Artificial intelligence news.
Straight from the original announcements.

Forrester Study: Private Connectivity Cut Public Internet Costs by 77%

October 20, 2026, 13:55Source: Equinix

A Forrester Consulting study commissioned by Equinix found that a composite organization saved $174,000 over three years by moving enterprise traffic off the public internet. According to the study, transferring 1 GB of data cost $0.085 on the public internet and $0.02 over private connectivity, a 77% reduction.

Key points
  • A Forrester study commissioned by Equinix found a composite organization saved $174,000 over three years by avoiding the public internet.
  • Per the study, 1 GB cost $0.085 over the public internet and $0.02 over private connectivity, a 77% reduction.
  • Interviewees also named agility, security, resilience and performance benefits that were not quantified.
  • Equinix Fabric One and Equinix Inference Exchange are in preview, with general availability scheduled for Q1 2027.

What the study measured

The Total Economic Impact™ of Equinix Networking Solutions study was conducted by Forrester Consulting and commissioned by Equinix. It is based on interviews with networking leaders and builds a composite organization that moves from traditional networking to private, software-defined interconnection using Equinix Fabric.

According to the study, the composite organization saved $174,000 over a three-year period by avoiding network transit costs on the public internet. These costs include cloud egress charges and additional operational overhead. Equinix says the figure is risk-adjusted to provide a conservative estimate.

The per-gigabyte comparison is the core of the finding: $0.085 for 1 GB over the public internet against $0.02 over private connectivity, which the study describes as a 77% cost reduction.

Why Equinix says the internet falls short

Equinix argues that enterprise AI applications are moving from proof of concept to production, and that scaling them requires fast access to cloud providers, AI platforms and ecosystem partners, along with dependable network performance. In its view the public internet cannot deliver either.

The company lists three limitations:

  • It lacks performance SLAs, which makes a good user experience hard to ensure.
  • It uses unpredictable routing, which leads to higher latency and potential compliance issues.
  • It is public, so sensitive data in transit cannot be protected.

Equinix says many enterprises stay on the public internet because they see it as the quickest and easiest route to cloud providers and partners, while private connectivity can look complex and expensive. The study, according to Equinix, shows the reverse: sticking with the status quo costs significantly more.

Benefits beyond direct savings

The blog ties the findings to the idea of a "coordination tax", the unexpected costs that arise when enterprises take a cloud-first approach to AI infrastructure and overlook the connectivity needed to make it work. The study gives a dollar figure for it, but interviewees also named benefits that could not be quantified: network agility, performance, security, resilience and flexibility.

  • Direct routing and observability on private networks help reduce compliance and governance risk.
  • Network agility speeds access to ecosystem partners and new services.
  • Redundant connectivity and simplified failover help maintain service continuity during outages.
  • More predictable performance enables lower latency and a better end-user experience, which Equinix links to higher customer satisfaction and retention.

Equinix concludes that the $174,000 figure is "almost certainly understating the true value" of avoiding the public internet, in its own words.

Financial services and new products

Several interviewees came from financial services. Equinix says institutions on the public internet cannot control how quickly transactions are processed or where sensitive traffic is routed. It cites Dow Jones Newswires, whose high-frequency trading customers needed insights within microseconds. The company used Equinix Fabric for private connectivity and deployed inside Equinix colocation data centers, in New York and London, and can now replicate the model in other financial hubs.

A vice president of core network service at a financial services firm said in the study: "We've shaved off tens, and in some cases, hundreds of milliseconds on every transaction when using [Equinix] services, including Equinix Fabric, for our connectivity."

Equinix also pointed to two recently announced solutions, both in preview and scheduled for general availability in Q1 2027. Equinix Fabric One enables intent-based, any-to-any connectivity without time-consuming reconfiguration. Equinix Inference Exchange lets customers deploy AI inference infrastructure from leading providers with preconfigured connectivity to clouds and AI platforms.

What it means for enterprises

The findings come from a composite organization and a vendor-commissioned study, so the figures illustrate a modeled case rather than a guaranteed result for any one company. For enterprises weighing AI deployments, the study frames connectivity as a cost and performance question, not only a technical one. Equinix recommends private, software-defined connectivity through Equinix Fabric and points readers to the full Forrester report for details.

Related news

Databricks beta streams Unity Catalog data directly to GPUs
For developersNovember 9, 2026, 16:00

Databricks beta streams Unity Catalog data directly to GPUs

Databricks has released UCDataset in beta on AI Runtime, a native PyTorch data loader that streams Unity Catalog tables directly to GPUs. The company says it can cut…

Equinix Opens Singapore Data Center for AI and High-Performance Computing
Chips and infrastructureNovember 6, 2026, 09:32

Equinix Opens Singapore Data Center for AI and High-Performance Computing

Equinix has opened SG6, its sixth International Business Exchange data center in Singapore, expanding capacity for artificial intelligence, high-performance computing…

Products and tools
Products and toolsOctober 29, 2026, 23:22

Equinix Inference Exchange, Built With NVIDIA and Together AI, Due in Q1 2027

Equinix will present Equinix Inference Exchange, a service built with NVIDIA and Together AI, at booth 206 during NVIDIA GTC in Washington, D.C., from November 30 to…