Your MPLS costs $2,000 per site per month — SD-WAN does the same job over a $35 fiber line

In July 2019, the MEF (Metro Ethernet Forum) published the MEF 70 standard, defining for the first time what an SD-WAN actually is. In November 2023, Broadcom acquired VMware for $61 billion, picking up VeloCloud — the market-leading SD-WAN solution — in the process. In 2025, the open-source project FlexiWAN crossed 4,000 accounts, proving a per-Mbps licensing model is not the only way. These three dates tell the same story: MPLS is dead, it just has not received the memo yet.

The principle behind SD-WAN is simple to explain and brutal in its budget implications. Where MPLS locks you into a single contract with a single carrier for guaranteed bandwidth at a premium price, SD-WAN takes any internet connection — fiber, 4G/5G, Starlink, DSL if you are nostalgic — and turns it into a segment of a unified private network, managed from a single dashboard. The result: you replace a $1,800 MPLS link with two $40 fiber lines, gaining redundancy and agility that MPLS cannot offer at any price.

MPLS was great in 2005

MPLS (Multiprotocol Label Switching) dominated enterprise WANs for two decades for a good reason: it guarantees throughput, latency, and quality of service that the public internet could not match. A carrier sells you a dedicated line between your headquarters and a branch office, with a contractual SLA — 99.95% availability, jitter under 5 ms, guaranteed repair within 4 hours. In exchange, you pay whatever the carrier asks.

The numbers are known to every CIO but rarely written down. A 10 Mbps MPLS link between two sites in the US costs between $500 and $1,200 per month depending on the carrier and distance. A 100 Mbps link climbs to $1,500–3,000 per month. For a network of 50 sites, the annual bill easily crosses $1 million — before you have bought a single router.

And this model has three structural flaws that SD-WAN attacks head-on.

First, rigidity. Switching carriers on an MPLS link takes between 30 and 90 days — time to terminate, order a new circuit, test it, and migrate traffic. SD-WAN adds a new connection in a few hours: you plug a 5G modem into the SD-WAN router, the control plane discovers it automatically, and traffic fails over.

Second, backhaul cost. With MPLS, all internet traffic from a branch office routes back to the central datacenter before reaching the web — the classic hub-and-spoke model. The result: a user in a Seattle branch accessing Office 365 sees their packets travel Seattle → Dallas datacenter → internet → Dallas → Seattle, with doubled latency and wasted bandwidth. SD-WAN performs local breakout: traffic destined for Office 365, Zoom, or Salesforce exits directly through the local fiber connection, never touching the WAN link.

Third, the absence of real redundancy. Adding a second MPLS link for redundancy doubles the bill. With SD-WAN, you combine a $40 fiber line and a $15 4G SIM, and the control plane distributes traffic based on real-time link quality — packet loss, latency, jitter — measured 10 to 20 times per second.

What SD-WAN actually changes

An SD-WAN rests on three pillars that MPLS lacks entirely.

The network overlay. SD-WAN creates an encrypted virtual network (IPsec or WireGuard) on top of your physical internet links, called the underlay. This overlay tunnel is transport-agnostic — Verizon fiber, Comcast coax, T-Mobile 5G, Starlink, it does not matter. The control plane sees every link as an interchangeable “transport” and selects the best path packet by packet. This is the layer that makes SD-WAN multi-carrier and multi-technology.

The centralized control plane. Where an MPLS router is configured via CLI, site by site, an SD-WAN is managed from a single interface — cloud or on-prem. You define policies (“Teams traffic gets priority on the lowest-jitter link; backup replication uses the cheapest link”) and the control plane pushes them to every edge device in seconds. Zero-touch provisioning (ZTP) lets you deploy a new site by plugging a box into power and Ethernet — the router announces itself to the controller, downloads its configuration, and builds the tunnels.

WAN optimization. SD-WAN does not just pick the best path — it fixes the path’s defects in real time. Forward Error Correction (FEC) sends redundant packets to reconstruct losses without waiting for TCP retransmission — critical for VoIP and video conferencing over a DSL link with 1% packet loss. Packet duplication sends voice packets in duplicate over two links simultaneously and keeps whichever arrives first. Dynamic path selection measures jitter, latency, and packet loss 10 to 20 times per second and switches the flow before the user perceives a degradation.

The MEF 70 standard (2019) precisely defines these functions and serves as the reference for comparing SD-WAN solutions. A product that does not perform dynamic policy-based routing and FEC is not an SD-WAN — it is a multi-WAN router with a nice UI.

The killer number: $1,800 vs $70

Reduce the debate to price and there is no debate left.

Take a network of 20 sites. With MPLS, using a 20 Mbps link per site and a 200 Mbps link at headquarters, the monthly bill is roughly:

• 19 MPLS 20 Mbps links × $600 = $11,400 per month
• 1 MPLS 200 Mbps link for headquarters = $2,200 per month
• Total MPLS: $13,600 per month, or $163,200 per year, hardware not included

With SD-WAN, on the same network:

• 20 business fiber lines (Verizon, AT&T, Comcast) at $40–60 per month = $1,200 per month
• 20 4G/5G backup SIMs at $15 per month = $300 per month
• SD-WAN license (Cisco Catalyst SD-WAN, list price): ~$150 per site per month = $3,000 per month
• Total SD-WAN: $4,500 per month, or $54,000 per year

SD-WAN costs 67% less than MPLS, delivers 10 to 100 times more bandwidth (a business fiber line provides 200 Mbps to 1 Gbps, versus 20 Mbps on MPLS), includes 4G redundancy that MPLS does not provide, and can be deployed in hours instead of weeks.

And the gap widens if you use an open-source solution like FlexiWAN — the license fee disappears, leaving only the cost of the connections and the hardware. At that point, MPLS becomes an accounting anomaly.

Four solutions to replace your MPLS

The SD-WAN market has organized into four approaches, from the data center to the homelab.

Cisco Catalyst SD-WAN (formerly Viptela, acquired by Cisco in 2017 for $610 million) is the “full Cisco” solution for large enterprises. The control plane is hosted in Cisco’s cloud or on-prem via vManage, with edge devices in the Catalyst 8000 router family. Cisco pushes tight integration with Umbrella (DNS security), ThousandEyes (internet visibility), and Duo (MFA) to build a full SASE offering. The entry ticket is high — expect $150–300 per site per month in licensing — but native integration between SD-WAN, firewall, and DNS security simplifies life for teams already 100% Cisco. Nestlé migrated 1,700 sites across 185 countries onto this solution.

VMware VeloCloud (now Broadcom VeloCloud since November 2023) is the historical market-share leader, with an architecture that places VMware SD-WAN Gateways in data centers and major cloud providers (AWS, Azure, GCP). These gateways act as aggregation points for SD-WAN traffic and enable deployment without an on-prem controller — only a VeloCloud Edge (physical or virtual) is installed on site. The Broadcom acquisition introduced uncertainty around licensing and support that many enterprises felt as a warning shot.

FlexiWAN is the only open-source SD-WAN on the market, founded in 2019 and available as SaaS or self-hosted. Unlike every other vendor, FlexiWAN charges per instance ($3, $9, or $29 per month depending on tier), not per Mbps — a pricing model that breaks the “the more you consume, the more you pay” logic of the big players. The architecture is modular: the FlexiWAN router can dynamically load third-party applications (firewall, routing, WAN optimization) the way a smartphone loads apps. The project claims Intel validation on Xeon Scalable and a partnership with Google Cloud for a joint offering. German operator O2 Telefonica uses it for its “Business Smart Network” service.

pfSense (and its successor OPNsense) is not an SD-WAN in the MEF 70 sense — it does not do FEC or packet duplication — but covers 80% of what an SMB needs to replace an MPLS link with two fiber connections and automatic failover. pfSense’s multi-WAN supports load balancing, failover, and policy-based routing (Teams over fiber, backup over 4G, web browsing over coax). Configuration is GUI-based, Netgate appliances start at $189, and the pfSense+ license is free for personal use with $129 per year for professional support (TAC Lite). For an SMB with 5 sites, pfSense does the job for under $3,000 in one-time hardware costs, with no recurring subscription — whereas an equivalent Cisco deployment costs $150 × 5 × 12 = $9,000 per year in licensing alone.

The choice depends on your scale, not your budget — SD-WAN is always cheaper.

What SD-WAN does not solve

Adopting SD-WAN does not fix everything. Security is not included by default. SD-WAN encrypts the tunnels (IPsec) but does not inspect the content flowing through them. If you replace MPLS with SD-WAN without adding an NGFW at every edge device, you have replaced a secured private line with a tunnel over the public internet — with no inspection. Cisco and VMware know this and push their SASE bundles (SD-WAN + SWG + CASB + ZTNA) — but that is an additional product, not a function of SD-WAN.

End-to-end QoS is only guaranteed within your overlay. SD-WAN works wonders on what it controls (link selection, DSCP marking, FEC), but once the packet leaves your edge device, it crosses the public internet or a carrier backbone — and there, best-effort takes over. MPLS remains superior on one precise point: deterministic end-to-end jitter, contractually guaranteed by a single carrier. For high-frequency trading or real-time telemedicine, SD-WAN alone is not enough — you either keep a residual MPLS link for that specific traffic, or step up to SD-WAN over private backbone (Aryaka, Cato Networks).

Self-hosting has a human cost. FlexiWAN and pfSense eliminate licensing fees but make you responsible for tuning, updates, and troubleshooting. For a two-person network team already managing 200 sites, a $150 per month Cisco license may be cheaper than the time spent debugging an IPsec tunnel at 3 a.m.

The verdict

If you have an active MPLS contract, run the numbers now. Take your monthly cost per site, compare it to two business fiber lines ($40 × 2 = $80) plus an SD-WAN license ($150 for Cisco, $29 for FlexiWAN), and look at the delta. The only case where MPLS remains defensible in 2026 is if your traffic demands sub-5 ms jitter end-to-end with an enforceable SLA — and that covers neither your ERP, nor your VoIP, nor your backup replication.

If you are an SMB with fewer than 20 sites, do not even look at Cisco or VMware. Go with FlexiWAN (if you have a network team that can read documentation) or pfSense with multi-WAN (if you want point-and-click). The price difference with an MPLS deployment is so large it finances the migration engineer’s salary in under six months.

If you are a large enterprise with 500 sites and a 100% Cisco stack, go with Cisco Catalyst SD-WAN. It is not the cheapest, but the native integration with the rest of the Cisco ecosystem (ISE, Umbrella, ThousandEyes) reduces the learning curve to zero. Your network team already knows the syntax — SD-WAN becomes a feature of your routers, not a new product to learn.

In every case, the question is not “is SD-WAN ready for my network.” The answer has been yes since 2019. The question is: how many more months of MPLS are you going to pay for before you hit the button?

References

• MEF 70 — SD-WAN Service Attributes and Service Framework, MEF, July 2019.
• Gartner Forecast Analysis: SD-WAN, Worldwide, Gartner, 2024.
• Cisco SD-WAN — Solution Overview, Cisco, accessed July 2026.
• VMware SD-WAN by VeloCloud — Documentation, VMware/Broadcom, accessed July 2026.
• flexiWAN — The World’s First Open Source SD-WAN & SASE, flexiWAN, accessed July 2026.
• pfSense Multi-WAN — Documentation, Netgate, accessed July 2026.
• Broadcom Completes Acquisition of VMware, Broadcom, November 22, 2023.
• Cisco Completes Acquisition of Viptela, Cisco, August 1, 2017.
• Nestlé revamps global network with Cisco SD-WAN, Cisco Blog, accessed July 2026.
• O2 Telefonica selects flexiWAN for Business Smart Network, flexiWAN, accessed July 2026.