The megabit is not the product

The commodity test

Here is a simple test for whether something is a commodity: can you swap the supplier in a quarter without changing how your business works? Electricity passes. Diesel passes. A gigabit of IP transit passes. A managed enterprise WAN, in most organisations, does not, and that failure is deliberate. The technical layer is interchangeable; the commercial layer is engineered to be sticky.

We buy connectivity for eight of our own domains across Amsterdam, Antwerp and Southeast Asia. The wholesale price of moving a bit between two well-connected cities is now close to a rounding error. What clients pay is almost never about the bit. It is about the wrapper: the portal, the SLA language, the account manager, the change process, and the quiet assumption that nobody will re-tender because re-tendering is painful.

This is not a new trick. Telecom incumbents have run the same playbook since the days of leased E1 lines: bundle the scarce thing (the last mile) with the abundant thing (capacity) and price the bundle as if both were scarce. What changed is the size of the gap. In 1999 that markup was defensible because long-haul capacity itself was expensive and finite. In 2026 it is not, and the invoice has simply not caught up with the physics.

Where the money actually goes

Break a typical multi-site WAN invoice into three parts and the picture clears up fast. Roughly a tenth of it is physical access, the local loop, the fibre in the ground, the only part that is genuinely scarce. Another slice is capacity, which is now near-commodity. The rest is management: configuration, monitoring, ticket handling, and the risk premium a carrier charges for promising uptime it does not control end to end.

• Access (last mile): genuinely scarce, genuinely worth paying for. Negotiate it locally, per site.
• Capacity: a commodity. Buy it short, buy it twice, from two providers who do not share a duct.
• Management: this is software. Price it like software, not like a headcount.
• Risk: an SLA that refunds 5% of a monthly fee for four hours of downtime is not insurance. It is theatre.

The reason SD-WAN mattered was never the acronym. It was that it decoupled those four lines from each other. Once the intelligence sits in software you control, the transport underneath becomes genuinely swappable, and a swappable supplier is a supplier with normal margins.

We once audited a client's 40-site WAN and found the same MPLS circuit priced three different ways across three regional contracts signed by three different procurement teams, none of whom had spoken to each other. The spread between the cheapest and most expensive per-Mbps price for functionally identical service was 4.7x. Nobody had lied. Nobody had even negotiated badly by the standards of their own region. The contracts had simply never been compared, because nobody owned the comparison.

What a commodity-first WAN looks like

In the redesigns we have run, the pattern is consistent. Two dumb, cheap, diverse circuits per site beat one expensive managed circuit on both availability and cost. Diversity does the work that the SLA promised to do. A dual-carrier site with automated failover typically lands in the four-nines range on measured availability, while costing 30 to 50% less than the single premium link it replaced.

The catch is organisational, not technical. Someone inside the company has to own the policy layer, routing intent, security posture, application priority, because that is the part that is genuinely yours. Outsource the fibre, never the intent.

The re-tender playbook, in order

Most WAN re-tenders fail not because the alternative pricing is unavailable but because the incumbent's exit terms were never checked at signing. Do the boring work first: read the termination clause before you shop for a replacement, not after you find one you like. We have watched a client discover a 14-month notice period two weeks before a board meeting where they had promised savings by Q1.

• Inventory every circuit, its carrier, its true capacity and its all-in monthly cost, including the ones procurement forgot were still billing.
• Separate the four cost lines (access, capacity, management, risk) on every invoice, even if the vendor bundles them.
• Run a shadow tender with two carriers per site before the incumbent's renewal date, not after.
• Negotiate exit terms as hard as entry pricing. A 90-day out is worth more than a 5% discount.

None of this requires new technology. It requires someone willing to spend three weeks reading contracts that were written to discourage exactly that. The return on those three weeks, in our own portfolio, has consistently outperformed anything we have done to the application layer in the same period.

The uncomfortable part for buyers

Procurement departments are optimised to reduce the number of suppliers. Commodity thinking pushes the other way: more suppliers, shorter contracts, more substitution. That is a harder spreadsheet to defend internally, right up to the first outage where the second carrier quietly carries the day.

There is also a career-risk dimension nobody puts in the business case. The person who signs a 36-month contract with a name-brand carrier is rarely blamed if the price is high, because 'nobody got fired for buying the incumbent'. The person who signs two 12-month contracts with regional carriers takes on visible personal risk if either one has a bad quarter, even though the aggregate risk to the business is lower. Fixing enterprise WAN economics is partly a fix to that incentive, not just to the invoice.