BAND · Research
Carriers Took 60% of Bandwidth's Growth. The 12% Business Left Over Just Started Paying Its Owners
Bandwidth's 22% headline growth is mostly carrier surcharges passed through at cost. The owned network grew 11.7%, sales spending barely moved, and owner cash after stock pay swung from negative $13.0M to positive $17.0M in nine months.
- Share of revenue increase from carrier surcharges
- 60.5%
- Cloud communications revenue growth
- 11.7%
- Owner cash
- $34.2M
1. Summary: what this piece answers
Bandwidth printed record third-quarter revenue of $234.0M, up 22.0%, and raised the full-year 2026 revenue guide to $905M to $910M [1][2]. Read only the headline and it looks like a communications company that found a second wind.
This piece asks a narrower question: how much of that growth belongs to Bandwidth's owners?
Here is what you will learn. Of the $42.1M increase in quarterly revenue, $25.5M, or 60.5%, was messaging surcharges: fees that mobile carriers impose on business texts, which Bandwidth bills to customers and hands straight back to the carriers at cost [1].
The business Bandwidth actually owns, cloud communications, grew 11.7% to $158.5M, our calculation [1]. That is the number the market should be valuing.
And on that smaller, slower line something important changed in 2026: sales and marketing grew 3.5% while cloud revenue grew 11.7%, non-GAAP gross profit rose 17.6%, and owner cash, meaning operating cash flow less all capex less stock pay, swung from negative $13.0M in the first nine months of 2025 to positive $17.0M in the first nine months of 2026, our calculation [1][3].
Our view: the surcharge line makes Bandwidth look faster and less profitable than it is. Strip it out and you find a 12% grower with a real owned network that is starting, for the first sustained stretch, to turn usage into cash for owners.
2. Thesis and the toll-gate test
Superintelligence agents that answer phones and send texts still need a phone number, a carrier interconnect, emergency routing and someone who carries the regulatory liability. Bandwidth runs an owned-and-operated network covering about 70 countries and 90% of global GDP, and sells it through APIs to the cloud contact center and unified communications platforms, hyperscalers and Global 2000 enterprises [1][2].
The question is whether owning that road lets it keep a toll, or whether carriers and platforms take most of it.
| Test | Result | Evidence |
|---|---|---|
| Owns the record | Partial | Bandwidth holds phone number inventory, 911-enabled numbers and porting relationships; one customer relies on it for millions of phone numbers [2]. It does not own the conversation data or the customer relationship, which sit with the contact center platforms it serves [1] |
| Owns the connection | Pass | Owner-operated network across about 70 countries; customers cite its low-latency owned network and 5x-redundant toll-free service; Salesforce named it the first certified international carrier for its native Contact Center [1][2] |
| Prices close to outcomes | Partial | 75% of cloud communications revenue comes from per-minute and per-message usage, so revenue scales with agent traffic rather than seats [1]. But it bills the minute, not the resolved call, and most customers have no long-term commitment [1] |
| R&D on the gate vs features | Pass | R&D of $112.3M over nine months, 17.0% of total revenue and 24.4% of cloud communications revenue, our calculation, funds the Universal Platform and Maestro orchestration on its own network [1] |
Two passes and two partials. The partials define the thesis: Bandwidth owns the pipe and is paid by the drop, but the carriers at one end and the platforms at the other decide how much of each drop it keeps.
3. What Bandwidth sells
Cloud communications is the business. It has three main product lines [1].
Global Voice Plans sell calling, phone numbers and 911 service, mostly to software platforms that embed calling in their products; this line grew 8% in the quarter on higher voice traffic. Programmable Messaging sells text and multimedia messaging through APIs and grew 26%, helped by about $3M of political campaign texting ahead of the November 2026 midterms that Bandwidth does not expect to recur in 2027. Enterprise Voice sells calling directly to large companies, increasingly through Maestro, and grew 15% [1].
Maestro is the product to watch. It is a software orchestration layer that sits on Bandwidth's network and lets an enterprise route calls among contact center vendors, carriers and third-party voice agents, with observability built in [1].
The 10-Q says customers increasingly chose Maestro for its software-driven orchestration, and the risk factors confirm that integrating third-party agent technology is a prominent feature of it [1]. In plain terms, Maestro lets a bank plug a voice agent into its contact center without rebuilding its phone system, and Bandwidth earns both a platform fee and the minutes.
The quarter's customer wins show the pattern [2]. An all-digital bank chose Bandwidth to modernize its contact center and add a voice agent, displacing a major rival.
A consumer privacy company launched an agent-powered call screening service on Bandwidth's voice infrastructure. A communications provider serving state and local governments adopted the platform voice product, adding recurring platform fees on top of usage.
4. How each line makes money, customers and pricing
Pricing is mostly usage. Bandwidth charges per minute of voice and per message, plus monthly recurring charges for phone numbers, 911-enabled numbers and software services [1].
In the third quarter 75% of cloud communications revenue came from those usage sources, up from 74% a year earlier [1]. Unbilled revenue was 60% of receivables, which tells you how much of the business is metered and billed after the fact [1].
Surcharges work differently. Carriers impose fees on business messages; Bandwidth invoices them to customers and passes them through [1].
They rose 51% on higher messaging activity and higher carrier fees [1]. They count as revenue and as cost of revenue, so they add volume and no profit.
They are why total gross margin fell to 36% from 38% even as gross profit rose 14.9% [1][2]. Management's own non-GAAP gross margin divides by cloud communications revenue only, and on that basis it rose to 61% from 58%
5. Growth
| Measure | Value | Change |
|---|---|---|
| Q3 2026 total revenue | $234.0M | +22.0% [1] |
| Q3 2026 cloud communications revenue | $158.5M | +11.7% [1] |
| Nine-month total revenue | $662.7M | +21.3% [1] |
| Trailing twelve-month total revenue | $870.4M | our calculation [1][3] |
| Fiscal 2025 total revenue | $753.8M | +0.7% [3] |
| Full-year 2026 revenue guide | $905M to $910M | company guidance [2] |
| 2027 outlook | double-digit cloud communications growth | company outlook, from voice usage demand [2] |
The growth story has two speeds. Total revenue went from 0.7% growth in fiscal 2025 to 21.3% in the first nine months of 2026, but most of that acceleration is surcharges [1][3].
Cloud communications growth of 12.0% over nine months is the steady line. The 2027 outlook matters because it names voice, not messaging, as the driver, which is where Maestro and agent traffic show up and where the political texting tailwind does not [2].
6. Retention and the proxy we use
Bandwidth discloses a net retention rate: revenue from last year's customer cohort this year, divided by what the same cohort paid a year earlier, averaged over four quarters [1]. It was 113% in the third quarter of 2026, up from 105% a year earlier [1].
The 10-Q attributes the increase primarily to higher commercial messaging activity [1]. Because the figure is measured on total revenue, surcharges inflate it too, so we treat 113% as an upper bound.
The cleaner proxy is cloud communications growth of 11.7% over the same quarter, which includes some new-customer revenue and so overstates retention on the owned line slightly.
7. Operations: costs, leverage and efficiency
The operating story is selling efficiency. Over nine months sales and marketing actually fell slightly, to $75.6M from $76.0M, while cloud communications revenue rose $49.4M [1].
Growth is coming from existing customers sending more traffic, which is what a 113% retention rate and a usage model should produce. Research and development grew faster than revenue on the owned line in the nine months, up 17.7%, so the company is spending its savings on the platform rather than on salespeople [1].
GAAP operating income understates the change. Amortization of acquired technology in cost of revenue rose to $5.0M in the quarter from $2.1M because of the shortened Voxbone life, which added $2.9M of expense in the quarter and $8.9M over nine months [1][2].
Without it, nine-month operating income would have been about $1.5M instead of a $7.4M loss, our calculation [1]. That charge is non-cash and runs off once the two-year life ends, around early 2028 on our reading of the change [1].
Total depreciation and amortization was $52.6M over nine months, up 32.7%
8. Cash flow and owner cash
| Line | Nine months 2026 | Nine months 2025 | Trailing twelve months | Fiscal 2025 |
|---|---|---|---|---|
| Operating cash flow | $75.2M | $50.9M | $113.9M | $89.5M |
| Less property and equipment | $12.4M | $17.2M | $17.5M | $22.3M |
| Less capitalized software | $6.7M | $8.2M | $9.2M | $10.7M |
| Less stock pay | $39.1M | $38.4M | $53.0M | $52.3M |
| Owner cash | $17.0M | -$13.0M | $34.2M | $4.2M |
Sources: [1][3], our calculation. Owner cash improved by $30.0M year over year in the first nine months.
Three things drove it: operating cash flow rose $24.4M, capex fell $6.3M, and stock pay was nearly flat [1]. The longer record is uneven: owner cash was negative $17.9M in fiscal 2023, positive $10.1M in fiscal 2024 and positive $4.2M in fiscal 2025, our calculation [3].
We checked the quality of the improvement. Deferred revenue fell over the nine months, so customer prepayments did not inflate operating cash [1].
Net interest expense was only $1.8M, so there is no interest income on corporate cash to strip out [1]. Working capital was mixed: receivables absorbed $27.1M as usage billing grew, while accrued liabilities added $21.0M, partly surcharges owed to carriers [1].
Those roughly offset, which leaves owner cash driven by the business rather than by timing. Stock pay remains the largest claim: over the trailing twelve months it took 46.5% of operating cash flow, our calculation
9. Balance sheet and capital allocation
Bandwidth rebuilt its debt this year [1]. On Jun 18 it issued $316.3M of 0% convertible notes due Jul 1, 2032, netting about $281.5M after capped calls and costs.
It used the cash to repurchase about $243M of principal on its 2028 notes for about $229M, booking a $13.2M gain, and settled the last $8M of 2026 notes at maturity in March. About $6.7M of 2028 notes remain.
On Oct 26 it raised the revolver to $200.0M and extended it to Oct 2031.
The result is a longer maturity and no cash coupon, at the cost of a larger convertible balance and potential dilution: the third-quarter diluted share count included 4.4M shares from convertible debt [2]. Convertible principal of about $323M less cash of $171.3M leaves net debt of about $152M, our calculation, excluding marketable securities.
The board also authorized an $80M buyback in February; Bandwidth bought back $30.0M in nine months and has about $50M left [1]. Because total convertible debt rose while the buyback ran, we treat the buyback as funded partly with borrowed money, which is leverage rather than a return of surplus cash.
10. Acquisitions in the last 24 months
None. The cash flow statement shows no acquisitions in 2025 or 2026, only $0.4M of proceeds from a small business sale in 2025 [1].
The most recent deal, Voxbone in 2020, matters today only through the accelerated amortization described above [1].
11. Superintelligence angle
Tailwind, with one reason: every voice agent that calls a customer, and every customer who calls a voice agent, generates billable minutes on a regulated network that agent builders do not want to own. Management frames the opportunity the same way, and Enterprise Voice growth of 15% plus the Maestro wins suggest it has started [1][2].
The threat is price. If platforms commoditize the minute or route around Bandwidth with their own carrier deals, usage rises and margin does not; the 10-Q notes that net retention falls when Bandwidth lowers prices [1].
Agents raise the volume of conversations. Whether Bandwidth keeps the toll depends on Maestro making it the routing layer, not just the pipe.
12. Leadership and control
David Morken co-founded Bandwidth in 1999 and serves as chief executive and chairman [1][4]. Daryl Raiford is chief financial officer [2][4].
Bandwidth has two classes; Class B carries ten votes. Morken and co-founder Henry Kaestner hold substantially all Class B shares and control about 40% of combined voting power [1]; the proxy lists Morken at 23.2% and Kaestner at 17.3% of voting power as of Mar 15, 2026 [4].
BlackRock held 6.3% of Class A shares, per its Schedule 13G [4]. There is no lead independent director, and the chief executive also chairs the board [4].
The style is founder-led and patient: the company spent years building its own network rather than reselling carriers, and its 2026 capital moves favored long-dated, zero-coupon funding.
13. Risks and catalysts
Risks: surcharge growth that adds revenue without profit and flatters retention [1]; about $3M a quarter of political texting that does not recur in 2027 [1]; concentration in a limited number of large customers who can cut usage without penalty [1]; price competition on minutes and messages [1]; a lawsuit alleging failure to bill and remit 911 taxes and surcharges in Illinois [1]; dilution from convertible notes and stock pay [1][2].
Catalysts: Needham Tech Week investor meetings on Nov 18, 2026 (confirmed) and a Wells Fargo TMT Summit fireside chat on Dec 8, 2026 (confirmed) [2]; the November 2026 midterm elections, after which political messaging rolls off (confirmed date) [1]; fourth-quarter 2026 results and formal 2027 guidance, expected late February 2027 (unconfirmed); the end of the accelerated Voxbone amortization around early 2028 (unconfirmed timing, our reading) [1].
14. Scorecard, questions and predictions
| Metric | Value |
|---|---|
| Verdict | Watch. Anchor: trailing owner cash of $34.2M; no base value set until three positive full years |
| R&D share of revenue | 17.0% of total, 24.4% of cloud communications, nine months [1] |
| GAAP net margin | 1.8% over nine months; about -0.2% excluding the $13.2M debt gain, our calculation [1] |
| API quality | 4 of 5: documented public REST APIs with keys, public pricing, SDKs and a changelog, and a large partner ecosystem; a dedicated agent tool server is not confirmed in the filings, our scoring |
| Net retention rate | 113%, Q3 2026, includes surcharges [1] |
| Toll-gate score | 2 of 4 pass, 2 partial |
Questions for the owner. First, should we value Bandwidth on cloud communications revenue alone and treat surcharges as a pass-through with zero value?
Second, how much of the owner cash swing survives once political texting rolls off and receivables stop growing? Third, will Maestro earn platform fees large enough to move pricing from the minute toward the outcome?
Fourth, conviction on a 1 to 5 scale: our view is 3, the network and operating leverage are real but owner cash has only two positive full years, so we watch with a trigger of fiscal 2026 owner cash above $30M and 2027 cloud communications growth at double digits.
Predictions to register: fourth-quarter 2026 cloud communications revenue grows 10% or more (60%); net retention stays at or above 110% at the next print (65%); fiscal 2026 owner cash finishes positive (80%) and above $25M (60%); messaging surcharges again exceed half of the total revenue increase in the fourth quarter (65%).
Stock Profiles & Comparable Companies
BAND Stock Profile →More Published Research
View All →Tenable Maps Every Weakness Agents Will Hunt. It Bought Back $516M of Stock on $81M of Owner Cash, and Borrowed the Difference
Tenable owns the exposure record and owner cash is finally positive, but growth is slowing toward 7%, the share count shrank 11.5% on convertible debt, and stockholders' equity turned negative.
SentinelOne Guards the Device Where Agents Run. Owners Lost $291M Last Year, and the Gap Is Widening
Revenue grows near 20% and the products target agent security, but stock pay was 5.6 times operating cash flow, owner cash worsened to -$291.0M, and a new convertible could add 46.1M shares.