ARR: the tech world’s favourite three-letter flex. LinkedIn is abuzz with discussion about which company has the highest, and who got there the quickest.
The current record holder is Swedish vibe coding startup Lovable. Last month, founder Anton Osika announced the company had gone from $1m ARR — which stands for annual recurring revenue — to $100m in just eight months. According to the company, that makes it the fastest-growing startup, not just in Europe, but in the world.
Other companies with accelerating ARR growth include Berlin-based N8n, which has said its ARR grew fivefold in 12 months, reportedly hitting $40m ARR in July. Topping off the frenzy, one startup built using Lovable was this week hailed on LinkedIn for hitting $100k ARR just a week after launching.
Despite the hype, a growing number of people are concerned that the industry’s focus on ARR is masking potential problems with churn and low margins.
“You just can’t calculate ARR based on a monthly subscription, without a churn rate,” says Christoph Gerber — founder of SaaS startup Talon.One and delivery platform Lieferando. “People are over indexing on one metric.”
The ARR hype
ARR represents how quickly a company’s subscription base is growing and provides a forecast for how much money a company is expected to generate in the next twelve months. ARR is typically used by investors to value tech companies — but has not been boasted about so commonly as a growth metric before.
The current crop of AI companies have blown previous ARR standards out the water. Back in 2020, investors told Sifted that hitting $100m ARR in five years was the gold standard for software companies. The same year, online events platform Hopin was lauded by VCs as the “fastest-growing company” for hitting $20m in eight months.
Lovable investor Judith Dada, general partner at Visionaries Club, says the company’s rapid ascent to $100m ARR is particularly notable because of its team size.
“They’ve achieved $100m ARR with an incredibly lean team,” she says. “What took traditional SaaS companies hundreds or thousands of employees, Lovable is doing with a team of 45 people.” (The company is actively hiring, with 16 open roles on its website.)
The company’s competitors have also leaned on ARR as a metric to describe their success. In June, US-based vibe coding startup Replit announced it had gone from $5.5m ARR to $100m in five months. In February, US AI coding startup Cursor said it had gone from $1m to $100m in ARR in 12 months (a record then trumped by Lovable’s growth).
Churn
ARR is a useful signal, says Talon.One founder Gerber, but only if put in context alongside other metrics. Churn rate is crucial: just because a company’s MRR (monthly recurring revenue) is high, there’s nothing to say that rate will continue across a year and translate into ARR.
If companies do have high churn, they also end up having to spend more on sales and marketing just to replace lost customers.
It’s a problem angel investor Fabrice Grinda summed up in a recent interview on the 20VC podcast.
“I don’t know if people remember that AI company where everyone for a month basically changed their profile photo using AI […] their MRR (monthly recurring revenue) went from $250k a month to $30m, and [after experiencing] 99% churn a month later they’re back to $500k. And they raised at that moment in time. So these things are riskier than people think they are,” Grinda said.
“ARR used to be just shorthand for the next 12 months of true recurring subscription dollars (not usage, not transactions and definitely not services),” says Jag Singh, managing partner at Angel Invest.
“When the LinkedIn heroes flex ‘hitting $1m ARR in 24 hours’, I often have to dig deeper to ask if that’s cARR (contracted but not yet billed revenue) or just convenient arithmetic based on one hot-week’s run rate multiplied by 52.”
Lovable told Sifted it’s “using industry-standard ARR recognition in alignment with (its) investors’ norms and expectations” — but declined to comment on its churn, or what percentage of its users are on monthly vs. annual subscriptions.
Thomas Cuvelier, investor at RTP and an angel investor in Lovable, says while free or low-paying users like students may churn once their projects end, enterprise users churn less and are the more appealing user base. (Lovable’s enterprise clients currently include Klarna, Hubspot and Photoroom.)
Meanwhile, a source with direct knowledge of Lovable tells Sifted that the company believes those serious, higher value customers won’t churn, even if those paying lower fees do.
Margins
As well as churn, there’s the small issue of margins, Gerber says. Companies might be gaining an impressive number of customers, but they can still be losing cash on each one if the fundamentals don’t make sense.
“If I were to sell gold for 90 cents on the dollar, I would be the biggest gold dealer the world has ever seen, probably within a week. But I would be paying people to buy my product,” says Gerber.
“Investors always take numbers in context. Why have they decoupled the fundamentals all of a sudden?” he questions.
Gerber likens the situation to the speedy grocery craze that gripped the tech world in 2021 and 2022, where VCs poured billions into companies offering to deliver groceries in 15 minutes or less. The companies reported impressive customer acquisition rates, but struggled to reach a positive margin on their deliveries.
The difference with software, Singh points out, is that in software, “the marginal economics are supposed to flip to a very high margin if retention holds.”
Angel investor Cuvelier says AI companies typically need to raise large rounds to fund the users who are trying out their products for free, in the hope that they switch to a paid plan.
Startups like Lovable pay other AI companies to use their models — one of the expenses that will factor into their margins. Lovable uses US company Anthropic’s model, Claude, to power its app and website builder. The source with direct knowledge of the company says Lovable’s margin had improved slightly after Anthropic released a new version of its model, Claude, in May this year.
That said, the person speculated that Lovable’s margin could take a dip after its switch to an agentic model earlier this month. Agentic models — which figure out what needs to be done and carry it out step by step rather than relying on successive prompts — are typically more expensive to run.
Lovable declined to comment on its margins.
The challenge ahead
The question remains whether the AI companies of the moment can turn ARR growth into profitable growth over time.
That means not just adding new customers, but keeping them — and squeezing more money out of them over time. It also means improving margins so that each cent of ARR actually contributes to the company’s profit.
“The challenge with AI companies we’re seeing is that they’re often built on thin margins and low switching costs (the costs businesses incur when switching from one provider to another). The test for the next set of Lovables is whether they evolve into platforms that also own workflows (hosting, deployment, payments) and are not just in the vibe-coding space,” says Singh.
“The other thing to consider is that VCs have definitely learned from the Gorillas and Hopin days, so we’re all competing to give term sheets to teams that (hopefully) know how to adapt just as quickly as they’re growing their revenues.”
With reporting by Sifted senior reporter Kai Nicol-Schwarz
Read the orginal article: https://sifted.eu/articles/ai-bubble-arr-churn/
