Chien warns AI winners won’t sell the models

Venture capitalist Chi-Hua Chien says the “model layer” is already being commoditized, and that the next AI winners will likely sell personalization and real-world experiences—not access to AI itself. In an edited discussion, he also argues Americans won’t tru

Chi-Hua Chien has spent more than two decades as a venture capitalist, but he doesn’t talk about artificial intelligence the way many fund managers do. He speaks like someone trained to notice human behavior—at scale, in culture, in habits—and then predict where value will actually stick.

As a co-founder of Goodwater Capital. a firm focused exclusively on consumer and prosumer technology. he’s built a portfolio across entertainment. healthcare. fintech. and live experiences. Investments he points to include MIDI Health. Fever. and Monzo—and his perspective is shaped by unusual origin stories as well. In the late 1990s. as a 27-year-old associate at Accel. he was the person who initially found a six-person company launched from Harvard called The Facebook.

In a conversation edited for length and clarity, Chien makes a pair of arguments that tug in opposite directions. One is bullish: the gap between what the best AI models can do and what you can run on your phone is shrinking fast. The other is blunt: the biggest winners of the AI era won’t be the companies selling AI at all.

A long view on investor behavior

Chien traces the recent surge in public grievances from founders and investors back to a cultural shift. He calls it the “meme-ification of everything,” where political-style amplification is bleeding into business. He also links the volume of outspoken criticism to venture firms getting bigger and more vertically integrated.

In his telling. the biggest firms don’t need syndicate partners as much as they used to. because they already have enough capital to lead deals on their own. That. he says. has reduced the old decorum around preserving relationships with co-investors—relationships that still mattered when capital was coordinated across multiple firms.

He also addresses “fast follow” rounds—an approach where investors put in a large chunk at a given valuation and then add a smaller amount weeks later at a higher valuation, creating headline numbers that look stronger than the structure might suggest. For Chien, it’s not new.

He says the fastest-growing companies now raise successive rounds quickly. with gaps often limited to three to six months. while valuations shift rapidly. In his view. valuations are being marketed aggressively as signals of market leadership—messages aimed at attracting talent and potentially blocking competition. He doesn’t dismiss the strategy as pure manipulation. but he does describe “frothiness. ” driven by an imbalance between demand and supply. If an investor can enter. set a price. finish a financing. and still find excess demand a couple of weeks later. a company can immediately price a new round higher.

Infrastructure gets commoditized; apps capture value

Chien’s theory of the cycle is anchored in past technology eras. He argues that infrastructure companies eventually get commoditized, while applications capture most of the value over time.

He points to the PC cycle, web cycle, and mobile cycle for similar patterns. He says infrastructure market caps peaked in the year 2000. and that even after 25 or 26 years. the market cap of infrastructure companies hasn’t surpassed that 2000 peak in nominal dollar terms. He then offers specific comparisons: in the web era. infrastructure new entrants produced $400 billion of new market cap. while application companies created $3.1 trillion—88% of the new value. In the mobile era, infrastructure produced about $700 billion while application companies produced $3.7 trillion. He names examples including Netflix, Spotify, Meta, Uber, and Airbnb.

The subscription battlefield is already changing

The shift he’s watching isn’t theoretical. He points to Google’s subscription AI product, saying it announced that the price is dropping from $7.99 a month to $4.99 a month while doubling the storage. For Chien, this fits an emerging era of price competition.

He argues that companies with structural advantages in vertical integration and distribution—like Google—can bundle and price-attack the average consumer.

Where AI becomes more personal

Chien returns again and again to personalization as the practical through line. He says hyper-personalization can drive higher customer satisfaction, deeper engagement, and higher ARPUs over time.

He describes entertainment companies in his portfolio—Triumph. Ritten. and Flow GPT—where customers aren’t necessarily saying “this is an AI application.” Instead. they’re buying an entertainment application. He claims these companies are reaching 100 million. 400 million. and 600 million of ARR quickly with high margins. because AI makes experiences more customizable and personalized—even if the core capability being sold isn’t framed as AI itself.

He makes the point more sharply with women’s health. One of the companies he highlights is Midi Health. He says a fundamental constraint in women’s health is that there aren’t many providers well trained in hormone replacement therapy for perimenopausal women. By using AI. he says. the company can expand the supply of care. treat hundreds of thousands of patients who otherwise couldn’t be reached. and do it cost effectively—expanding access to a market that was previously supply constrained. He also suggests that this could play out across every supply-constrained category where human expertise is the bottleneck.

Phones are catching up faster than people think

On the technical timetable, Chien says the future isn’t far away. He believes you can run locally on your phone now models that are as good as the best models were about six months ago, and that the lag is shrinking.

He recalls that two years ago, the lag between what you could run locally and what was available in the cloud with frontier models might have been 18 to 24 months. Now, he says, it’s six months. He adds that it’s “probably getting down to three months by this time next year.”

What’s missing, he argues, isn’t only model capability. It’s use cases that are clearly defined. He compares the moment to mobile adoption: when the iPhone launched in 2007. he says many people expected it to be mostly a port of web applications. Instead, it took time for entrepreneurs to figure out what the new platform made possible.

He also describes the underlying “two things” that LLMs enable. He says they make it possible to process large amounts of context and make sense of it, and they allow cost-effective personalization down to the individual through a feedback loop that makes the product better over time.

Why a “super app” is still a trust problem

Chien’s skepticism about merging social and financial services echoes an earlier prediction: Americans won’t trust a single app that covers both social life and finances.

He frames it around Facebook’s long, uneasy attempts at a super app. He points to multiple efforts: Facebook Credits, which launched in 2009; Facebook Pay; and Libra. None, in his view, realized a true super app.

He explains the hurdle through what he calls an intuitive perspective on trust. There’s a trust gap between entertainment and social products, and commerce, banking, and financial services—especially in the Western world.

He contrasts how people experience the two categories. Financial transactions, he says, carry seriousness and are relatively high monetization and relatively low time. Social platforms, by contrast, can turn attention into value even though they monetize in a different way. The psychological expectations are the key challenge: customers “don’t want to hang out” in a banking app—they want to transact and be done. with extremely high confidence in the security and reliability of that transaction.

Real-world connection as the counterreaction

In the final stretch of the conversation, Chien points to another kind of differentiation—one that doesn’t live inside a model layer at all.

He says his team believes people crave real human contact and real-world experiences in a world where there’s an infinite supply of digital content. In his formulation, the thing people want most is what is most constrained.

He ties that belief to investments in physical-world interaction: Bump, based in Paris, built an interface that allows people to interact in the physical world catalyzed by digital information. He says Bump traces back to the original founders of Zenly, which was acquired by Snap.

He also highlights Fever, based in London and Madrid, which he calls “essentially the Live Nation of Europe.” Fever, he says, started with smaller, quirky events like candlelight concerts and the Bridgerton Experience, before going mainstream.

Chien says the swing is already happening back toward in-person consumption, with AI as an enabling technology that can anticipate interests—where someone goes, who they hang out with, and where they tend to spend time—making experiences more useful and more personal. He calls it “super exciting.”

And even as the conversation stays rooted in venture theory and market cycles. the through line is practical: the next wave of winners may not look like the loudest sellers of AI. They may look like businesses that translate AI into something people feel—whether that’s a better entertainment experience. more accessible healthcare. or a night out that feels tailor-made.

Chi-Hua Chien Goodwater Capital AI personalization LLMs on mobile commoditization of infrastructure venture capital fast follow rounds Google subscription AI price drop Midi Health Fever Bump