regulation.
Excluded Downtime shall not be counted toward Downtime for purposes of calculating the Monthly Uptime Percentage.
