AudioCodes reported second quarter results this morning, alongside a semi-annual dividend of 20 cents per share.
The question we set in advance was singular: is Voice AI visible in the revenue line yet. The answer is complicated.
What was reported
| Quarter | Year ago | Change | |
|---|---|---|---|
| Revenue | 63.0 | 61.1 | +3.1% |
| Of which: services | 34.6 | 32.6 | +6.2% |
| Of which: products | 28.4 | 28.5 | -0.4% |
| Gross margin | 65.7% | 64.1% | +160bp |
| Operating income | 3.2 | 2.6 | +23% |
| Non-GAAP operating income | 4.6 | 4.4 | +4.5% |
| Non-GAAP EPS | $0.15 | $0.14 | - |
In millions of dollars
The gap that explains everything
The company reports that its conversational AI business grew more than 50% year over year, and that it booked record orders for Voice AI Connect and Live.
And total revenue rose 3.1%.
And why the two numbers cannot be bridged
AudioCodes publishes the growth rate of conversational AI - but not the absolute figure. Neither this year nor last.
The practical consequence: there is no way to compute what it contributed. 50% growth on a small base and 50% growth on a meaningful base are entirely different things, and the report does not allow the reader to tell them apart.
What can be inferred from the arithmetic: if a line growing more than 50% sits inside a company growing 3.1% overall, it is still small enough not to move the picture.
And the products-versus-services split sharpens it: all the growth came from services, up 6.2% to $34.6 million. Product sales actually declined slightly - from $28.5 million to $28.4 million.
What is improving
Margins. Gross margin rose to 65.7% from 64.1%, and operating margin to 5.1% from 4.3%.
And annual recurring revenue reached $84 million, up 20% - far faster than the company overall, and a better gauge of the direction the business is moving.
The Microsoft Teams business grew 5%.
The point demanding attention
EPS rose - and profit fell
Non-GAAP net income fell to $3.9 million, from $4.1 million a year ago.
And non-GAAP EPS rose, to $0.15 from $0.14.
The reason is not profitability but a reduced share count. During the quarter the company repurchased 950,133 shares for $8.9 million, and diluted shares fell to 24.6 million.
This is not deception - it is the legitimate arithmetic of a buyback. But anyone reading only the EPS line sees an improvement that did not exist in profit itself.
And cash: $64.2 million at 30 June, down $11.5 million from end-2025. Operating cash flow in the quarter was $6.1 million. $17.4 million remains under the buyback authorisation.
The dividend
$0.20 per share, roughly $4.8 million in total. Record date 19 August, payment 3 September. Israeli withholding tax of 25%, and 30% for holders of 10% or more.
That is exactly the same amount distributed with last year's second-quarter report - the dividend was not increased.
What was not in the release
The results release contains no updated full-year guidance. The standing guidance, reaffirmed with the first-quarter report on 5 May, is revenue of $247-255 million and non-GAAP EPS of $0.60-0.75.
The conference call was scheduled for 15:30 Israel time - where an update may be given.
My Angle
A personal opinion of Ilan Abramov - not advice, not a recommendation
AudioCodes is a good test case for the question that has run through this whole earnings season: who actually pays for AI.
The answer here is "someone, but not enough". A line growing more than 50% is real, and recurring revenue up 20% to $84 million confirms the transition is happening. But a company growing 3.1% overall is telling you that line is still small.
And what bothers me is not the size but the disclosure. If conversational AI is the story, why not publish the number? A company that publishes a growth rate without a base denies the reader precisely the calculation that would let them judge it.
And two technical points worth remembering: EPS rose while profit fell - that is a buyback, not an improvement. And cash fell $11.5 million in half a year at a company paying a dividend and buying back stock at the same time.






