Five9 is converting a contact center software franchise into an AI-native customer experience platform. The second quarter of 2026 is the first print where the revenue engine and the cost base visibly disagree with each other, and the rest of this report works out which side of that disagreement the equity is being asked to fund.
Subscription revenue grew at a double-digit clip, the third straight quarter of acceleration, while adjusted gross margin slipped a point and a half from a year earlier. The spread between top-line momentum and cost absorption is now the central variable.
A nine-figure new-customer contract closed through the Google Marketplace in the same quarter as the $200 million buyback authorization. The question the next two quarters resolve is whether the AI revenue mix lifts growth without re-pricing the cost of serving a customer.
Five9 competes in the cloud contact center software, or CCaaS, market, where the relevant peer set includes Genesys, NICE CXone, Verint, and the bundled contact center suites inside the hyperscaler cloud marketplaces. The structural shift the company is navigating runs deeper than vendor competition. As generative AI absorbs a growing share of routine customer interactions, the unit of economic value in a contact center is migrating from the licensed human seat to the consumed AI interaction. The risk factors in the quarterly filing state the tension plainly: if AI solutions perform an increasing proportion of interactions, the company is expected to replace decreasing license revenue with consumption-based AI revenue at a pace that supports the historical growth rate.
The company addresses this through a revenue-commitment model that moves away from pure per-seat licensing toward contracted spend across human and AI channels. More than three thousand organizations subscribe to the platform, and no single customer exceeds a tenth of revenue, which keeps the growth math from depending on any one relationship. The strategic posture is to be the orchestration layer where every interaction, human or machine, routes through the same system of record, so that each conversation strengthens the next through accumulated data.
The peer comparison matters because the larger CCaaS incumbents carry the same seat-to-AI transition but with heavier legacy install bases. Five9's smaller scale and its partnership posture, including the expanded Google Cloud Enterprise CX AI solution announced in January 2026, position it as the platform that enterprises buy through the cloud procurement channels they already use. That positioning converts the hyperscaler relationship from a distribution convenience into a commercial wedge against both Genesys and the hyperscalers' own native contact center offerings.
The geographic mix reinforces the enterprise focus. United States customers accounted for roughly eight tenths of quarterly revenue, with international contributing the remainder from a base that grew faster year over year. The international expansion carries macro and tariff exposures that the filing flags as an open risk, but the revenue concentration in domestic enterprise accounts keeps the near-term growth profile tied to U.S. contact center budgets, which are the segment most directly affected by the AI substitution economics.
The product story of the quarter is the launch of Five9 Voice AI Agents, a capability for human-like, real-time voice conversations with enterprise-grade governance and seamless handoff between AI and human agents. The product matters because voice has historically been the most labor-intensive and least automatable channel in customer service. Automating voice at enterprise scale converts the largest remaining pool of human agent minutes into consumption-priced AI usage, which is exactly the revenue mix shift the company needs to offset seat-based license erosion.
The moat argument rests on three compounding assets. First, the platform is the system of record for full conversation history, which means the AI agents inherit context that standalone virtual assistant products built outside the contact center flow cannot. Second, the Genius AI suite is priced on a consumption basis, so the more an enterprise deploys the product, the larger the switching cost becomes because usage data and workflow integrations accumulate. Third, the partnership network of more than 1,450 technology providers extends the platform into integrations that a single-vendor build cannot match.
The data flywheel framing is the forward-looking claim that needs evidence. Each interaction, AI or human, feeds the same orchestration engine, which improves routing, personalization, and automation accuracy over time. The falsification test is whether retention and net revenue expansion among the AI-adopting subset of customers outpaces the base. The trailing twelve-month subscription dollar-based retention rate stood at 107 percent as of the June quarter. A year earlier it had read 109 percent, and management attributes the decline to a single large new customer ramping through 2024.
The concentration of the decline in one account is a useful detail because it suggests the base erosion is not broad-based license shrinkage but a timing artifact that should fade as that account's ramp matures. If the next two retention prints return to the high end of the historical range, the argument that AI adoption is destroying net revenue within existing accounts loses its empirical footing. If retention continues to drift, the revenue-commitment model protects reported revenue but conceals a deeper usage decline that would surface at renewal time, when committed-spend customers reassess their targets against actual interaction volumes.
The income statement for the quarter shows revenue of $312.4 million, up ten percent from a year earlier. Cost of revenue rose faster, at a fourteen percent pace, which is the single most important line in the print. The gross profit line absorbed the difference, leaving GAAP gross margin at 53.4 percent, down from the prior-year quarter. Operating expenses of $164.8 million put the company just into positive GAAP operating income of $2.0 million, a swing from an operating loss a year earlier that the smaller base flatters. GAAP net income of $3.4 million, or $0.04 per diluted share, is a thin margin of revenue. The year-ago quarter produced a smaller GAAP net income figure, but that result had been lifted by a smaller depreciation base and had not yet felt the full cost of the AI capacity build now hitting the income statement.
The adjusted EBITDA line carries the real story. Adjusted EBITDA of $70.1 million, or 22.4 percent of revenue, is up only a hair from the prior-year figure. The prior-year adjusted EBITDA margin was 24.0 percent, so the current print shows a one-and-a-half-point compression on a dollar figure that grew only three percent on ten percent revenue growth. The gap between the top-line momentum and the margin print is the single most important feature of the quarter.
The margin gap is the quarter's structural signal. Roughly seventy basis points of the EBITDA compression traces to the AI and capacity build, while the remainder reflects the absence of year-ago restructuring charges that had depressed the base period. The filing itemizes the drivers of the higher cost of revenue, with third-party telecom costs rising on customer activity and depreciation, data center spend, and software amortization carrying the rest of the increase. Personnel costs fell by $2.8 million, the clearest trace of the 2025 reduction-in-force plan that cut the global workforce by about four percent. The decline shows the labor reset now landing in the expense base rather than in the headcount line.
Stock-based compensation for the quarter came down from $41.9 million a year earlier, confirming the same reset. The balance sheet funds the transition without external pressure. Cash and marketable investments total $654.1 million, against roughly $737 million of convertible senior notes. The coupon on that note keeps quarterly interest expense near $3.5 million, and the company's stated position is that existing cash is sufficient for at least the next twelve months of working capital and capital expenditure needs. The cash story rounds out the quarter. First-half operating cash flow of $106.0 million, up from $83.4 million a year earlier, is the cleanest evidence that the business funds its own AI build. The cash flow covers the capital expenditure and capitalized software development of the period with room to spare. The quarter's $8.4 million headquarters consolidation impairment, booked in general and administrative expense, is a one-time non-cash charge that marks the company's footprint right-sizing as part of the same productivity program that drove the earlier headcount reduction.
The third-quarter revenue guide of $316.0 million to $322.0 million implies a sequential increase over the second quarter and continued double-digit annual growth at the midpoint. The full-year 2026 guide, roughly $1.27 billion at the midpoint, sits at the upper end of the prior outlook. The non-GAAP diluted EPS range for the full year, about $3.25 at the midpoint, implies the second half is expected to produce meaningfully better margins than the first half, because the first half already delivered non-GAAP net income of $112.1 million. The execution risk is concentrated in the cost side of that math.
The June executive appointments are the other load-bearing event of the quarter. Niranjan Vijayaragavan arrived as Chief Technology Officer from Nintex, Rob Hornish took the Chief Sales Officer role, and Sven Linsmaier joined as Executive Vice President of Transformation and Strategy. The CTO appointment carries the most thesis weight because the R&D organization was restructured around it during the quarter, consolidating product engineering, AI automation, and architecture under one leadership structure. A product organization re-platformed around a new CTO mid-year is a classic delivery-risk setup, and the falsifiable signal is whether the Voice AI Agents roadmap cadence holds through the second half.
The Google Marketplace win is the commercial proof point that the hyperscaler channel is not merely a listing but a procurement mechanism. The nine-figure total contract value agreement closed through that channel with a new Fortune 100 financial services customer. The deal matters for three reasons. The customer class is the top-of-funnel segment where per-interaction volume is highest, the channel removes the enterprise security and procurement friction that historically slowed Five9's large-account sales cycles, and the deal structure under the revenue-commitment model converts a single seat-based negotiation into a multi-year consumption commitment. For shareholders, the consequence is a visible anchor for the second-half subscription growth trajectory that management has been citing as accelerating for three consecutive quarters.
The guidance cadence itself is a monitoring variable. The full-year revenue guide implies roughly $635 million for the second half against $617.8 million for the first half, a sequential step-up that the Q3 guide supports. If the Q3 print lands at the low end and the full-year guide moves in October, the margin story gets a second look. If it lands at the high end with reaffirmation, the AI revenue mix argument gains a full quarter of evidence behind it.
The principal downside scenario is that the AI revenue substitution arrives slower than the license revenue erosion. If seat-based licenses continue to decline because customers redirect headcount budgets toward AI consumption, and the consumption pricing does not clear at the same revenue per interaction, the top line could stall in the mid-single digits even as the product mix looks healthier. The data signal is the subscription dollar-based retention rate, which still sits above a full hundred percent. The two-point year-over-year decline, even if concentrated in one large account, is the first crack in a metric that had hovered near a high of 109 percent.
The securities class action filed in December 2024 in the Northern District of California is the litigation overhang. The court granted in part and denied in part the motion to dismiss on February 23, 2026, which means the amended complaint survives in material part and the case proceeds toward discovery. The related derivative actions in California and Delaware were stayed and consolidated pending the securities case. The company disclosed legal fees of $854,000 in the quarter related to the action and cannot estimate a reasonably possible loss range. For a company with $654 million of cash and marketable securities, a settlement would be absorbable, but the discovery timeline extends into 2027 and the governance claims against the board keep the matter in the diligence files of institutional investors regardless of the eventual dollar outcome.
The convertible note maturity in March 2029 is the capital structure event to watch, because it is the date on which the conversion-versus-cash question resolves. The $747.5 million of 1.00 percent notes carry a conversion structure that depends on the stock price relative to the conversion price at maturity. The probability of a cash settlement has declined, because the share price now sits at 52-week highs well above the levels at which the 2026 accelerated share repurchase programs were struck. A cash settlement of roughly $747 million would consume more than the current cash and marketable securities balance unless the company draws on operating cash flow over the intervening years. The resolution of this question is the single most important capital-structure event on the company's near-term calendar.
The company generated $106.0 million of operating cash flow in the first half of 2026, a pace that would make a cash refinancing feasible if margins hold. The scenario to avoid is the combination of margin compression, a weak 2027, and a conversion price that sits above the market at maturity. The customer concentration risk is bounded but present. While no single customer exceeds a tenth of revenue, the retention-rate attribution to a single large new customer's ramp pattern shows that one account can move the headline metric by two points. The revenue-commitment model amplifies this: a committed-spend customer that under-delivers on its own end-user demand still pays, which protects revenue, but a committed-spend customer that renegotiates downward at renewal would show up as a retention miss in the trailing twelve-month window. The hyperscaler exposure cuts both ways in the same way: the Google Marketplace channel that opened the Fortune 100 door also positions the hyperscaler's own native voice AI as a future substitute at the same consumption price point.
At the early-September share price near $34, the market capitalization is roughly $2.55 billion. That valuation sits against full-year revenue of about $1.27 billion at the guide midpoint, a multiple near 2.0x revenue. On the full-year non-GAAP net income midpoint of approximately $248 million, the multiple is about 10.3x. The first-half adjusted EBITDA run rate of $144.5 million implies an annualized EBITDA multiple around 8.9x. These multiples sit at a level that reflects the AI transition still in progress, with the cost side of the equation yet to be proven.
The bear case values the equity on the GAAP print and the margin trajectory. GAAP net income of $21.8 million for the first half anchors the low end. The adjusted EBITDA margin compressed 160 basis points year over year in the second quarter. A de-rating toward 7x EBITDA at a stabilized level near $285 million annually implies an enterprise value near $2.0 billion before cash. After the 2029 convertible notes, the equity value in the bear case comes in closer to $1.4 billion. That is the value the margin reset would support if the AI build never pays back in operating leverage.
The base case holds the multiple at 9x a normalized adjusted EBITDA of roughly $290 million, supported by the revenue-commitment model stabilizing the large-account base and the AI consumption revenue reaching the high teens of subscription revenue. At 9x, the enterprise value is approximately $2.6 billion, and with net cash the equity value is near $2.0 billion. The base case is the scenario in which the margin compression of the first half of 2026 is the trough, driven by the capacity build and the AI deployment cost base, and the second half shows operating leverage as the fixed infrastructure spend stops scaling with each additional AI interaction. The base case is where the equity sits today, pricing in a transition that is real but not yet fully proven.
The bull case prices the equity at 12x a normalized EBITDA that re-expands toward $315 million. That mix reaches 20 percent of subscription revenue and retention recovers to 108 percent. At 12x, the enterprise value is near $3.8 billion. The equity value approaches $3.1 billion, roughly 35 percent above the early-September share price. The counterargument to the bull case is that the hyperscaler partnership that opens the enterprise procurement channel also exposes the company to the hyperscalers' own native AI voice offerings, which would compete at the margin on the same consumption pricing model. The capital allocation program is the valuation floor. The prior repurchase program of $150 million was fully retired, and the board then authorized a new $200 million program with no expiration. The repurchase authorization, combined with the net cash balance, gives the equity a structural downside cushion that the pure multiple analysis does not capture.
The quarter revealed that Five9's AI transition is real on the revenue side and still unproven on the cost side, and the equity now prices in a version of the story where the two eventually reconcile.
The strategic initiatives that the quarter made load-bearing are the Voice AI Agents launch, the nine-figure Google Marketplace contract with a Fortune 100 financial services customer, the revenue-commitment model shift, and the June leadership restructure around the new CTO. Together they describe a company that is re-architecting its commercial model, its product portfolio, and its management team in the same twelve-month window, which is the signature of a deliberate platform transition rather than a reactive product push.
The variables that the next two quarters resolve are the subscription dollar-based retention rate, the adjusted EBITDA margin trajectory as the first-half capacity build stops adding cost with each incremental AI interaction, the Voice AI Agents adoption cadence under the new CTO's restructured R&D organization, the AI share of subscription revenue, the pace of the repurchase program against the current share price, and the discovery timeline of the securities class action now proceeding in the Northern District of California. The 107 percent retention reading is the single number to watch, because it determines whether the AI adoption story is destroying or building net revenue within existing accounts.