Table of Contents
Bandwidth grew revenue 22%, adjusted EBITDA grew 27%, first-half free cash flow nearly doubled, and non-political NRR was a very healthy 113%. Even after stripping away the surcharge optics (see below), Bandwidth continues to become a more profitable business.

Bandwidth is still fundamentally a voice company. Global Voice accounts for 71% of Cloud Communications revenue, Enterprise Voice another 9%, and Messaging 20%. But software is growing much faster, customers are ramping beyond their original contracts, and AI is starting to change both how Bandwidth sells and what it puts inside its network.
The Good
Bandwidth’s revenue increased by roughly $40M from a year ago. Nearly $24M of that came from higher messaging surcharges, which jumped from $44.2M to $68M. Put differently, about 60% of the YoY revenue growth came from higher carrier surcharges that come in one door and go out the other.
That is not all that different from its peers, where carrier fees can similarly make the reported top line look better than the economics underneath it.
The more interesting growth might happen after the contract gets signed. Of the six $1M+ customers Bandwidth signed in 2025, two are now expected to exit 2026 at nearly twice their original estimated contract value.
That is one advantage of a consumption-based model: Revenue can overshoot bookings as customers ramp.
Bandwidth essentially refinanced its house. It paid off debt coming due in 2028 as well as its $50M revolver, replacing them with $316M of 0% convertibles, i.e., no interest until 2032. The catch is potential equity dilution, though Bandwidth bought protection against much of it until the share price roughly doubles. That’s a pretty good refi.
The Interesting
Bandwidth’s Build product lets AI agents sign up, provision services, and begin generating traffic. Like Twilio and Sinch, Bandwidth is courting AI developers, but as David put it, “Instead of API for developers, we’re living through CLI for agents.”
Bandwidth has also begun putting GPUs into its points of presence. It’s just cool to see pieces of the AI compute stack moving directly into a telecom network.
Voice AI is showing up in customer ramps too. Customers are moving from prototypes into production, and they have additional Voice AI workloads planned for later in their contracts. It looks increasingly similar to what Twilio reported this quarter: Voice gets the workload in the door, and then more usage and software can follow.
Salesforce may add another leg. Bandwidth is already carrying traffic for Agentforce Contact Center, and David said some opportunities developing through that relationship could be larger than its typical enterprise deals.
The Unknown
We still do not know how much of the 66% growth in attached software services is AI, how large Build is, or what Salesforce may eventually contribute.
There is enough evidence to say AI is moving beyond experimentation. There is not enough to model its financial impact separately, however.
And RCS: It only had a one-liner that said a major client was expecting to ramp up on RCS in 2027. Across the industry, everyone’s asking RCS: Where is the money?
Finally
One good thing about having a public blog is that it’s hard to hide from what I’ve said before. And I’ve written enough about Bandwidth. I was a reference client for its IPO. I told the bankers then, as I believe now, that Bandwidth was Twilio before Twilio was Twilio. Back then, Bandwidth was the only one with an API to do anything for CPaaS (number provisioning). But somewhere along the way, it lost the narrative. That is changing.
Bandwidth is finding a groove that is uniquely its own. From its focus on building a high-quality voice network to pouring $100M into its Raleigh office, Bandwidth is making the same bet: If you build it, they will come.