From Madhur Jain | Product & Market Analysis

Twilio Q3 2026: Voice AI Is Real, and Net Expansion Is About 110% Without Fees

On this page

Twilio reported a 114% dollar-based net expansion rate for Q3 2026, its strongest run in years, and the company credits voice AI for much of the lift. Its own CFO said roughly 4 of those points came from US carrier fees passed through to customers. Strip them out and Twilio's net expansion is about 110%, against 109% a year earlier. Voice AI is showing up in usage revenue. It is showing up smaller than the headline.

Key takeaways

  • Twilio's 114% Q3 2026 net expansion includes about 4 points of carrier fees. CFO Aidan Viggiano said so on the 29 October 2026 call, which puts the underlying rate near 110% against 109% in Q3 2025.
  • Organic growth is strong but not accelerating. Organic revenue grew 12%, 16%, 17% and 16% over the last four quarters, and Twilio guided Q4 2026 to 11% to 12% organic.
  • Twilio does not disclose voice AI revenue. The only hard figures are growth rates, a few customer run rates, and a CEO estimate that AI holds about 1% of total available voice minutes.
  • Voice AI is not squeezing Twilio's margin. Excluding pass-through fees, non-GAAP gross margin rose to about 51% in Q3 2026 from 50.1%, because the AI customer, not Twilio, pays the model vendor.
114%Dollar-based net expansion rate, Q3 2026, up from 109%. Source: Twilio 8-K, October 2026.
$53.1MUS carrier A2P fees Twilio excluded from Q3 2026 organic revenue, 3.4% of reported revenue. Source: Twilio 8-K reconciliation table.
1%Share of total available voice minutes the CEO says AI has reached so far. Directional, a management estimate. Source: Q3 2026 call transcript.

What Twilio reported in its Q3 2026 earnings

Twilio's Q3 2026 results, released on 29 October 2026, were a clean beat on every line the company guides to. The question for anyone building on voice AI is which lines the AI story actually moved.

Revenue was $1.56 billion, up 20% reported and 16% organic. The gap between those two numbers is almost entirely carrier fees. US mobile carriers raised application-to-person (A2P) messaging fees, and Twilio passes them through to customers as revenue. Twilio strips them out of its organic figure, along with a small amount of acquisition revenue.

The reconciliation in the Q3 2026 earnings release filed with the SEC removes $53.1 million of A2P revenue and $1.8 million of acquisition revenue. Organic revenue lands at $1.505 billion. Against $1.30 billion a year earlier, that is the 16% organic growth rate the company reports.

Non-GAAP income from operations was $325.7 million, up 39%, for a 21% operating margin. Free cash flow was $369.6 million, a 24% margin. Twilio raised full-year 2026 organic growth guidance to 15%, from 13% to 13.5%. On the call it also gave a first look at 2027: organic revenue growth of at least 10%, and non-GAAP operating income of at least $1.35 billion.

The release itself says little about AI beyond positioning. Its one AI sentence is CEO Khozema Shipchandler's: "Consumer AI has accelerated AI development, and its adoption demands fast, reliable infrastructure built for AI speed". The specifics live on the earnings call, which is where every AI number in this post comes from.

Twilio organic growth, net expansion and AI disclosures over four quarters

One quarter is a headline. Four quarters is a trend you can test. The table below lines up every quarter since Twilio began talking about voice AI as a named growth driver. It uses the filed releases for the financial lines and the call transcripts for the AI lines.

Twilio quarterly results and AI disclosures, Q4 2025 to Q3 2026.
QuarterRevenue, growthOrganicA2P fees outDBNE (prior year)Non-GAAP marginAI disclosure (quoted)
Q4 2025$1.366B, +14%12%$23.1M109% (106%)49.9%CFO: voice AI "revenue growth accelerated above 60% year-over-year".
Q1 2026$1.407B, +20%16%$46.1M114% (107%)About 49.6%Release: "foundational infrastructure layer in the era of AI". No voice AI figure in the filing.
Q2 2026$1.499B, +22%17%$71.1M116% (108%)49.1%CFO: "Incremental carrier fees contributed roughly 5 points to DBNE". CEO: "very early innings".
Q3 2026$1.560B, +20%16%$53.1M114% (109%)49.2%CFO: "Incremental carrier fees contributed roughly 4 points to DBNE". CEO: AI at "1% penetration" of voice minutes.

Revenue, organic growth, A2P exclusions and net expansion are from Twilio's filed earnings releases. Q1 2026 gross margin is computed from the release (non-GAAP gross profit of $697.5M on revenue of $1,406.9M); the release rounds it to 50%. Quotes are from The Motley Fool's call transcripts, which carry their own accuracy disclaimer. Last verified 11 November 2026.

What changed between Q4 2025 and Q3 2026

Organic growth stepped up from 12% to 16% in Q1 2026 and has held in a 16% to 17% band since. That is a real change in the business. In February 2026 Twilio had guided full-year organic growth at 8% to 9%, and it has raised that range three times since, to 15%.

The guidance pattern is worth noticing too. Each quarter Twilio has guided the next one at 10% to 12% organic and then delivered 16% or 17%. On the Q2 call Viggiano was blunt that the 5-point beats were unusual: "We don't expect that to be the new norm for the business". On the Q3 call he put the historical beat at 2 to 4 points.

The filed releases never quantify AI. They use positioning language, "infrastructure for customer engagement in the AI era", and stop there. Every number about voice AI arrives orally, in prepared remarks or analyst Q&A, and none is defined in a filing. That matters for how much weight you put on them, a point developed in the ranking of how software companies disclose AI revenue. A growth rate with no base is a direction, not a size.

Twilio's net expansion rate without carrier fees is about 110%

Dollar-based net expansion (DBNE) is the cleanest test of whether existing customers are spending more. It is also the line most inflated by the carrier fees, because those fees land on existing messaging customers.

Twilio's definition compares this quarter's revenue from customer accounts active a year ago with the same accounts' revenue in that earlier quarter. New customers are excluded. Acquired revenue is excluded until its first anniversary. Pass-through carrier fees are not excluded, so a fee increase that customers pay looks exactly like customers buying more.

That is not a hidden trick. Management volunteered the fee contribution on both of the last two calls. The headline number simply carries it, and most coverage repeated the headline. For context on how 114% compares with other software companies' retention figures, see the 29-point spread across 12 Q2 filings.

Twilio dollar-based net expansion, reported and excluding carrier fees Percent. Axis starts at 100%. Ex-fee bars subtract the CFO's stated fee contribution; Q4 2025 and Q1 2026 have no verified figure. 100 110 106 109 Q4 2025 107 114 Q1 2026 108 116 ~111 Q2 2026 109 114 ~110 Q3 2026 Same quarter a year earlier Reported Excluding carrier fees (Zan Digital arithmetic). Source: Twilio earnings releases (Exhibit 99.1) and Q2 and Q3 2026 call transcripts.
Compare the red bar with the light bar beside it, not with the dark one. On an ex-fee basis Twilio's expansion improved by about 3 points year on year in Q2 and about 1 point in Q3.

The 110% number, and what it says

Subtracting the stated contribution gives about 111% for Q2 2026 and about 110% for Q3 2026. The year-earlier figures were 108% and 109%. So underlying expansion did improve, by roughly 3 points and then 1 point. Viggiano said as much on the Q2 call: "we did see DBNE accelerate even adjusting for the fees."

The improvement is real but modest, and it narrowed in Q3. That is the single most useful number in this post, because it is the closest public proxy for whether existing customers, including the voice AI ones, are compounding their spend. They are, at about 10% a year. That is healthy for a usage business. It is not the 14% to 16% the headline implies.

My view is that the ex-fee figure should be the one quoted. A rate that moves with a carrier's pricing decision tells you about AT&T and Verizon, not about Twilio's customers.

How much voice AI shows up in Twilio's usage revenue

The honest answer is that nobody outside Twilio can size it. The company discloses growth rates and anecdotes, never a revenue base. What can be assembled is a set of bounds and signals, laid out below.

What Twilio has and has not disclosed about voice AI

On the Q4 2025 call in February, Viggiano said voice revenue growth reached the high teens. He called it "its best growth rate since 2022", aided by voice AI revenue growth above 60%. By Q2 2026 voice revenue growth was above 20%, and in Q3 it was again "20% plus", which Viggiano described as "a balance of volume growth and software add-ons".

Customer-level numbers are thin and mostly unnamed. On the Q2 2026 call, the company described a horizontal conversational AI customer that started in Q1 2025 and reached a $6 million annual run rate, growing 65% a year. A verticalised conversational AI customer had reached $9 million. Twilio also said it signed an 8-figure deal with an unnamed "leading AI company".

In Q3 the named wins were Ushur, Eltropy, Gong and Meera.ai, plus an unnamed health insurer that became Twilio's first 8-figure Conversations deal. Coverage in CX Today on 3 November 2026 reported the same customer list.

There is no voice AI revenue figure, no voice AI customer count, and no share of revenue from AI-native companies. Voice revenue itself is not broken out in the filings. On the Q3 call, Viggiano said messaging is about 60% of revenue and self-serve about 15%. That leaves voice inside a residual that also holds email, software add-ons and everything else.

Asked on the Q2 call how concentrated the AI spend was, Shipchandler said: "we don't have a measure per se of where the different pockets of spend are". I take that at face value as a reporting gap, not evasion. It still means a reader cannot test concentration risk.

Putting a ceiling on it with the 1% figure

The one sizing statement came from the Q3 Q&A, as reported by The Motley Fool transcript. Shipchandler said voice AI is "still 1% penetration in terms of the total available voice minutes out there." He framed the remaining 99% as runway.

Read it the other way and it is also a ceiling on today's contribution. If AI handles about 1% of available voice minutes, voice AI cannot yet be a large share of a business where voice is the minority product. A 60% growth rate on a small base produces a few points of total growth, not the whole acceleration. For a sense of how the contact-centre software vendors downstream report the same shift, compare the NICE and Five9 disclosures on AI revenue against seats.

Twilio Q3 2026 revenue bridge: where the 20% came from $ millions. Axis starts at $1,200M to show the steps; bar heights are to scale above that line. $1,200M 1,300.4 Q3 2025 +204.2 Organic growth 1,504.6 Organic revenue +53.1 Carrier fees +1.8 Acquisition 1,559.5 Q3 2026 reported Source: Twilio Q3 2026 organic revenue reconciliation, Exhibit 99.1 to Form 8-K. Figures in $ millions.
The red step is about 4 points of the 20% reported growth and earns Twilio close to nothing. Voice AI sits somewhere inside the blue step, unsized.

Is a usage-billed CPaaS an agent-era winner or an inference pipe?

The bear case on communications platforms (CPaaS) in the agent era runs like this. Voice agents are mostly model inference. The model vendors capture the margin, and the phone line becomes a commodity pipe billed by the minute. The Q3 numbers let you test that directly.

Who pays the model bill

A voice agent built on Twilio typically has at least two cost lines for its builder: the model vendor's inference bill, and Twilio's per-minute voice and add-on charges. Twilio does not buy the inference. The AI company does. So the inference-pipe risk is not that model costs land in Twilio's cost of revenue. It is that the builder's total cost per call is dominated by inference, which caps what Twilio can charge. The same squeeze is examined from the builder's side in the analysis of inference costs and AI margins.

No public figure splits a voice agent's cost per minute between model and channel. I am not going to estimate one. What can be checked is whether Twilio's own margin is moving the way a commoditised pipe's would.

Gross margin without the fees says Twilio is not being squeezed

Reported non-GAAP gross margin fell to 49.2% from 50.1%. That looks like compression. It is not. Management said the pass-through fees carry essentially no gross profit. Removing them from revenue, Twilio's Q3 non-GAAP gross profit of $767.8 million on $1,506.4 million of non-fee revenue gives a margin of about 51.0%. Viggiano put the ex-fee change at up 80 basis points year on year, which matches.

The drivers he named were mix toward "higher-margin products" and cost work, including a finished email migration. Branded Calling and Conversational Intelligence grew at triple-digit rates in Q2. A pipe being commoditised by its largest new customer class does not usually expand gross margin and sell more software on top.

Shipchandler's pitch is neutrality. "We're the neutral party," he said in Q3, adding that application-layer AI players "are running off of Twilio's rails". On the Q2 call he called Twilio "the Switzerland of it all" in response to a question about an OpenAI voice product.

I think the neutrality argument is right for now and fragile later. Carrier interconnects, number inventory, compliance and branded caller ID are hard to replicate. They are also exactly what a well-funded model vendor would buy or partner for if voice became core. The table below sets out how I read each test today.

Agent-era winner or inference pipe: the evidence in Twilio's Q3 2026 numbers.
TestWhat a pipe would showWhat Twilio showsVerdict
Gross margin excluding feesFalling as AI volume growsAbout 51.0%, up from 50.1%Winner, for now.
Software attach on voiceRaw minutes onlyAdd-ons growing faster than minutesWinner.
Net expansion excluding feesFlat or fallingAbout 110%, up about 1 pointModestly positive.
Disclosed AI revenue baseNot applicableNone disclosedCannot judge.
Where the pipe case winsModel vendors bundle telephonyOpenAI voice product treated as partnerOpen risk.

Where this argument is weakest

Three parts of this analysis rest on less than a filing, and each could move the conclusion.

The ex-fee arithmetic is mine, built on spoken figures

The roughly 110% and 111% net expansion figures subtract a CFO's rounded verbal estimate from a filed number. "Roughly 4 points" could be 3.5 or 4.5. Twilio has never filed an ex-fee DBNE, and the transcripts I used are third-party and carry a disclaimer about errors. Treat the ex-fee figure as directional to within a point. The ex-fee gross margin has the same weakness: it assumes the fees carry zero gross profit, which management stated but did not quantify.

Concentration risk is invisible, in both directions

If a handful of voice AI startups drive most of the voice acceleration, a funding downturn among them would hit Twilio faster than the numbers suggest. Shipchandler acknowledged "there are probably going to be some ups and downs as the AI story unfolds". The opposite is also possible: AI-driven expansion could be broad enough that my reading undersells it. Twilio's own disclosure does not let either case be proven, which is why the takeaway above says "cannot judge" rather than guessing.

The Q4 guide may simply be conservative

Organic growth guided to 11% to 12% looks like a sharp deceleration from 16%. Twilio has guided this way for three straight quarters and beaten by 4 to 7 points each time. Viggiano called Q4 "a more volatile quarter" for a usage business. If Q4 lands near 15% again, the trend reading in this post is too cautious. The case for metered revenue holding up better than seat revenue is also stronger than I give it here; see Datadog's Q3 split of metered and seat growth for a parallel.

What to watch in Twilio's Q4 2026 results

Twilio reports Q4 2026 in early 2027 and has an investor day on 13 and 14 April 2027. Four lines will settle most of the open questions above.

Twilio Q4 2026: the lines that matter for the voice AI thesis.
WatchWhy it mattersBad reading
Fee contribution to DBNEQ4 guidance assumes about $75 million of incremental fees, the largest quarter yetEx-fee DBNE at or below 109%.
Organic growth against the 11% to 12% guideShows whether usage momentum survives the holiday quarterA beat of 2 points or less.
Voice revenue growthVoice faces tougher comparisons in the second half, per the CFOVoice growth below the high teens.
Any voice AI revenue baseInvestor day is the natural moment to disclose oneAnother quarter of growth rates with no base.

The fee line deserves the closest reading. With roughly $75 million of incremental fees assumed in Q4 against $53 million in Q3, the reported DBNE could rise even if the underlying rate slips. A headline of 115% next quarter would tell you nothing on its own. How pass-through and usage meters distort retention metrics more generally is covered in the piece on hybrid pricing and the NRR gap.

Frequently asked questions

What were Twilio's Q3 2026 earnings results?

Twilio reported Q3 2026 revenue of $1.56 billion, up 20% reported and 16% organic, on 29 October 2026. Non-GAAP operating income was $325.7 million and free cash flow was $369.6 million. Dollar-based net expansion was 114%. The company raised full-year 2026 organic growth guidance to 15% and guided Q4 organic growth at 11% to 12%.

What is Twilio's dollar-based net expansion rate?

Twilio's dollar-based net expansion rate was 114% in Q3 2026, against 109% a year earlier. It compares revenue from existing customer accounts with the same accounts a year before. The CFO said US carrier fees passed through to customers contributed about 4 points, so the underlying rate is roughly 110%. In Q2 2026 the reported rate was 116%, with about 5 points from fees.

How much revenue does Twilio get from voice AI?

Twilio does not disclose voice AI revenue. It has disclosed growth rates, including voice AI revenue growth above 60% in Q4 2025 and total voice growth above 20% in 2026. The CEO said in Q3 2026 that AI accounts for about 1% of total available voice minutes. Without a revenue base, the dollar contribution cannot be calculated from public data.

Why is Twilio's organic growth lower than reported growth?

Twilio excludes pass-through fees that major US carriers charge for application-to-person messaging, plus revenue from recent acquisitions. In Q3 2026 those exclusions were $53.1 million of carrier fees and $1.8 million of acquisition revenue. The fees add revenue but almost no gross profit, so organic growth of 16% is the better measure of demand than 20% reported.

Do AI model vendors take Twilio's margin on voice agents?

Not on current evidence. The AI customer pays the model vendor directly, so inference costs do not sit in Twilio's cost of revenue. Excluding carrier fees, Twilio's Q3 2026 non-GAAP gross margin was about 51%, up from 50.1% a year earlier. The longer-term risk is that model vendors bundle telephony themselves, which has not shown up yet.

What is Twilio's guidance for 2027?

On the Q3 2026 call, Twilio said it expects 2027 organic revenue growth of at least 10%, above its 2025 investor day framework of 7% to 8%. It expects non-GAAP income from operations and free cash flow of at least $1.35 billion each. Detailed 2027 guidance is due with Q4 results and at an investor day in April 2027.

Where to start if you run voice agents on Twilio

If you build voice agents, pull your last six Twilio invoices this week. Split each one into three buckets: carrier fees, raw minutes, and software add-ons. The fee line is now large enough to distort your own unit economics the same way it distorts Twilio's net expansion, and your cost-per-call model should carry it separately.

Then put your inference bill next to your Twilio bill for the same calls. Whichever line is larger tells you where your negotiating effort belongs. For most teams I would expect it to be the model, which is why the margin questions in the breakdown of real AI gross margins matter more to you than Twilio's pricing does.

Related on usage-billed infrastructure

For how other metered businesses report the AI shift, read Datadog's Q3 metered versus seat growth and what NICE and Five9 disclose about AI revenue.

References

  1. Twilio Inc., Twilio Announces Third Quarter 2026 Results, Exhibit 99.1 to Form 8-K, 29 October 2026. Used for all Q3 2026 financials, net expansion and the organic reconciliation.
  2. Twilio Inc., Twilio Announces Second Quarter 2026 Results, Exhibit 99.1 to Form 8-K, 6 August 2026. Used for Q2 2026 figures.
  3. Twilio Inc., Twilio Announces First Quarter 2026 Results, Form 8-K exhibit as mirrored by Last10K, 30 April 2026. Used for Q1 2026 figures.
  4. Twilio Inc., Twilio Announces Fourth Quarter and Full Year 2025 Results, Form 8-K exhibit as mirrored by Last10K, 12 February 2026. Used for Q4 2025 figures and initial 2026 guidance.
  5. The Motley Fool, Twilio (TWLO) Q3 2026 earnings call transcript, call held 29 October 2026. Used for fee contribution to DBNE, the 1% penetration remark and 2027 outlook.
  6. The Motley Fool, Twilio (TWLO) Q2 2026 earnings call transcript, call held 6 August 2026. Used for Q2 fee contribution, customer cohort figures and neutrality remarks.
  7. The Motley Fool, Twilio (TWLO) Q4 2025 earnings call transcript, 12 February 2026. Used for voice AI revenue growth above 60%.
  8. CX Today, Twilio sees CX demand rise for AI voice and agent identity, 3 November 2026. Used to corroborate named Q3 customers.

Weakest point in the source base: every voice AI figure is spoken, not filed, and the transcripts are third-party with a stated error disclaimer. Financial figures last verified against the filed releases on 11 November 2026.

MJ
Madhur Jain
Writes for Zan Digital about AI product economics, usage-based software and what company filings say once the headline numbers are taken apart.

Related reading