October 19, 2026
Tolan just raised $30 million for its AI companion app, which TechCrunch reports has already pulled in more than 250,000 active users. On the same day, we learned that Eugenia Kuyda -- the founder of Replika, arguably the company that proved this category could work -- has built a new startup called Wabi, a free personal AI assistant that fulfills a pitch she first made at Disrupt twelve years ago. Two separate stories, same underlying signal: money is still flowing into AI that exists primarily to build a relationship with you, not to finish a spreadsheet.
The easy reaction is skepticism -- haven't we been here before with chatbot fatigue and app graveyards full of 'companion' products that never found a business model? Fair point. But the persistence of this category matters precisely because it keeps clearing funding bars that plenty of 'serious' enterprise AI startups don't. Investors are implicitly betting that emotional engagement is stickier than productivity, and stickiness is what turns a free download into a paying subscriber. For founders building consumer AI, the lesson isn't 'add a personality to your product.' It's that retention, not novelty, is what gets you your next check. Kuyda waiting twelve years to ship the thing she originally pitched is also a reminder that timing -- not just the idea -- is half the business.
Meanwhile, Tenstorrent closed a $1.5 billion round that pushes its valuation to $6.2 billion, according to TechCrunch, turning down acquisition interest to stay independent as an open-weight alternative to Nvidia's dominant AI chip stack. This is the less glamorous half of the same story. Every companion app, every assistant, every agent needs compute underneath it, and right now that compute market has exactly one dominant vendor. A well-funded challenger betting on open-weight chips is good news for anyone who worries about concentration risk in AI infrastructure -- pricing power, supply constraints, and roadmap decisions all currently sit with one company.
I'd argue this is actually the more consequential headline of the day, even though it will get a fraction of the attention Tolan and Wabi get. Consumer AI apps are easy to understand and fun to write about. Chip independence is boring until the day a supply shock or a pricing change hits every AI product you depend on at once, at which point it's the only thing that matters. Businesses evaluating AI vendors right now should be asking not just 'what does this tool do' but 'what is it built on, and what happens if that changes.' That question applies whether you're picking a consumer app or choosing a platform to run your own operations -- it's part of why we think through platform and security considerations together rather than treating infrastructure as someone else's problem.
If you run a business and none of this seems to apply to you, consider the pattern rather than the products: capital is rewarding AI tools that get deep, habitual usage over tools that just automate a task once. That's a useful lens for internal tooling decisions too. A tool your team opens once a week and resents is a worse investment than one they use daily without thinking about it, even if the second one looks less impressive in a demo. We see this constantly when teams compare rebuilding their stack versus adopting workflow automation that people actually stick with.
Do you think AI companions are a durable category or a venture-funded detour -- and would you trust an open-weight chip challenger over the incumbent if it meant a less mature ecosystem?
Sources