Chi-Hua Chien has spent more than 20 years as a venture capitalist, but he thinks like a cultural anthropologist. As co-founder of Goodwater Capital, a company focused on consumer and prosumer technology, he has invested in companies such as MIDI Health, Fever, and Monzo, building a portfolio spanning entertainment, healthcare, fintech, and live experiences. As a 27-year-old employee at Accel, he also first founded The Facebook, a six-person company out of Harvard University.
His ability to read human behavior at scale speaks to everything from his view that Americans will never trust a single app for both their social lives and finances to his belief that the gap between cutting-edge AI models and what they can run on their phones, once as long as two years, will shrink to three months within the next year.
These days, many venture capitalists are willing to say out loud what they just think: that the commoditization of the model layer is already underway, and that the biggest winners in the AI era will not be the companies selling AI.
We spoke last week. This interview has been edited for length and clarity.
Lately, more founders and investors have been publicly sharing their frustrations with VCs. What has changed?
It's part of the meme-ification of everything, and you see what's happening in the political world seeping into the business side. It's also probably a sign of peakiness in the market. The reason some of these outspoken investors are becoming more public is because venture companies are almost vertically integrated, so the really big companies have enough capital that they're not necessarily looking for syndication partners. There used to be a courtesy of wanting to maintain good relationships with other co-investors. This is because we have to cooperate with them in various situations. As companies grew larger and became more vertically integrated, that need became less necessary.
What about “fast follow” rounds? Are companies investing large sums of money at one valuation, then a small sum a few weeks later at a much higher valuation, making the headline numbers look more impressive than they really are? Is this really new? How widespread is it?
I think it's been going on for quite a while. The best companies raise their next round very quickly. Today, there may only be 3-6 months between rounds, but valuations can change very quickly. Valuations are very aggressively marketed as a way to demonstrate market leadership, attract talent, and potentially deter competition. What these rapid financings best represent is that there is much more demand than supply, so there is probably some element of frothiness. Once the investors come and set the price and close the funding, there is still excess demand weeks later, allowing the company to immediately price a new round at a higher price.
You argued that infrastructure companies will become commoditized and that over time applications will capture most of the value. Are we already seeing that play out in this cycle?
If you look at the PC cycle, the web cycle, and the mobile cycle, they all follow a fairly consistent pattern. Infrastructure market capitalization actually peaked in 2000, but 25, 26 years later, in nominal terms, the market capitalization of these infrastructure companies has not exceeded its 2000 peak. In the web era, infrastructure new entrants created $400 billion in new market capitalization. Application companies created $3.1 trillion, or 88% of new value. The same goes for the mobile era, where infrastructure generated about $700 billion, while application companies generated $3.7 trillion. Companies like Netflix, Spotify, Meta, Uber, and Airbnb.
and [last week] I saw something very interesting. Google announced that it will reduce the price of its subscription AI products from $7.99 per month to $4.99 per month and double the storage. We are already in an era of price competition. And companies like Google, with their vertical integration and structural advantages in distribution, may begin to bundle and compete on price for the average consumer.
It keeps coming back to personalization. Will that be what separates the next wave of winners?
Hyper-personalization is definitely important throughout the line. What does personalization give you? Done right, it increases customer satisfaction, deepens engagement, and increases ARPU over time.
Our portfolio includes entertainment companies like Triumph, Ritten, and Flow GPT, but our customers aren't saying, “This is an AI application.” They say it's an entertainment application. These companies are reaping significant benefits, with ARR reaching $100 million, $400 million, and $600 million very quickly, as AI makes the experience more customizable and more personalized. But that's not the basic feature they're selling.
There is also a women's health company called Midi Health. One of the fundamental constraints in women's health is that not many health care providers are adequately trained to administer hormone replacement therapy to perimenopausal women. Using AI, we can significantly expand the supply of care and treat hundreds of thousands of patients who would otherwise be unreachable. It can also be done cost-effectively, increasing access to markets where supply has previously been limited. This can be driven across all supply-constrained categories where human expertise is the bottleneck.
How far are we from truly personal and atmospheric AI?
I don't think we're that far away. You can now run smartphone AI models locally that are as good as the best models from about six months ago, and that lag is shrinking. If you go back two years, the Frontier model might have had an 18-24 month lag between what you could run locally and what you could run on the cloud. It's been 6 months now. By this time next year, it will probably have been shortened to three months.
What we don't have yet is use cases that are very clearly defined. When the iPhone was released in 2007, people largely assumed that all web applications would be ported to mobile. It takes time for entrepreneurs to instill what is currently possible.
What the LLM does is basically two things when you extrapolate from the mechanism to estimate its function. One, it allows us to process large amounts of context and make sense of it all, and two, it allows us to cost-effectively personalize down to the individual, with feedback loops that make the product better over time.
You've seen Facebook experiment for years trying to build a super app. Why is it so difficult for American consumers to blend financial services and social entertainment?
They took multiple shots on goal – Facebook Credit launched in 2009…Facebook Pay, Libra…they failed to deliver a true super app. I think people have an intuitive view of trust, but there's a trust gap between entertainment and social products and commercial, banking and financial services, especially in Western countries.
There is a seriousness to financial transactions that is very different from the triviality of social media. Don't get me wrong. That little thing has created a company worth over $1 trillion. However, financial services are actually the complete opposite. Audience time is very long and monetization is relatively low, whereas financial services transactions are highly monetized and time is relatively short. You don't have to keep using your banking app. You want to make and complete a trade, but you need a high level of confidence in the security and reliability of that trade. It is very difficult to satisfy the psychological expectations of customers.
Are you betting that the reaction to all this is that people are craving in-person connection?
We truly believe this. In a world with an endless supply of digital content, what do people crave? They crave the least constrained: real human contact, real-world experiences.
We're investing in a company called Bump, which is based in Paris. The company comes from the original founders of Zenly, which was acquired by Snap. They have built interfaces that allow people to interact in the physical world, using digital information as a catalyst. We also have Fever based in London and Madrid. This is essentially a European live nation. It started with small, quirky events like Candlelight Concerts and the Bridgerton Experience, and has since gone mainstream.
I think we're moving back in a different direction away from pure online consumption. By knowing where we go, who we spend time with, and where we tend to spend our time, AI as an enabling technology will be able to infer a wealth of information related to relevant interests that will make our real-world experiences more convenient and more personal. That's very exciting for us.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.