TL:DR
- Legacy network provisioning is too slow and rigid for enterprise AI, creating costly delays every time organizations need to update or expand their digital infrastructure.
- Software-defined interconnection cuts provisioning time from 32 hours to 4 hours per instance, freeing network engineers to focus on higher-value transformation work.
- A Forrester study found switching to software-defined connectivity delivered savings across provisioning, circuit replacement and dynamic bandwidth scaling over a three-year period.
The drawbacks of traditional network provisioning are clear and obvious. Companies have to work with a network carrier to schedule a physical circuit turn-up, which could take weeks. In contrast, deploying software-defined networking solutions is faster and more cost-efficient. However, we’ve never been able to measure exactly how much faster and more cost-efficient. Until now, that is.
Equinix commissioned Forrester Consulting to conduct the Total Economic Impact™ Of Equinix Networking Solutions study.[1] This study is based on interviews with four decision-makers representing real organizations that have recently adopted software-defined networking. It shows the quantifiable impact of switching from physical network infrastructure to software-defined connectivity, including the benefits of faster provisioning.
The interviews indicated that provisioning infrastructure manually took 32 hours of network engineering work per instance. Deploying connections via an automated portal took only 4 hours per instance, an 86% reduction in provisioning time.
All this avoided effort has a real impact on the bottom line. The study found that a composite organization based on the interview responses would save about $360,000 in labor costs over a three-year period, assuming 50 instances provisioned annually.
What happens when network procurement isn’t agile enough?
As my Equinix colleague argued in a recent blog post, the advent of enterprise AI has exposed the limitations of internet-based connectivity and MPLS networks. In particular, MPLS networks are too rigid to keep up with the always-changing AI landscape. When organizations need to spin up new GPUs in hours, they don’t want to spend weeks or even months deploying the MPLS circuits to connect those GPUs.
The respondents in the Forrester study consistently named the same challenges mentioned in that previous blog post: Traditional deployment methods are complex and time-consuming because they require coordinating with providers, procuring the needed hardware, and performing the physical installation. Every time they needed to adjust their network quickly to meet a new requirement—such as connecting with an emerging AI provider—they lacked the agility to do so.
Legacy provisioning is no longer just an inconvenience; it’s now a fixed tax that comes due every time the network needs to change. Simply put, it’s incompatible with enterprise AI.
What are the real benefits of automated network provisioning?
Network engineers are knowledgeable professionals. Their time is valuable, and they should spend as much of that time as possible performing high-value work. Instead of struggling to clear a provisioning backlog that just gets longer, these engineers should redeploy that time toward higher-value infrastructure, network and transformation work that drives real business results.
The study found that automated provisioning saves roughly 28 hours of work for each instance. At the hourly rate that an experienced network engineer could command, an organization would save several thousand dollars every single time they needed to update their network.
Easy access to the portal, requesting services, and provisioning—that was second to none.”AVP of engineering, insurance/financial services
Whenever enterprises need to connect with new partners, they can do so quickly, without the need for a major network rebuild. This agility can help address many of the business challenges they may be facing today:
- They can access cloud providers in specific locations to enable expansion into new markets while also meeting sovereignty requirements.
- They can break through AI supply-chain bottlenecks by connecting to GPU as a Service providers, and access new model providers quickly for optimal choice and flexibility in their AI strategies.
- They can connect to technology partners to enable digital transformation.
The true cost of physical networking
Faster provisioning is just one of the financial benefits mentioned in the Forrester study. The cost savings continued even after the connections were live. That’s because software-defined connections can scale up more cost-effectively than traditional physical infrastructure.
For each MPLS connection replaced with a software-defined connection, the Forrester study placed the average annual savings at $12,000. This isn’t a one-time benefit: The more connections deployed, the bigger the savings. For instance, the composite organization added new network connections from year to year:
- Year 1: 120
- Year 2: 140
- Year 3: 160
They also scaled up the percentage of those connections that were software-defined instead of physical:
- Year 1: 20%
- Year 2: 40%
- Year 3: 60%
Both of these factors compounded the total cost savings. Over the three-year period, the study found that the composite organization cut their total network infrastructure costs by $1.5 million as a result of switching to software-defined networking.
We believe that this figure quantifies what we’ve long known to be true: Physical networks weren’t meant for AI. They were meant for traditional IT workloads that didn’t change often or grow quickly.
In contrast, AI workloads demand agility that physical networks can’t provide. This mismatch between workloads and network now has a specific cost attached to it. That cost will only continue to grow for any organization that doesn’t capitalize on software-defined connectivity.
[Equinix's solution] is easy to use; it's flexible. It helped us facilitate our connection to the clouds.”Operations director, financial services infrastructure
Bandwidth should be an elastic resource, not a fixed cost
Many enterprises are experiencing higher demand for bandwidth as they scale out AI. Once again, software-defined connectivity is the cost-effective way to meet this need.
Traditional networks only scale up in one way: by adding physical infrastructure. We already know this scaling process is slow, but it’s also hard to undo. When an organization deploys enough bandwidth to meet their peak demand, they lock themselves into higher costs going forward, even when a significant chunk of that bandwidth is sitting unused. The Forrester study found that the composite organization’s average unused capacity on each physical connection was 30%.
To address this problem, software-defined networking enables dynamic bandwidth scaling. The organization can simply wait until peak demand hits, scale up to meet that demand, and then scale back down to avoid overpaying. The Forrester study found that the composite organization saved $1.1 million over the three-year period thanks to dynamic bandwidth scaling.
There are times you need extra bandwidth. Stretching the pipelines [and] the ability to scale up and scale down become important during these scenarios.”AVP of engineering, insurance/financial services
One architectural change, three ways to save
This post highlights the different ways that the composite organization saved money:
- Faster provisioning: $360,000 saved
- Replacing MPLS circuits: $1.5 million saved
- Dynamic bandwidth scaling: $1.1 million saved
These are three different benefits that all come from the same underlying shift: Using Equinix Fabric® for software-defined interconnection that can be provisioned on demand and scaled easily.
Software-defined connectivity also provides the foundation for Equinix® Connected Cloud, our neutral digital infrastructure that helps customers connect to the clouds, networks, AI providers and ecosystem partners they depend on.
This same foundation also supports our newly announced solutions:
- Equinix® Fabric One™ can automatically build the network that aligns to your business goals, instead of you re-architecting it manually every time your needs change.
- Equinix® Inference Exchange helps stand up AI inference in days rather than months. It’s pre-connected to the clouds and AI platforms you already depend on.
Both these solutions are currently in preview and scheduled for general availability in early 2027.
Faster networking provisioning isn’t just about productivity
The difference between 32 hours to provision a physical connection and 4 hours to provision a software-defined connection is clear, and it comes with quantifiable cost savings attached. It’s also the first step toward further savings: replacing physical circuits and avoiding idle bandwidth.
More importantly, automated provisioning could mean the difference between a network that’s agile enough to scale and change with your AI initiative and a network that slows you down.
This is where the differentiated value that Equinix delivers can have a major impact. With our unique blend of on-demand connectivity, elastic bandwidth and global reach, our customers can modernize their networks and enable AI-driven growth.
Learn more about the benefits that go along with software-defined networking: Read “The Total Economic Impact™ Of Equinix Networking Solutions” from Forrester Consulting.
[1] The Total Economic Impact™ Of Equinix Networking Solutions, A Forrester Total Economic Impact study commissioned by Equinix, September 2026.
Subscribe to the Equinix Blog