AudioCodes 2025: Services Are Now the Majority, but the New Engine Still Lacks Leverage

Getting To Know The Company

AudioCodes is no longer just a gateway, SBC and IP telephony vendor. It is trying to turn its installed base into a broader voice infrastructure, managed services and Voice AI platform built around Microsoft Teams, Webex, Zoom and contact center workflows. That sounds like a natural shift from hardware to services, but 2025 shows that the transition still looks more expensive than elegant.

What is working right now is fairly clear. Revenue stayed broadly stable, products still grew 2.6%, services already account for 53.2% of revenue, and the company still ended the year with $72.9 million of liquid assets and no classic refinancing drama. At the same time, management pushed capital returns aggressively: $30.6 million of buybacks and $10.9 million of dividends in 2025, followed by another $13.3 million of buybacks and a $5.3 million dividend after the balance-sheet date.

But a superficial read can miss the active bottleneck. Services are growing, leverage still is not. Operating margin fell to 5.7% from 7.1%, service gross margin slipped to 68.0% from 69.6%, selling and marketing expense jumped 8.5%, and revenue per employee edged down despite higher headcount. This is not the profile of a business already harvesting a cleaner software model. It is the profile of a company financing a strategic transition before the new layer has proven better economics.

The gap between the narrative and the revenue base matters. The company talks about Live Platform, Voca CIC, Meeting Insights, Operator Connect and Webex Cloud Connect, yet it explicitly says that revenue is still generated primarily from on-premises deployments. So this is not a clean SaaS story that has already crossed over. It is an incumbent voice business trying to pull customers from infrastructure into cloud delivery, managed services and AI applications while the legacy engine still carries a meaningful share of the result.

AudioCodes’ 2025 economic map looks like this:

Events And Triggers

The first trigger: Live Platform got a real commercial upgrade in 2025. AudioCodes completed certifications as an approved provider of Teams Operator Connect and Webex Cloud Connect, in addition to its Microsoft Operator Connect Accelerator, Webex Cloud Connect Enablement Partner and Zoom Provider Exchange Accelerator status. That matters because it pushes the company from being just another component or SBC vendor toward becoming a service-enablement layer for partners.

The second trigger: the company also launched a global private peering network designed to let operators, partners and service providers connect local networks and telephony services directly into the AudioCodes cloud. This is not a minor product announcement. It is an attempt to build recurring distribution and service rails around Teams, Webex and Zoom rather than sell a one-off box.

The third trigger: 2025 also saw stronger demand for analog gateways in North America because of the FCC-driven PSTN shutdown. That is a genuine near-term tailwind, especially in sensitive use cases such as fire alarms and elevator intercoms that still need to remain analog. But it needs to be read correctly. This is not a new structural growth engine. It is transition-driven demand supporting legacy products while the broader market migrates.

The fourth trigger: the company expanded its Voice AI and productivity layer in 2025 through Meeting Insights, Voca CIC, Interaction Insights and real-time AI analytics capabilities inside Live Platform. That sharpens the strategy, but it still does not provide enough proof at the consolidated P&L level.

The fifth trigger: the capital-markets message was unusually strong. AudioCodes repurchased 3.15 million shares for about $30.6 million in 2025, paid $10.9 million of dividends, and then declared another dividend of about $5.3 million in February 2026. After year-end it repurchased another 1.695 million shares for $13.3 million. That signals confidence, but it also underscores that the balance sheet is being used to improve the per-share story while the new engine is still proving itself.

Efficiency, Profitability And Competition

The central insight is that the shift toward services is real, but for now it is adding operating layers faster than it is adding leverage.

Services are growing, but the quality of profit is not improving with them

Total revenue rose only 1.4% to $245.6 million. That is not a bad number, but it is too soft for a story that wants to lean on cloud transition, Voice AI and managed services. The growth came from two very different buckets. Products rose 2.6% to $114.9 million, mainly on stable demand for core connectivity products such as gateways, SBCs and IP phones, with some incremental pull-through from managed services engagements. Services increased only 0.4% to $130.7 million.

One of the more important gaps in the filing sits here. Product support declined by about 5%, mainly because customers are relying less on legacy telephony equipment and moving toward cloud-based end-to-end service models rather than ongoing maintenance of on-premises systems. In other words, the old recurring layer is fading. What kept services broadly stable was growth in professional services and managed services, mainly through AudioCodes Live and enterprise UC demand around Microsoft Teams.

That distinction matters because not every service dollar has the same economics. A decline in maintenance revenue replaced by implementation hours and managed delivery does not automatically create a better business. The 2025 numbers show exactly that. Product gross margin improved to 61.5% from 60.3% even with tariff pressure, while service gross margin fell to 68.0% from 69.6% because of higher support and delivery personnel costs.

The cost of the strategic pivot is already visible in the P&L

Selling and marketing expense rose 8.5% to $77.2 million, or 31.4% of revenue versus 29.4% a year earlier. R&D expense edged up to $52.6 million, mainly because of higher cloud expenses. At the same time, the company received almost no IIA support in 2025. So AudioCodes is effectively self-funding the push into the newer stack.

That also shows up in headcount. Total employees increased to 981 from 946. Most of the increase came from sales, service and support, which rose to 514 from 494, and from R&D, which rose to 331 from 320. Revenue per employee slipped to roughly $250 thousand from roughly $256 thousand in 2024. That is not a crisis, but it is another sign that the company is layering in cost before it has proven a stronger growth slope.

Net income weakened, but not all of the decline was operational

Operating income fell 18.3% to $14.0 million, and operating margin dropped to 5.7%. That already matters. But net income fell much harder, to just $9.0 million from $15.3 million in 2024, and this is where a straight headline read becomes misleading.

Part of that swing is tax-driven. In 2024 the company benefited from a tax release, including the release of part of a valuation allowance, while in 2025 tax expense climbed to $4.6 million because of lower deferred tax assets and higher taxable income in certain subsidiaries. So a reader looking only at the collapse in net income will overstate the operational damage. This is not a story of operating implosion. It is a story of a slow-growth business spending more on selling and services while also facing a less favorable tax year.

Even after making that adjustment, the picture is still not clean. General and administrative expense actually fell 10.8%, mainly because 2024 included a one-time settlement tied to the termination of the previous headquarters lease. So 2025 benefited from an easier comparison there, and operating income still fell. That is the core issue. The new engine has not yet produced the leverage the company wants investors to believe is coming.

Competition gets tougher precisely where the company wants to grow

AudioCodes is not escaping competition. It is moving into it. In the newer layers of the business it is up against Twilio, Vonage, Telnyx, Vapi.ai and Retell.ai in voice AI connectivity and orchestration. In Meeting Insights it faces Microsoft Copilot for Teams, Zoom AI Companion, Avoma and Otter. In contact center and compliance recording it runs into NICE, Genesys, Five9, Verint and others.

The economic meaning is straightforward. The company is moving from edge hardware into categories that require more selling effort, more integration, more partnerships and more proof of value. Strategic partners, especially Microsoft, Cisco, Zoom and Genesys, are both distribution channels and points of dependency. That is not automatically negative, but it does explain why 2025 looks like a build year rather than a harvest year.

Cash Flow, Balance Sheet And Capital Structure

This is not a leverage-stress story, but it is a story of very aggressive cash use.

Cash flow

Cash flow from operating activities came in at $29.4 million, down from $35.3 million in 2024. The gap versus net income is meaningful, but most of it does not point to a problem. On the positive side, there were non-cash items such as $6.5 million of share-based compensation, $4.2 million of depreciation and amortization, and changes in derivatives and deferred taxes. On the less favorable side, working capital was not as clean: trade receivables increased by $8.7 million, mainly because several contracts were signed close to year-end, and other receivables and prepaid expenses rose by $6.1 million.

Relief came from inventory, which declined by $9.3 million, but that also needs to be read carefully. The company still carries $22.0 million of inventory, and it explicitly says the model sometimes requires holding inventory ahead of demand in order to meet short lead-time delivery requirements. It wrote off about $3.3 million of inventory in 2025 after $4.8 million in 2024. So the inventory unwind helps cash, but it sits inside a business that still lives with ongoing obsolescence risk.

The real cash picture

It is important to define the framing clearly. I am using an all-in cash flexibility view here. That means starting from reported operating cash flow, which already includes $7.1 million of lease-related operating cash, and then checking what remains after CAPEX, dividends and buybacks.

On that basis, AudioCodes generated $29.4 million of operating cash, spent $6.5 million on property and equipment, paid $10.9 million of dividends and used $30.6 million on buybacks. The result is about negative $18.6 million after all those cash uses. That almost perfectly explains the drop in liquid assets from $90.9 million to $72.9 million.

Without buybacks and dividends, the picture is much more comfortable: about $22.9 million remained after CAPEX. So the point is not that the business is weak on cash generation. The point is that management is choosing to front-load shareholder returns while the operating model is still proving whether the cloud and Voice AI transition can generate cleaner growth and margins.

No classic debt problem, but there are obligations to keep in mind

The balance sheet is relatively clean of classic financial debt, and that is a real advantage. But that does not mean there are no claims on cash. The company carries $38.0 million of lease liabilities, with $7.1 million of operating lease cash paid in 2025. It also has non-cancelable purchase obligations of about $16.2 million for excess inventory items and another $4.8 million for SaaS subscription services expected to be used through 2026 and 2027. This is not distress, but it is also not completely free cash.

On the positive side, the deferred-revenue layer does give some support. Deferred revenues stood at $57.9 million, and the company disclosed remaining performance obligations of $38.2 million for 2026, $9.8 million for 2027 and $9.9 million thereafter. That shows there is a real contractual base. At the same time, the company also says backlog is limited, which means this visibility still covers only part of annual revenue. In other words, there is some contractual depth, but not yet the kind of deep subscription visibility that would justify a much cleaner recurring-revenue read.

Capital allocation is now part of the thesis

The number of shares outstanding fell to 27.09 million from 29.68 million, an 8.7% decline in a single year. That is material. Management is effectively telling the market that even if growth is not yet strong enough, it can improve the per-share story through sustained buybacks.

That can work as long as two conditions hold. First, the legacy revenue base remains stable enough. Second, the new services layer does not consume the old cash advantage. For now, both conditions hold only partially, not fully. So capital returns strengthen the thesis, but they do not replace the thesis.

Outlook

Before looking ahead, four less-obvious findings are worth locking in:

1. Services are already the majority, but their margin moved in the wrong direction.
2. 2025 still got help from legacy products, especially analog gateways in North America, not only from the newer engines.
3. The collapse in net income overstates the operational deterioration because of tax noise, but even the cleaner operating view is not strong enough.
4. The company has liquidity and some contractual revenue support, but not enough to call this a deep, proven recurring model.

2026 looks like a proof year

The company does not provide explicit numerical guidance, but the direction is clear. It expects the UCaaS shift to continue in 2026 and beyond, expects AudioCodes Live to keep gaining traction, and frames Voice AI applications as a new growth engine for the near and long term. That sounds promising, but the real test is not the number of strategic announcements. The real test is what starts to flow through the income statement.

For the story to get cleaner, three things need to happen almost together. First, faster growth in managed services and applications such as Voca CIC and Meeting Insights. Second, better quality of growth, meaning more recurring revenue and less dependence on labor-heavy service delivery. Third, an operating margin that starts moving up again. Without that, the transition remains a mix story rather than a superior-economics story.

What can still weigh even if revenue moves higher

The company explicitly warns that the shift to SaaS and pay-per-use can pressure short-term revenue recognition. That matters. Not every strategic win will hit the P&L in the same way a hardware or license sale used to. So it is entirely possible to see the strategic narrative improve before the numbers fully catch up.

A second pressure point is competition. In the categories where AudioCodes wants to grow, it is facing platform owners, communications specialists and contact center and AI vendors with fast product cycles. So even if demand is there, the cost to sell and support the new stack can stay elevated.

A third pressure point is FX. In 2025 the NIS appreciated sharply against the dollar, and the company explicitly says that this increased the dollar cost of its Israeli operations. It does run a hedging program, with forward and option contracts carrying a $45.0 million notional amount and a $7.47 million balance-sheet asset at year-end, but that only reduces volatility. It does not eliminate the structural cost exposure. If the NIS remains strong, the cost base will keep pressing on margins.

What needs to happen over the next 2 to 4 quarters

The certifications and platforms need to become numbers. That means more recurring revenue from Live Platform, more contribution from Teams, Webex and Zoom-related bundles, and more proof that Voice AI is adding real economic value rather than just strategic color. Trade receivables also need to behave better after the year-end build. And if legacy products are currently benefiting from a PSTN migration window, the company needs to show that it can use that cash support to fund the transition rather than postpone it.

Risks

The central risk is that the company stays stuck in the middle. On one side, the legacy product base continues to erode with the cloud shift. On the other side, the new services and application layers do not grow fast enough, or profitably enough, to replace it. That is a dangerous setup because it can look attractive in strategy language while remaining underwhelming in the numbers.

A second risk is channel concentration. Westcon alone accounts for 13.8% of revenue, ScanSource for another 9.3%, and the top five customers together account for 36.1%. Beyond that, the company says it sells mostly through intermediaries, which means less direct visibility into end-user needs. That channel is both a moat and a real dependency.

A third risk is working-capital quality. The company relies on inventory to meet fast delivery needs, explicitly says backlog is limited, and still recorded $3.3 million of inventory write-downs in 2025. Together with more than $21 million of non-cancelable purchase commitments, that means the demand backdrop and the technology transition both need to remain supportive for inventory not to become a heavier burden again.

A fourth risk is the Israeli cost base. 511 of 981 employees are in Israel, and the company explicitly states that a significant portion of its personnel and facilities costs are incurred in NIS. That already hurt in 2025. On top of that, the regional security situation and reserve-duty exposure remain operational risks even though facilities and supply chains have not yet seen a material disruption.

A fifth risk is that management keeps leaning on the balance sheet to support shareholder yield before the new engine is ready to carry the story on its own. That is not necessarily the wrong decision, but if growth stays muted for several more quarters, the market may start reading the buyback less as confidence and more as a shortage of high-return reinvestment opportunities.

Conclusions

AudioCodes enters 2026 as a company that understands where the market is going, but has not yet proven that it can monetize the new direction better than it monetized the old one. Services are already larger than products, Voice AI sits more centrally in the strategy, and Live Platform got meaningful commercial validation. But leverage is deteriorating, selling costs are rising, and the aggressive capital-return program is partly covering for the lack of stronger growth.

Current thesis in one line: AudioCodes is a transition company with a comfortable balance sheet and strong capital returns, but 2025 showed that the shift to cloud and Voice AI still has not translated into convincing operating leverage.

What has changed versus the older reading of the business is not only the mix, but the question investors need to ask. The old question was whether the legacy voice infrastructure business was stable. The new question is whether the services and application layer can become a better business, not just a different one.

The strongest counter-thesis is that the market may simply be too impatient. There is a recurring-revenue base here, a solid cash position, no meaningful classic financial debt, a buyback that is reducing the share count quickly, and a product set increasingly aligned with a market that wants voice infrastructure for AI rather than legacy telephony. If management is right, 2025 is only an interim year before a sharper improvement.

What can change the market read in the short to medium term is not another certification or product announcement. It is a combination of faster organic growth, better operating margin, cleaner receivables behavior and evidence that the newer services are not just replacing old support revenue but actually creating higher customer value and higher company economics.

One sentence on why this matters: if AudioCodes can turn this transition into an economic transition rather than just a technological one, it can move from being read as a mature infrastructure vendor toward being read as a voice services and AI platform. If it cannot, shareholders are left with dividends and buybacks but not a clean growth engine.

Over the next 2 to 4 quarters the thesis gets stronger if Live Platform, Voca CIC and Meeting Insights keep growing without further pressuring service delivery and selling efficiency, and if recurring revenue starts to look deeper than it does today. It weakens if the legacy product base loses pace faster, if services keep requiring more people to stand still, or if buybacks remain the only clearly visible positive headline.