Skip to content

VPC Interconnect

Your existing cloud, just with affordable GPUs.

iFrame instances peer into your AWS, Azure, or GCP VPC like any other subnet. Same security groups. Same IAM. Same observability stack. Your platform team doesn't manage a separate cloud — they get a new region with more affordable GPUs.

Talk to salesRead the architecture

Why it matters

The thing that's actually expensive about a new GPU vendor isn't price.

It's the operational cost: separate IAM, separate security review, separate observability, separate procurement, separate audit. VPC interconnect removes all of those.

Same security model.

iFrame instances appear in your VPC under your existing security groups. IAM roles work as expected. Service-to-service auth uses your existing identity provider.

Same observability.

CloudWatch, Azure Monitor, GCP Cloud Logging — your existing stack picks up iFrame instances natively. No new dashboards, no new alerting.

Same procurement.

Workloads run in your accounts under your contracts. Egress, billing reconciliation, and audit follow your existing processes.

VPC interconnect

Your existing cloud — just with affordable GPUs.

iFrame instances peer into your VPC like any other subnet. Same security groups. Same IAM roles. Same audit logs. Same observability stack. Your platform team doesn't have to learn a new cloud — they just have a new region with more affordable GPUs.

  • AWSDirect Connect + VPC peeringGA
  • AzureExpressRoute + VNet peeringGA
  • GCPCloud Interconnect + VPCGA
  • OracleFastConnectBeta
See the architectureTalk to sales
CUSTOMER VPC · AWS / AZURE / GCPapp-subnet10.0.1.0/24 · ECS · Lambda · App servicesdata-subnet10.0.2.0/24 · S3 · RDS · Object storageIAM · security groupsInherited unchanged across the peeringPEERINGDirect ConnectIFRAME · US-WESTgpu-subnetB200 · B300 · H200 · H100bare-metal-subnetDedicated nodes · NVLink · Reservedaudit · observabilityCloudWatch · Azure Monitor · GCP Logging

Diagram: customer VPC on the left (subnets, security groups, IAM) peers via the supported interconnect product to iFrame bare-metal and GPU subnets on the right. Audit and observability flow across.

Per-hyperscaler integration

Production on the three majors. Beta on Oracle.#

HyperscalerIntegrationStatus
AWSDirect Connect + VPC peeringGA
AzureExpressRoute + VNet peeringGA
GCPCloud Interconnect + VPCGA
OracleFastConnectBeta

Setup process

Two to four weeks from discovery to production.#

Joint configuration with your platform team. We handle the iFrame side; you keep your existing security review.

  1. 01

    Discovery call.

    We map your existing topology and the workloads you want to run on iFrame.

  2. 02

    Network configuration.

    Joint setup of the interconnect, peering, and security policy.

  3. 03

    Validation.

    A pilot workload runs end-to-end to verify networking, observability, and audit.

  4. 04

    Production.

    Capacity scales to contracted levels. Named support engineer attached to the account.

Reference architectures

Three patterns we ship most often.#

Burst-to-iFrame.

Steady state runs on hyperscalers; spikes burst to iFrame for cost. Same observability picks up the burst capacity natively.

Hybrid training.

Training runs on iFrame; checkpoints and models stored in your existing S3 / Blob / GCS object storage.

Inference-only.

Application logic stays on the hyperscaler; inference traffic peers to iFrame's managed endpoint at one-third the per-token cost.

Compliance

Inherit your posture. Add ours.#

VPC interconnect inherits your existing compliance posture — the workloads run inside your accounts, under your contracts, against your audit logs. iFrame's own compliance certifications cover the underlying platform.

Trust center

FAQ

Networking and security questions.#

Map your topology with our team.

A 30-minute call clarifies whether VPC interconnect fits your stack. No deck. Architecture diagrams welcome.

Talk to salesRead the trust center