In the world of venture capital a signed term sheet is treated as a word of honour. It is not binding in court, but whoever walks away after signing one is remembered for it. Listen Labs signed such a term sheet for 125 million dollars (around 107 million euros) at a valuation of 1.5 billion - and then simply walked away from it.

The round was to be led by Menlo Ventures. It never closed. The reason, according to people familiar with events, sits at a different address: Salesforce has been in talks to buy the company for around two billion dollars. The talks are not finished and may come to nothing. But they were enough for somebody to calculate that a sale is worth more than the reputation attached to a signed document.

What does Listen Labs do? It runs user interviews with artificial intelligence - it drafts the questions, conducts the conversation over audio or video, and then assembles reports and presentations out of those conversations, almost identical to what marketing agency staff have produced for decades. Its clients include Microsoft, Canva, Anthropic and Sweetgreen. The company is three years old and has around thirty million dollars in annual revenue.

Hold on to that figure, because it is the whole story. Two billion dollars against thirty million in revenue means the buyer is paying sixty-seven times annual turnover. Someone with experience negotiating sales to Salesforce told the source that this is exactly what could kill the deal - the multiple is too steep even for a company that wants to take its competitor's place.

In January the same company was worth five hundred million dollars. In July the competitor Simile raised 200 million at a valuation of two billion - with three times less revenue than Listen Labs. Somebody in that room worked out that if the weaker one is worth two billion, the stronger one cannot stay at one and a half. Eight months, a fourfold difference in price, the same product. That is not business growth, that is growth in expectation.

If the Salesforce talks collapse, investors say Listen Labs will come back to the market asking for two billion or more. Then walking away from the signed term sheet will look like a calculated move rather than a mistake. And if they do not collapse - somebody will pay sixty-seven years of revenue for software that imagines what the customer is thinking. For that kind of money, would it not be cheaper to ask the customer directly?