밴드위스(BAND) 2026년 3분기 실적 발표 콘퍼런스 콜: 매출 22% 증가, 가이던스 상향

핵심 요약

• 밴드위드(NASDAQ: BAND)가 2026년 3분기 매출액 2억 3,400만 달러를 기록하며 전년 동기 대비 22% 증가해 역대 최고치를 달성했다고 발표했다. 클라우드 커뮤니케이션 매출은 1억 5,800만 달러로 12% 증가했다.
• 조정 EBITDA는 43% 증가한 3,500만 달러로 역대 최고치를 기록했으며, 마진율 역시 사상 최고치인 22%를 달성했다. 잉여현금흐름(FCF)은 150% 이상 증가한 3,300만 달러에 달했다.
• 엔터프라이즈 음성 매출은 15% 성장했고, 프로그래머블 메시징 매출은 26% 증가했다. 소프트웨어 서비스 매출은 68% 증가한 700만 달러를 기록했으며, 연간 반복 매출(ARR)은 3,000만 달러를 초과했다.
• 밴드위드는 2026 전체 연도 가이던스 모든 지표의 중간값을 상향 조정했다. 현재 총매출은 9억 500만 달러~9억 1,000만 달러, 조정 EBITDA는 1억 2,400만 달러~1억 2,500만 달러로 예상하고 있다.
• 경영진은 2026년 정치 메시징 매출과의 비교 기저효과에도 불구하고, 2027 회계연도에 클라우드 커뮤니케이션 매출이 두 자릿수 성장을 기록할 것으로 전망하고 있다. 이미 체결된 엔터프라이즈 계약과 기존 고객이 이러한 성장의 대부분을 견인할 것으로 예상된다.
• 밴드위드는 2026년 첫 9개월 동안 건당 100만 달러가 넘는 계약을 11건 체결했으며, 이는 연간 계약 가치(ACV) 기준 약 1,900만 달러에 달한다. 이 매출의 대부분은 아직 보고된 실적에 반영되지 않았다.

주요 재무 실적

밴드위드는 1억 7,600만 달러의 현금 및 유가증권을 보유하며 분기를 마감했다. 순레버리지는 전 분기 1.6배에서 1.3배로 하락했다.

회사는 3분기 동안 평균 단가 44.20달러에 주식 225,000주를 1,000만 달러에 자사주 매입했다. 올해 누적 자사주 매입액은 평균 단가 37.35달러에 총 3,000만 달러에 달한다. 또한 밴드위드는 2028년 만기 전환사채 2,100만 달러를 상환해 미상환 잔액은 약 700만 달러가 되었다.

비즈니스 및 영업 성과

상업용 메시징 매출은 약 300만 달러의 선거 캠페인 매출에 힘입어 14% 성장했다. 고객 수요 또한 엔터프라이즈 음성의 15% 성장을 지원하며 음성 사업 내 주요 성장 동력 자리를 유지했다.

소프트웨어 서비스 ARR은 9개월 동안 두 배 이상 증가하여 3,000만 달러를 넘어섰다. 소프트웨어가 클라우드 커뮤니케이션 매출에서 차지하는 비중은 약 5%에 불과했지만, 경영진은 이를 기업의 AI 기반 커뮤니케이션 채택 및 관련 사용량을 보여주는 지표라고 설명했다.

공시된 순유지율(Net Retention)은 113%였다. 선거 캠페인의 주기적 영향을 제외한 상업용 순유지율은 114%였다. 고객 유지율은 99% 이상을 유지했으며, 고객당 평균 연간 매출은 역대 최고치인 27만 3,000달러를 기록했다.

밴드위드는 3분기 중 건당 100만 달러 이상의 신규 계약 및 확장을 4건 기록하며 2026년 누적 건수를 11건으로 늘렸다. 이 가운데는 AI 기반 콜센터를 구축하는 디지털 은행, AI 기반 통화 스크리닝을 위해 밴드위드의 음성 인프라를 사용하는 소비자 개인정보 보호 기업, 플랫폼 및 국제 전화 서비스를 확장하는 통신 서비스 제공업체 등이 포함되어 있다.

회사는 세일즈포스와의 파트너십도 확대했다. 밴드위드는 유럽, 영국, 호주, 뉴질랜드 등 25개국 이상에서 세일즈포스의 자체 콜센터 글로벌 출시를 지원하는 최초의 인증 통신사가 되었다. 경영진은 세일즈포스가 2027년부터 매출에 기여하기 시작할 것으로 예상하고 있다.

경영진 가이던스

밴드위드는 상업용 클라우드 커뮤니케이션 매출 예상치를 400만 달러 올렸다. 반면 올해 선거 캠페인 예상 매출은 300만 달러 하향 조정하여 1,000만 달러로 수정했다.

2027 회계연도의 경우 경영진은 2026년 정치 메시징 수혜를 제외하더라도 클라우드 커뮤니케이션 매출이 두 자릿수 성장을 달성할 것으로 전망하고 있다. 회사는 이러한 성장의 대부분이 기존 고객 기반 및 체결된 엔터프라이즈 계약으로 뒷받침된다고 밝혔다. 경영진은 엔터프라이즈 음성 매출이 2026년 4분기에 20% 이상 성장하고 2027년 내내 견조한 성장을 유지할 것으로 예상하고 있다.

리스크 및 주시 항목

• 정치 메시징은 주기적 특성을 갖는다. 밴드위드는 2026년 선거 캠페인 매출 전망치를 300만 달러 하향 조정했으며, 2027 회계연도는 올해의 기여분과 비교 대상이 된다.
• 대형 엔터프라이즈 도입은 단계적으로 진행된다. 경영진은 2026년 첫 9개월 동안 체결된 연간 계약 가치(ACV) 약 1,900만 달러 중 실적에 반영된 부분은 매우 적으며, 도입은 2027년 중반을 향해 본격화될 것으로 보고 있다.
• 글로벌 음성 요금제는 약 4억 5,000만 달러의 매출 기반에서 운용된다. 경영진은 다수의 기반 연결 서비스에 대한 시장 성장률인 3%~8%를 크게 상회하는 단기 가속화는 기대하지 않고 있다.
• 소비자용 AI 음성 에이전트는 아직 초기 단계에 있으며, 테스트가 한정적이고 초기 물량도 적다. 에이전트 간 프로토콜이 기존 음성 채널 외부에서 일부 저가치·고볼륨 상호작용을 처리할 수도 있다.
• 통신사 메시징 부가요금 매출 증가로 인해 GAAP 기준 매출총이익률이 200bps 감소했으나, Non-GAAP 기준 마진 확장을 막지는 못했다.

애널리스트 Q&A 하이라이트

경영진은 2027 회계연도에 예상되는 클라우드 커뮤니케이션 성장의 대부분이 이미 체결된 계약과 기존 고객 기반으로 뒷받침된다고 밝혔다. 2025년 초 이후 밴드위드는 건당 100만 달러 이상의 계약을 17건 체결했으며, 이는 총 연간 계약 가치(ACV)로 3,000만 달러가 넘는 규모다.

회사는 2027년 성장 궤적이 하반기에 과도하게 쏠리기보다 비교적 균형을 이룰 것으로 예상한다. 경영진은 이러한 가시성의 배경으로 구체적인 고객 이전 계획, 용량 계획, 명확한 사용 사례를 꼽았다.

세일즈포스에 대해 경영진은 파트너십을 통해 세일즈포스가 2028년 말까지 밴드위드의 최대 고객이 될 수 있음을 재확인했다. 이를 달성하려면 자체 콜센터 구축 및 해외 자가 통신사 선택(BYOC) 기회에 힘입어 세일즈포스 관련 매출이 연간 약 2,000만 달러의 실행 연율(Annual Run Rate)에 도달해야 한다.

경영진은 또한 기업들이 AI 사용 사례를 도입함에 따라 기존 음성 인프라를 재검토하는 비중이 늘고 있다고 설명했다. 이러한 프로젝트는 기존 통신사를 대체하는 동시에 마에스트로 오케스트레이션, 소프트웨어 서비스, 사용량 기반 커뮤니케이션 매출을 추가할 수 있는 기회를 창출한다.

실적 발표 컨퍼런스 콜 전문

전체 실적 발표 컨퍼런스 콜 녹취록

경영진 발표

Operator

Good day, and welcome to the Bandwidth Inc. Third Quarter 2026 Earnings Conference Call. [Operator Instructions] Please note, this event is being recorded.

I would now like to turn the conference over to Nils Erdmann, Senior Vice President of Investor Relations. Please go ahead.

Nils Erdmann

Thank you, operator. Good morning, and welcome to Bandwidth's Third Quarter 2026 Earnings Conference Call. I'm joined today by David Morken, Chief Executive Officer; and Daryl Raiford, Chief Financial Officer. We will begin with prepared remarks and then open up the call for Q&A.

During this call, we will make forward-looking statements, including statements related to our future financial performance, expectations regarding the growth of our business. These forward-looking statements are subject to certain risks, uncertainties and assumptions, and actual results may differ materially. These forward-looking statements represent our beliefs and assumptions only as of today, and we undertake no obligation to revise or update any statements to reflect changes that occur after this call.

Further information on factors and other risks that could cause actual results to differ materially from these forward-looking statements is included in our reports filed with the SEC, including our most recent Form 10-K and our forthcoming Form 10-Q and in our third quarter earnings press release and earnings presentation that were issued this morning.

During the call, we'll refer to certain metrics, including non-GAAP financial measures. Definitions of these non-GAAP metrics and other operating metrics as well as reconciliations with the most comparable historical GAAP measures are available in our earnings press release and our earnings presentation, both of which can be found on the Investor Relations page of our website at investors.bandwidth.com.

With that, I will now turn the discussion over to David.

David Morken

Thank you, and welcome, everyone. We are pleased to report solid third quarter results with record revenue of $234 million, up 22% and record adjusted EBITDA of $35 million. Daryl will walk through the financial details shortly.

I want to start by thanking our customers. The trust you place in us grows with every call and message we carry together. To our Bandmates, thank you for the craft and care you put into serving others so well. And I thank God for the gift of serving alongside our exceptional team.

Now we'd like to kick off this call with a few words about where Bandwidth is headed. Today, we are raising the midpoint of our full year 2026 outlook across all of our guidance metrics. Just as important, we have growing confidence in achieving double-digit cloud communications revenue growth in fiscal 2027 even as we lap this year's contributions from political messaging revenue.

We see the fourth quarter of 2026 as the turning point in our growth rates this year with momentum building throughout 2027, and that confidence rests on what is already in hand. Specifically, we've signed 11 deals greater than $1 million of ACV since the start of the year, and they are deploying in phases and will layer into our results over the next 12 months. We're also seeing early traction with Salesforce, which, while not in our 2026 outlook, we expect will be a contributor next year. That's the near-term picture.

The question then is why we believe it lasts. And that brings me to the topic of AI or Super Intelligence, SI, to use the new official terminology. Across the technology sector this year, the same questions keep coming up about super intelligence. Who wins? Does the layer beneath the models become a commodity? How does SI turn into revenue and when? We think the answers are structural, favorable and durable for Bandwidth and come down to 3 things.

First, ownership. We own and operate a global communications network. That gives us the licensed ability to connect, to attest who is calling and to do it with the speed of real-time conversation demands. An SI model cannot replicate that, and SI makes it more valuable, not less.

Second, economics. We monetize interactions, not seats. The industry is still working out how to price SI and what it costs to train and run models. Our revenue follows the conversation, whether a person or an agent is having it and SI creates more conversations.

Third, reliance. Every SI voice or messaging agent needs a number, a trusted identity and access to a human being. That is what we provide, and it's why customers find us easy to deploy and hard to leave. I'll take each of these structural advantages in turn and show you where it's becoming visible in our results.

Let's start with ownership. Software is going headless. Vendors are opening their platforms to third-party agents and giving up the interface. And the fair question is what stays differentiated beneath it. For application vendors, the answer is context.

For Bandwidth, it is execution authority. That means a licensed, owned and operated network, number inventory, caller identity attestation, direct carrier interconnections and local regulatory expertise in nearly 70 countries. These are regulated network assets, not API features. A model layer or an application layer cannot replicate them any more than a generic connector can recreate a CRM.

Context tells an agent what to say. The network decides whether a call connects, whether the customer trusts it and whether it's delivered. We are the trusted infrastructure layer that enables the entire voice ecosystem.

SI is raising the value of that layer. For most of its history, the legacy carrier network was a means of connecting 2 people on a call and all the network had to do was put them in contact with one another.

An SI conversation is different. There is an agent listening and responding. Other applications are, at the same time, transcribing, translating, checking for fraud or pulling up a customer record and all while the caller is talking. Many applications are now part of the conversation and acting on it in real time. That puts new demands on the underlying network. The network has to take part in the call.

On Bandwidth's network, a caller's live audio can be streamed to an agent and any other application surrounding it and the responses stream back in real time. A legacy carrier connection cannot do that. The network has to keep pace. An agent has much less than a second to hear, think and answer.

We can't make the model faster, but because we own the path, we can make sure the network never spends more milliseconds than the model needs. And the network has to be trusted. As SI drives more outbound calling, enterprises will need to prove their calls are legitimate and consumers will need to know who is calling. Voice SI raises the bar for latency, trust and compliance. We're deploying compute in our points of presence to deliver all 3.

You can see evidence of this in the growing number of enterprises that are choosing Bandwidth, all of whom have adopted Maestro, enabling them to take further advantage of our orchestration layer to solve complex integration, routing and communication workflows. It has an intuitive UX that allows enterprises to plug into conversational SI engines and to swap or add SI tools without custom middleware code.

Maestro's prebuilt integrations and network-level security and fraud defenses are without equal, and they were a significant factor in winning a key $1 million-plus financial services customer this quarter. This rapidly expanding all-digital banking platform sought to modernize its customer contact center while integrating an SI agent.

The bank needed low-latency connectivity for real-time SI interactions, a native integration with Genesys and resilient toll-free calling that could scale to support its growing customer base. These requirements highlighted the advantage of direct control over the underlying network and call routing rather than relying on a virtual provider's third-party carrier infrastructure.

Bandwidth displaced a major CPaaS competitor, winning on the strength of our low latency own network, our 5x redundant toll-free solution with Maestro advanced routing and observability, our long-standing Genesys partnership and expert support. This win will translate into high-value enterprise voice traffic on our network with the opportunity to grow alongside the bank as it scales its SI-powered customer interactions.

The second structural advantage we have is economics. Application vendors are still working out price. They're moving from seats to consumption to outcomes and most will tell you, they don't yet know the end state. Our price is set by network value and trust, not seat count. Our model is usage-based. Usage has always been our model, and SI is driving more of it.

Every call or text runs across our network. Whether a person or an SI agent is on the other end doesn't change that. And agentic voice adds legs to many interactions: sentiment analysis, transcription, translation, orchestration between systems. Each leg is expected to be new usage.

Fewer human agents doesn't mean fewer conversations. SI agents answer calls human teams never could, and they place outbound calls that were never economic before. Our network is built for humans and agents without exposure to the economics of the models. We run no freemium and there is no inference cost to us.

That is also why falling inference costs are a volume story for us. As models get cheaper, more SI voice use cases become economic and more use cases means more minutes. We are model agnostic. Maestro works whether a customer runs a cascaded or a speech-to-speech model, and we do not compete with our customers. Maestro gives us a strategic advantage. As enterprises adopt more SI, we can capture both the communications usage they generate and the value of orchestrating those interactions.

Third and finally, reliance. There are 2 parties to every call, and agents are now showing up on both ends. On the enterprise end, agents answer calls and place calls on behalf of business. That is where the growth is today, and it's what the Maestro adoption and the wins I'll describe reflect.

But on the consumer end, the large consumer platforms have just begun giving their personal agents the ability to pick up the phone and call a business on a person's behalf. It's early. It's in limited testing and the volumes will be small for a while, but every one of those calls is an inbound call to a business, into a toll-free number or a contact center, and we carry those calls for many of the enterprises and platforms that will answer them.

When consumers delegate the dialing, we are on the receiving end. Whichever end an agent is on, it needs a number and identity the other side will actually trust and a path to a human. And the business answering needs to know whose agent is calling. That is what we provide, and it is why customers find us easy to deploy with and hard to leave.

We built for that. Our network is a global licensed infrastructure that took decades to build. On top of it, Bandwidth Build and our MCP server let an agent provision its own numbers and configure its own services in plain language and our media streaming carries live audio into the agent and back.

Combined with enterprise-grade compliance and reliability, that is what makes voice the most powerful, secure channel for agentic conversations, and it's why more and more enterprises are turning to Bandwidth. We added 4 million-plus wins and expansions in the third quarter, bringing the total to 11 since the start of the year.

In addition to the digital bank I mentioned earlier, we expanded our relationship with a fast-growing consumer privacy company that now relies on Bandwidth for millions of phone numbers to help protect its customers' personal information. The company has also introduced an SI-powered call screening service that uses our voice infrastructure and voice API transcription to screen unwanted calls and engage callers through conversational SI when needed. Additional services are planned using our WebRTC solution, and we will benefit from opportunities to capture expanded voice and software usage as the customer's business scales.

We also expanded with a major enterprise network and communications provider serving businesses and state and local governments across the U.S. Building on a long-standing relationship that includes toll-free, 911 and other voice services, the customer adopted our platform voice solution, allowing them to retain their own phone numbers while using our network infrastructure for connectivity. This creates the potential for a significant new source of recurring platform fees and incremental usage-based revenue for Bandwidth.

In our fourth $1 million-plus win, a specialized communications provider serving thousands of businesses and government agencies through its reseller and managed service provider network selected Bandwidth to expand its calling services internationally. The customer is relying on our network footprint and our global routing capabilities to serve destinations its previous provider could not support.

Together, these wins demonstrate how our network and software investments create new revenue opportunities, deepen existing customer relationships and allow us to capture more usage as our customers scale.

And with Salesforce, in September, we announced the expansion of our collaboration with Bandwidth named as the first certified carrier to support the global rollout of Salesforce's native contact center. Following our North American partnership, this extends their CRM contact center capabilities to more than 25 countries across Europe, the U.K., Australia and New Zealand with additional markets to follow.

Enterprises will be able to connect to Salesforce's native contact center through a bring-your-own-carrier model, BYOC, enabling SI-driven customer engagement on a global scale. This is a relationship we believe in deeply. Salesforce's own vision is that every customer call should become a conference call with a human, an SI agent and the full context of that customer relationship on the line together.

We share that vision, and we built Maestro on top of our global network to make it possible. We're embedded in Salesforce's governed workflows today with the control, observability and integration depth their platform requires.

Looking forward, the 3 structural advantages that make Bandwidth unique in the SI space are: we own and operate the connectivity layer; we are compensated on interactions, not seats; and every SI voice or messaging agent will need a number, a trusted identity and access to a human being, all of which Bandwidth can provide.

As more and more consumers and businesses are SI and SI agents, these advantages will grow and strengthen. We are excited about solidifying our position of strength, expanding our global footprint, obtaining regulatory coverage in new markets and deploying edge and data center infrastructure to capture this demand.

Now I'll turn it over to Daryl to walk through the financial details of the quarter and our outlook for the remainder of the year.

Daryl Raiford

Thank you, David, and good morning, everyone. We are reporting another quarter with results at the high end of our expectations. Our third quarter 2026 results were as follows.

Total revenue was $234 million, an increase of 22% year-over-year. Cloud communications revenue, which is total revenue less messaging surcharge revenue, reached $158 million, a 12% year-over-year increase.

GAAP gross margin was 36% for the quarter, 200 basis points less than the prior year, solely driven by higher carrier messaging surcharge revenue. Non-GAAP gross margin improved nearly 300 basis points to 60.6%. Non-GAAP gross profit of $96 million increased 18% year-over-year, marking a gross profit yield of approximately 86% on cloud communications revenue growth.

Adjusted EBITDA grew by 43% to $35 million. Adjusted EBITDA margin reached a record 22%, demonstrating the scalability of our business model and our ability to convert revenue growth into meaningful profit expansion.

Non-GAAP net income was $20 million, while non-GAAP earnings per diluted share was $0.50. Free cash flow was $33 million, showing another quarter of strong execution and earnings and cash-generating power of our business.

Turning to cloud communications growth. Both programmable messaging and voice were at the high end of our expectations. In programmable messaging, revenue of $33 million grew 26% resulting from commercial messaging growth of 14%, combined with approximately $3 million of earned messaging revenue during the quarter from political campaign activities. Voice revenue was $125 million, growing 9%, comprising enterprise voice growth of 15% and global voice plan growth of 8%.

Software services revenue was $7 million in the quarter, up 68% year-over-year and exited the quarter at an ARR rate of over $30 million, more than double where it stood 9 months ago. It remains a small share of cloud communications revenue, approximately 5%. We highlight it because customers buy it when they put SI into their communications and the usage follows.

Turning to our operating metrics. Our reported net retention rate for the third quarter was 113%. Adjusted to normalize for the cyclical political campaign revenue impact, our commercial net retention rate was 114%. Customer name retention rate remained above 99%, reflecting the mission-critical nature of our Communications Cloud and the very low churn we consistently deliver. Average annual revenue per customer reached a record $273,000, reflecting both continued customer expansion and our success attracting larger enterprise customers to the platform.

Collectively, these metrics underscore the strength of our existing customer relationships as customers expand their usage, adopt more of our solutions and trust our platform for a broader set of mission-critical communications.

In terms of capital allocation activities, cash and marketable securities at the end of the quarter were $176 million. Free cash flow for the quarter was $33 million, up over 150% year-over-year and representing 21% of cloud communications revenue, driven by our strong operational execution and top line growth.

Our strong cash conversion provides us the strategic flexibility to execute on our capital allocation priorities. During the third quarter, we deployed $10 million to opportunistically buy back 225,000 shares at an average price of $44.20. This brings the total cash used this year for share repurchases to $30 million at an average purchase price of $37.35. Additionally, we repurchased $21 million in aggregate principal of our 2028 convertible notes at a discount to par, leaving approximately $7 million of principal remaining.

Our net leverage reduced to 1.3x at quarter end compared with 1.6x at the end of last quarter, while adding cash to the balance sheet, lowering financing costs and improving the maturity profile. This morning, we also announced that we upsized our revolving credit facility capacity to $200 million and extended the maturity date to October 26, 2031. We have no amounts drawn under the facility. Rather, this move provides us with increased financial flexibility going forward, commensurate with our accelerating growth profile.

Turning to our outlook for full year 2026. We are narrowing our guidance range to reflect our third quarter beat and sustained demand. We are encouraged by the broad-based trends we've experienced since the start of the year. Our confidence is further supported by our expectation for continued accelerating growth in enterprise voice, supported by strong usage demand and the ramp from our recent enterprise customer wins.

Since the beginning of last year, we have closed 17 greater-than-$1 million deals with an aggregate $30 million in annual contract value. The phased deployment of these large-scale implementations provides meaningful visibility into future revenue growth that we expect to ramp over the next 12 months.

For the full year 2026, we now expect total revenue to be in the range of $905 million and $910 million, representing 20% growth year-over-year at the midpoint; cloud communications in the range of $624 million and $626 million, representing 11% growth year-over-year at the midpoint.

Underpinning our guidance, we are raising our expectation for commercial cloud communications revenue by $4 million, while trimming our expectation for political campaign revenue by $3 million to a total of $10 million for the year; adjusted EBITDA to be in the range of $124 million and $125 million, representing 34% growth year-over-year at the midpoint and a 20% EBITDA margin; non-GAAP EPS to be in the range of $1.73 and $1.79, implying full year non-GAAP net income growth of 48%.

Additional modeling details underlying our full year 2026 outlook are as follows. We expect net interest income in the range of $0.6 million and $1.1 million; depreciation expense in the range of $39 million and $41 million; adjusted effective tax rate in the range of 20% and 22%; weighted average diluted shares outstanding of approximately 39 million; and capital expenditures in the range of $24 million and $26 million.

In summary, we are pleased with the strong growth we delivered in the third quarter, our early proof points in voice SI traction and our robust profitability and free cash flow. As we close out the remainder of 2026, we remain committed to delivering sustainable, profitable growth.

With that, I'll now turn the call back over to David.

David Morken

Thank you, Daryl. Before we conclude, I want to welcome Chip Schroeder, who joined our Board of Directors this week. Chip spent more than a decade at Silver Lake, advising the leaders of large technology businesses on go-to-market strategy and capital allocation. As Bandwidth builds on the opportunities we've discussed today, that experience will be invaluable, and we're very glad to have him.

Finally, I'd like to invite you to join us at Reverb 2026 here in Raleigh, North Carolina on November 10. Reverb has become our flagship customer and product event, where we'll showcase how enterprises are deploying SI into production and how Bandwidth is making it possible. We hope to see you then.

I'll now turn it over to the operator for questions and answers.

Operator

[Operator Instructions] The first question comes from Joshua Reilly with Needham.

질의응답

Joshua Reilly

As you look at the $1 million-plus customers you've added to the business here year-to-date, how do you go about determining what volumes to include in guidance to make a conservative estimate assumption there? And then secondarily, how should we think about the pipeline to finish out the year for $1 million-plus customer adds?

David Morken

Thanks, Josh. Our team has become very good at calibrating expected volumes from customers as well as assessing the onboarding time frames, and that's what's allowed us to be quite accurate again in 2026 in providing the guidance that we have. So as you look forward to how the $1 million-plus deals, which now year-to-date stand at 11, basically double '25 and maybe we even have a chance to triple that number by the end of the year, you can depend upon us to understand the onboarding time frames and the outcomes.

And whether that's because we have a firm grip on the use cases or minimum commits, what the team has demonstrated with the guidance accuracy really since inception as a public company is that we will dial it in tightly.

Joshua Reilly

Great. And then in terms of Agentforce initially ramping here, at Dreamforce, there was announcement you also won some international opportunities for that as well. Can you just discuss your expectations around the ramp for domestic versus international revenue opportunities from Salesforce?

David Morken

Yes. We're excited about Salesforce and have talked about our relationship with them yielding potentially them being our largest customer exiting 2028. And the work that we've done with them again in this last quarter continues to give us confidence in that.

One of those developments in the quarter, as you mentioned, was the announcement that we are the first BYOC partner for Salesforce now globally and internationally. And that's exciting for us and extends our capacity and our relationship with them.

Maestro becomes very, very important. Cross-border regulations and local networks are always hard to navigate. And so we expect our network and Maestro combination for Salesforce globally to incrementally add to the relationship that we've already begun with them. And I think there's more to follow.

Operator

The next question is from Erik Suppiger with B. Riley.

Erik Suppiger

First off, on the enterprise voice, do you expect that to accelerate in fiscal '27? Do you think -- what kind of growth are you anticipating, given the pipeline of enterprise deals that you've had? And then secondly, on the software ARR, I think you grew about $4 million sequentially. Is that kind of a reasonable growth rate going forward? Or how should we think of the growth on the software ARR going forward?

David Morken

Erik, thanks for the good question. When you look at what we've said about 2027 being a double-digit cloud communications growth year, then yes, you absolutely can assume enterprise voice growth rate will be healthy and growing during and throughout 2027. That's a safe assumption.

And on number two, let me ask Daryl to comment on the $4 million in the software services category.

Daryl Raiford

Erik, this is Daryl. We do think that we do look for it growing in -- continuing in the fourth quarter and into the 2027. It's implied in our outlook for 2027 double-digit cloud communications revenue growth, irrespective of the political campaign cycle. And so -- and it is a leading -- as I said, we do track it. While it's a small -- it's growing from a small base, it is a leading indicator in terms of how enterprises are adopting use cases in SI and really keenly becoming AI-enabled.

Operator

The next question is from Koji Ikeda with Bank of America.

Koji Ikeda

I wanted to ask another question on 2027. And so how much of the 2027 communications growth guidance is already secured by signed customers ramping? And what are the assumptions on pipeline conversion to make up the rest?

David Morken

Yes. The vast majority, Koji, is because of what we already have in hand, and that goes all the way back to the beginning of last year and the now 17 $1 million-plus deals on top of our installed base. So the biggest contribution by far, implicit in your question, is from both the existing customer base and the enterprise deals that we've signed so far.

And the second question you have, I think, is on pipeline and the pipeline contribution. Let me ask Daryl to respond to that.

Daryl Raiford

We do -- sorry, Koji, this is Daryl. We do have good visibility into the pipeline. I mentioned that over the last -- since 2025 and the 9 months of 2026, we have aggregated deal value -- annual contract value of greater than $30 million.

To put a fine point on that, in the 9 months with our 11 customers, that's almost $19 million of annual contract value. Very little of that has played into the forecast or into our results right now. We see that ramping into the mid-2027, and it gives us some really good confidence about that.

That's not to say that the pipeline, as David said, isn't robust going forward and the closure rate, we expect to be just as strong. But as a fact, we do have deals closed that would give us good visibility.

Koji Ikeda

Got it. And for a follow-up here, maybe more high level. So Salesforce has been very active on the M&A front. And so what is it specifically that Bandwidth offers that it's not a risk that you're -- what you're saying could be your largest customer in a couple of years goes out and buys a Bandwidth competitor to take what you offer in-house?

David Morken

I think the risk is tantamount to 0. The nature of the underlying network and platform combination that Salesforce values now, both domestically and internationally, is extremely rare. I mean like rare earth rare. So I would estimate it at 0.

Operator

The next question is from Elizabeth Porter with Morgan Stanley.

Elizabeth Elliott

I just wanted to follow up a little bit on the growth outlook for next year. So could you help us understand how these customers typically ramp their progress, and if any 2027 growth should be relatively consistent or more back-end loaded? It looks like the cloud comms, ex the campaign, slowed a little bit in the quarter. So I would just love to unpack your view on the shape of growth into next year. And as we think about the outlook, can you just put a finer point on where you have the most cushion in your assumptions and where you see the greatest potential for upside?

David Morken

Thanks, Elizabeth. Let me ask John Bell, our Chief Product Officer, to begin our answer.

John Bell

Yes. So I'll start to give more color on how we work with our customers. We work very closely with our customers. Many of them are doing a very sophisticated migration process to our platform. As part of that, we have a deep understanding of their use cases, their volumes, their customer base. Capacity planning is a natural part of that project.

And so as we work with them, we understand both in great detail, what they have today and the vision for the future and work with them in planning towards that. So as we look at these opportunities, we have deep plans we co-developed with these customers as we move forward with them.

David Morken

And I'll add, Elizabeth, the second part of your question, I think, was the continuity or the slope throughout 2027. And when we think about double-digit cloud comm growth that comes from pipeline, that comes from the projects John mentioned, that slope is reasonable. It isn't a back-loaded severe slope at all. It has to do with how we perform our annual AOP budget process here.

And this represents, in 2027, the 10th such year that we've executed this process and cycle internally, and that's yielded 36 straight quarters of achieving or exceeding guidance. So that's 9 years in a row. I think that's given us good credibility when we shape and look forward to 2027 as a reasonable year.

Elizabeth Elliott

Great. And then just as a follow-up, some of the customer highlights this quarter involved displacing competitors. And so are you seeing more opportunities like this as enterprises reassess their voice infrastructure for AI use cases? And how much of the growth opportunity is from competitive displacement versus expansion in the overall market?

John Bell

This is John again. Yes is the short answer. As more and more enterprises see the opportunity for SI in their businesses, they are forced to reconsider their legacy voice infrastructure. That is the opportunity for us. And so consistently, that is a conversation we see.

As mentioned before, there's always a migration as well as a vision towards the future as they expand what they're doing with voice and really looking at voice as a new digital channel that is oftentimes displacing other digital channels they've been using as well. So we're very excited about these opportunities because they are really very visionary.

David Morken

And that displacement, Elizabeth, that John mentioned is most frequently away from the incumbents. So when an enterprise has worked with AT&T or Lumen or Verizon for decades, this moment in time with SI and the SI readiness imperative is driving that displacement.

Operator

The next question is from Patrick Walravens with Citizens.

Patrick Walravens

Great. So Dave, I love my Grok Bot. And part of the reason I like my Grok Bot so much is because when I first started using it, I asked it to go get a phone number and it went to Bandwidth Build, and it got a phone number, and it's always the same phone number. So when I have my Grok Bot call someone, it announces itself as this is Patrick's chief of staff, and he would like to know whatever. And then it can get the answer and it can relay it to me. But that seems like a fantastic opportunity for you guys to me. So where are we on Bandwidth making it possible for personal agents to have their own phone number that persists?

And I guess my follow-up observation on that is Elon Musk was talking about a 1,000x increase in the usage of Starlink over the next 3 to 5 years. How do you feel about the potential increase in the usage of the Bandwidth network?

David Morken

Thanks, Pat. Q3 was day 1 for consumer voice SI agents. You were there at the beginning, and it is taking off. Whether it's Instinct, whether it's Muse, Dots, Grok Bot, Duo or the plethora of emerging other SI voice agents, the underlying speech-to-speech models are extraordinary right now and developing intuitive human-like interfaces, the intelligence, the speed. Soon, the focus will be on latency at a network level where we have an extraordinary value.

But Q3 was day 1. And where it goes from here is super exciting because the utility, the usefulness, the value of those conversations between super intelligence and people in a commercial context, what it represents for Bandwidth, is the first time in our history that we have served consumers, individuals like you with a Grok Bot coming to Bandwidth Build, getting a number, beginning to have their SI agents calling on their behalf, and that is exciting.

You'll have some of these SI agents at scale, integrate and go native. And what we've learned from the hyperscaler successful path that we've navigated is that, often they come to Bandwidth when they reach scale. They may use an API layer at launch, and that's easy and quick. But as they scale, they begin to value the uniqueness of our platform and network globally. And so Q3 was day 1. And the bright horizon here is Bandwidth has a consumer frontier to pursue with our platform and network, and that is super exciting for us.

Operator

The next question is from Arjun Bhatia with William Blair.

Arjun Bhatia

Perfect. David, maybe I'm going to stick on that topic of consumer agents. I think certainly, it seems like there's a pretty big opportunity for this to inflect voice volumes and for Bandwidth to play a role there. But we're also starting to hear about these sort of agent-to-agent protocols where if a consumer agent is calling an enterprise and they have their own AI agent that's answering inbound calls, maybe that call doesn't -- or it doesn't need to be a call, it doesn't need to go over telephony, maybe there's another way for agents to talk to each other, right, that aren't human.

So is that -- do you think that's going to scale and gain traction? And what does it mean for sort of Bandwidth's opportunity as agents proliferate both on the sort of the consumer and the enterprise side? Or is this just sort of a rising-tide-lifts-all-boats sort of moment?

David Morken

I think it's both. I think there is a massive rising tide right now of SI agents on our behalf making calls, and that is super exciting. You're right, agent-to-agent protocols should and we want them to emerge and develop successfully so that lower value, perhaps higher volume, less nuanced, less important transactions and conversations can occur at an agent-to-agent level.

But what that implies is the highest value exchanges often are going to be between SI and a person across all kinds of both relationships, transactions, you name it. And that's what we're excited about. It's a huge new tide. And we're okay and love the new agent-to-agent protocols because many things that agents do together should not be done in voice, but don't lose sight of the moment that we're in for SI voice agents. But let me ask John to respond as well.

John Bell

Yes. And let me just add there that today, it's pretty inefficient for an agent to use a website or other existing digital channels like mobile apps to communicate. I think that is actually the primary value is around some of those existing digital channels that we've had to use in the past. What's unique about the voice channel really is it's used for mission-critical applications where time matters and where people and SI voice agents are working together and collaborating.

And so there's some really unique aspects to that channel that we're excited about. And again -- so we're actually excited about these A2A interfaces. I think they're more impactful to legacy digital channels, but we look forward to working with them.

Arjun Bhatia

Perfect. That's very helpful. And then one thing I wanted to just touch on the outlook. We've talked about the enterprise ramp, and I think we can sort of start to see it happening already, and it sounds like Q4 is going to be another maybe inflection point in that going into 2027. How should we think about the other part of the voice business, global communications plans or global voice? Is there room for that to also sort of reaccelerate into 2027? Or are there different sort of dynamics at play there? I guess how should we think about that relative to the enterprise business?

Daryl Raiford

This is Daryl. I think the short answer is yes, to get to the punchline. I do think that there is some good context that goes with that just in terms of the mix and how we're thinking about our business because you did mention enterprise and then you mentioned global voice plans and the like. So this is the third quarter in a row that we have raised our communications revenue guide.

We've raised it $4 million into the fourth quarter now for the full year. We did trim back political campaign revenue -- messaging revenue by about $3 million. And that $4 million raise in cloud for the remaining 3 months of the year for cloud communications is commercial, and it's a combination of messaging and voice.

In terms of enterprise voice, to put a fine point on it, we are expecting it to grow greater than 20% in the fourth quarter, and we're expecting commensurate and growing rates in '27. We're staying basically at the headline of double-digit cloud communications growth in 2027, irrespective of the political campaign revenue we experienced this year, but we do expect enterprise to continue to grow when we report 2027.

In terms of global voice plans, it is a context, it is a very large base. It's a $450 million base, and it's both exposed to growing usage across contact centers as well as a fair amount of very sticky but stable product offerings, for example, emergency services, where we're #2 in the market share in the United States. We don't think a base of global voice plans at this moment in this current period in the next very near term, few quarters, is going to accelerate meaningfully above market rates.

Market rates for many of those sorts of base-level connectivity services is somewhere between 3% and 8%, and we feel pretty strong about that. Where we are really focused, again, is on the enterprise -- the large enterprise acquisition. And not only have we been focused, we've been very successful in bringing in so many of these large deals over really kind of setting the company's record.

I think just this year in 2026, in the 9 months, when we've announced 11 wins greater than $1 million, that's nearly $19 million in annual contract value in 9 months. And we do see that playing very nicely into our total voice growth and meaningfully into enterprise voice growth next year.

David Morken

And I'd just add to that. In the $1 million wins we mentioned this quarter, we continue to see innovation driven by SI in our existing GVP customers and continue to see a cohort of new GVP customers who are the SI voice agent customers themselves. As they scale and reach production, they become customers.

Operator

The next question is from James Fish with Piper Sandler.

James Fish

Circling back on the Agentforce CX relationship here. What are you guys seeing underneath with the move to Agentforce CX from another cloud provider versus some of the on-premise incumbents? And how should we think about that ramp as we go into '27, '28 in terms of how it impacts your segments underneath?

David Morken

James, this is David. I'll start, and I'll actually take your question in reverse order. For us to have a ramp with Salesforce that supports the conviction we have about them exiting '28 as potentially our largest customer, they then would need to, as we've talked about publicly, get to somewhere around a $20 million annual revenue run rate by then. So you can imagine that, that trajectory has to be in throughout '27, if not right now, and be healthy throughout '27 and then '28 to get there.

And that conviction is partly supported by the international BYOC relationship launch that was announced in the quarter that we are now powering the voice calling and messaging for both the native SI-driven contact center, as your question mentioned, but also now having the BYOC architecture available and Maestro available to facilitate the global connectivity for Salesforce. Both are important. Both are very different customer and customer use cases, often different customer sizes.

But you can imagine there are many, many global customers that have incumbent relationships for many, many years that need to be able to transition a complex environment using Bandwidth, using Maestro and using Salesforce. And so it's a market segment distinction and a geographic distinction that drives the 2 different relationships and paths to market. Both are modeled, both are understood, both pipelines are building, both are projected and inform how we have growing conviction, not lessening conviction, about Salesforce as a partner for us over the next 2 years plus.

James Fish

Got it. And then, Daryl, can you help us bridge the '27 comments here? I don't mean to belabor the point, but let's face that the 6 deals from last year entering this quarter were almost fully deployed. And I think you just said $19 million for the 11 $1 million-plus, year-to-date, but that still leaves a fairly large sort of go-get to sort of hit double-digit growth at least on the cloud communications side and understanding you face a headwind on messaging just in aggregate because of the political and related fees. But it still seems like that's a fairly large go-get when you kind of back out that $19 million.

Daryl Raiford

Right. Yes, I don't mean to leave you with the impression that our aspirations and projections for 2027 rely totally and solely upon enterprise voice growth rate. You've seen our net retention rate be very steady in the double digits over the course of the last several quarters. We have a growing base. We have other -- we have still commercial messaging. We expect to continue growing that market. We have other wins. We have many wins that are not above $1 million, right, so to speak.

And so I don't mean to leave you with that impression. Your point was around enterprise. And so I wanted to give you color into that. But no, just -- we have a number -- we have a very large base of customers, and we have a number of different product solutions, and we expect that to continue to demonstrate the growth that it has in the past.

Operator

[Operator Instructions] At this time, there are no further questions. This concludes our question-and-answer session. It also concludes our conference. Thank you for attending today's presentation. You may now disconnect.