Chi-Hua Chien: AI winners won't sell AI — they'll master personalization

At a glance:

• Venture capitalist Chi-Hua Chien predicts the AI model gap between cloud and mobile will shrink from 2 years to 3 months within 12 months.
• He argues AI infrastructure companies are being commoditized while applications capture 88% of new market value, citing $3.1T web vs $400B infrastructure.
• Personalization, not raw AI capability, will separate winners, with companies like Midi Health and Fever already hitting $100M+ ARR using AI as an enabling layer.

The commoditization thesis gains momentum

Chi-Hua Chien has spent over two decades as a venture capitalist, but his approach resembles that of a cultural anthropologist. As co-founder of Goodwater Capital, he focuses exclusively on consumer and prosumer technology, with investments spanning entertainment, healthcare, fintech, and live experiences. His portfolio includes companies like MIDI Health, Fever, and Monzo, and in his early days at Accel, he was the associate who first identified a six-person Harvard startup called The Facebook.

This background in reading human behavior at scale informs his contrarian view: the biggest AI winners won't be the companies selling AI models or infrastructure. Instead, he believes applications that master personalization will capture the value. His argument rests on historical patterns. Looking at the PC, web, and mobile cycles, he notes that infrastructure market caps peaked in 2000 and have since failed to surpass that high in nominal terms. In contrast, application companies generated $3.1 trillion in new market cap during the web era—88% of the total—while infrastructure contributed just $400 billion. The mobile era showed a similar split: infrastructure produced $700 billion, while applications generated $3.7 trillion.

This dynamic is already playing out in AI. Google's recent decision to cut its subscription AI service from $7.99 to $4.99 a month while doubling storage signals the arrival of price competition. Companies with structural advantages in vertical integration and distribution, like Google, can bundle and compete for the average consumer in ways that pure-play AI infrastructure firms cannot.

Personalization as the new battleground

Chien emphasizes that hyper-personalization is the key differentiator for the next wave of AI winners. Done correctly, personalization drives higher customer satisfaction, deeper engagement, and increased average revenue per user (ARPU) over time. His portfolio reflects this strategy. Companies like Triumph, Ritten, and Flow GPT—entertainment firms—are achieving $100 million to $600 million in annual recurring revenue at strong margins, not because they sell AI, but because AI makes their experiences more customizable and personalized.

In healthcare, Midi Health—a women's health company in his portfolio—uses AI to address a critical supply constraint. Hormone replacement therapy for perimenopausal women requires specialized providers, which are in short supply. AI enables Midi to expand care delivery to hundreds of thousands of patients cost-effectively, turning a human bottleneck into a scalable service.

Chien sees this model playing out across any domain where human expertise is the limiting factor. The question isn't whether AI can match frontier models, but how it can be deployed to make products more personally relevant.

The local AI revolution accelerates

The gap between frontier AI models and what can run locally on consumer devices is closing rapidly. Two years ago, the difference was 18-24 months: cloud models were that much more capable than phone-local versions. Today, that lag has shrunk to six months. Chien predicts it will narrow further to three months by this time next year.

This compression enables a new class of applications that blend cloud and local capabilities seamlessly. Users can now run AI models on their phones that match the quality of cloud models from six months ago, opening possibilities for real-time personalization without constant connectivity.

However, the use cases for this capability remain underdefined. Just as the iPhone launched in 2007 before entrepreneurs understood mobile-first applications, today's AI tools are waiting for creators to discover what's truly possible. LLMs, Chien argues, essentially do two things: they enable processing and sense-making of large contexts, and they allow cost-effective personalization at the individual level with feedback loops that improve products over time.

Why super apps fail in the west

Facebook's long struggle to build a super app offers lessons. The company tried multiple approaches—Facebook Credits (2009), Facebook Pay, Libra—but never succeeded. Chien attributes this to a fundamental trust gap between entertainment/social products and financial services, particularly in Western markets.

He explains that social media thrives on triviality and high time investment with low monetization, while financial services demand seriousness, high monetization, and low time commitment. Users want transactions completed quickly with absolute confidence in security and reliability. Bridging this psychological divide is extremely difficult.

American consumers, he suggests, are unlikely to trust a single app with both their social lives and finances—a lesson reinforced by the success of specialized platforms like Venmo for payments and Instagram for social media.

The return to physical experiences

Despite AI's digital dominance, Chien is betting on a counterreaction: demand for in-person connection. He argues that in a world of infinite digital content, the scarcest resource is real human contact and real-world experiences.

This belief drives investments in companies like Bump, a Paris-based startup from the original Zenly founders (acquired by Snap), which creates interfaces for physical world interaction catalyzed by digital information. Also in his portfolio is Fever, a London and Madrid-based company positioned as the Live Nation of Europe. Starting with niche events like candlelight concerts and Bridgerton-themed experiences, Fever has scaled to mainstream success.

AI enhances these physical experiences by leveraging data on location, social circles, and time patterns to surface relevant interests. The result is more personalized and useful real-world encounters, making the return to physical experiences not just nostalgic but intelligently curated.

Looking ahead

As AI infrastructure commoditizes, the winners will likely be those who master the art of personalization—not as a feature, but as the core product. The shrinking gap between cloud and local AI will unlock new application categories, while consumer appetite for authentic experiences may reshape how technology integrates into daily life. For investors like Chien, the opportunity lies not in backing the next AI model, but in funding companies that know how to turn AI into deeply personal, human-centered experiences.