A credible argument names its own limits. The New Enterprise model does not fit everything, and pretending it did would undermine everywhere it does fit.
There is a tier of organization that genuinely needs the full traditional stack, and it deserves to be named precisely. These are not merely “big companies.” They are companies that have effectively become their own carrier or internet service provider:
- Companies that operate as their own ISP. The clearest example is a global parks-studios-and-resorts operator that functions as a carrier for all of its own properties and business units. When a single entity is provisioning connectivity for tens of thousands of users and devices across a private domain it controls end to end, it is running a carrier-grade network, and it needs carrier-grade infrastructure.
- Data-center and hyperscaler fabrics. The networks inside a major data center, where massive east-west traffic between machines is the entire point, require exactly the high-capacity switching fabric the traditional core was built for.
- Latency-critical local compute. Manufacturing floors, trading systems, real-time industrial control, large hospital systems with on-site imaging and monitoring, environments where microseconds of local routing matter and the compute must stay on-site.
- Compliance and air-gap islands. Some regulated or secure environments require local infrastructure by mandate, regardless of where the rest of the industry is going.
For these organizations, the capacity, feature depth, and engineering of traditional enterprise and carrier networking are not overkill. They are the requirement. Nothing in this paper argues otherwise. The skill of designing and operating those networks is real and hard, and the vendors who do it well earn their place.
The argument of this paper is narrower and, we think, more useful: that tier is perhaps five percent of businesses. The other ninety-five percent have been sold, and have kept buying, an architecture built for a world they no longer live in. They asked for “enterprise” because they wanted reliability and scalability, and they were handed feature depth they will never use, priced like the carrier networks it was borrowed from. Most medium and large business networks have quietly crossed into the New Enterprise without their architecture catching up. They are still being designed and sold as if the building were full and the servers were down the hall. They are not. This paper is for them.
From the Field: from one project to the modern enterprise, 2008 to 2026.
“It started in 2008 with one environmental research company and a quarter-million-dollar quote I replaced for twenty-five thousand. Then healthcare in 2014, where the old enterprise could not do the job at any price. Then full enterprise networks, a multi-state health system, a construction company where we made the router the core and the outages stopped. Along the way we became one of the largest mobile-healthcare connectivity providers in the country, hundreds of dental and mammography and imaging units built on the same idea that started with two routers and a wind-farm study.
It is 2026 now, and we serve some of the largest enterprises in the world, organizations looking for mission-critical connectivity that is scalable, affordable, reliable, and accountable. The thing I want to be clear about is that this was never really a story about one vendor. It is a story about a dynamic shift in how the modern enterprise is being built. The companies that see it first will spend less and break less than the ones still buying the network their predecessors built.”
To say it once more, plainly: the argument is not that high-capacity switching fabric is obsolete. It belongs in the environments that still generate massive east-west machine traffic or operate as their own carrier. The error is architectural displacement — taking a design optimized for those environments and pushing it into the far larger set of sites whose workloads left for the cloud years ago.