AI startups ride the AI wave, gain traction, secure investment, and launch operations. If only it were that simple—some fade into irrelevance before they even get started!
What could be the reason behind these failures? What are the potential solutions? What are investors looking for in an AI startup in 2025? The questions are aplenty. Crystal Huang, general partner at Google Ventures, shed light on these in Google’s ‘Future of AI: Perspectives for Startups 2025’ report.
Boom and Bust Cycle is Going to Continue
Huang said never before has she seen so many companies pick up crazy traction at launch, get to tens of millions of annual revenue, and have people lose interest in a short while. She explained that since the tooling is so widely accessible for building an application this kind of cycle will continue.
“It’s difficult and costly to build foundation models so not many teams will be able to do it, but at the application layer, I think there’s going to be tons of disruption and rebirth,” she stated.
Huang further added, “It’s exciting that the 2025 landscape will look nothing like last year, and while generative AI is obviously an exciting territory for investors, the standard valuation framework still applies,” Huang added.
Stickiness as a Metric
Huang noted that she will be looking for products that are stickier. “If your product is easy to implement, it’s just as easy to uninstall. Products need to be stickier to create lasting value, which means being both indispensable and deeply integrated into the user’s workflow,” she explained.
Even though she sees all the urgency in AI, she believes that it often requires mutual effort between platforms and enterprises to create automation or workflows, and gaining difficult access to enterprise data to significantly boost performance.
Huang also highlighted that the fastest growth is observed in individuals willing to experiment with something new for $15 a month over an enterprise investing $20 million a year in legacy systems.
There have been talks about companies like Duolingo, which took the gamification route to add stickiness to their product. Startups can learn from such examples to add value to their products.
Hyperpersonalisation is the Future
Huang highlights a crucial shift in the AI landscape: the journey toward hyperpersonalisation.
While the promise of AI-driven, tailored experiences—from marketing to healthcare—has long been touted, its widespread adoption has been hindered by excessive costs. However, this is rapidly changing.
“Training expenses are dropping as smaller and more domain-specific models emerging, and inference costs are plummeting across the board,” Huang pointed out. This cost reduction is enabling personalised AI applications, which were once economically unfeasible, further helping AI companies cultivate user loyalty.
Huang emphasises the increasing sophistication of enterprise clients in AI. CIOs and CTOs are no longer swayed by mere novelty; they demand demonstrable ROI and a clear competitive edge. The commoditisation of certain AI capabilities means that companies must continually innovate and adapt.
This dynamic environment necessitates AI startups to move beyond simply securing funding and focus on building robust, revenue-generating products with defensible moats.
Matthieu Rouif, co-founder and CEO of Photoroom, added in the report, “AI will understand and adapt to human emotion. It will get better at understanding what triggers emotions in humans, allowing for stories and content to be personalised and adapted to individual emotional responses.”
David Friedberg, CEO of Ohalo Genetics, states that AI will also change the media landscape with personalised movies and video games with content generated on the fly, altering the value proposition.
Taking into account the views of some influential industry leaders, AI startups can benefit from having a clear understanding of what to focus on next to really connect with their audience and make their product shine.