AI has broken the ARR startup metric

When I worked as an investor in Singapore, one of the strangest investment pitches to cross my desk was from a peer-to-peer fintech lender in Indonesia. Its numbers looked spectacular: It was growing 3x to 4x faster than traditional lenders and almost twice as fast as other fintech players.

It did not take much digging to uncover the uncomfortable truth. Beneath the eye-popping growth rates was an accumulating pile of bad loans and flaws in the firm’s fundamental business model.

Image credit: Timmy Loen

Now, watching AI startups proudly announce similarly “impressive” figures across LinkedIn and Twitter brings a sense of deja vu. Some of this growth is undoubtedly real, driven by strong product-market fit.

However, history teaches us that in moments of hype, it is easy to ignore deeper structural problems. The AI software-as-a-service (SaaS) sector is rapidly approaching such a crossroad, where investors and founders must dig deeper beyond shiny narratives.

Illusions

Annual recurring revenue (ARR), the amount of predictable revenue a company expects to make from subscriptions each year, has long been the holy grail for SaaS businesses. It signals predictable, subscription-driven cash flow, and startup valuations are often anchored around this number.

Public SaaS companies typically trade at around 10x ARR, while private companies can command even higher multiples based on growth expectations. However, the AI-enabled SaaS sector may be rewriting these rules.

See also: He was a pioneer in streaming. Now he says AI will end pop stars

The surge of interest around new AI tools is supercharging adoption curves. Subscriber bases are exploding, and recurring revenue is following suit.

Yet the question remains: Are these stellar ARR numbers comparable in quality to what investors have historically encountered?

Not all ARR is created equal

In traditional SaaS models, ARR has always been subject to different levels of churn – the attrition rate of a company’s customers – depending on the nature of the product.

At one end of the spectrum, you have large enterprise software systems like hospital management tools, which are deeply embedded into operations and rarely churn. At the other end, you have consumer SaaS products like meditation apps, where the use is more discretionary, leading to higher churn.

Image credit: Timmy Loen

The hidden risk

Retention matters more

Building for durability