Forget Selling Models: Real AI Value Moves to Smart Apps

Chi-Hua Chien has spent more than twenty years working as a venture capitalist, but he approaches the market more like a cultural anthropologist. As a co-founder of Goodwater Capital, a firm focused entirely on consumer and prosumer technology, he has built a portfolio spanning entertainment, healthcare, fintech, and live experiences. His investments include notable companies like MIDI Health, Fever, and Monzo. Long before this, as a twenty-seven-year-old associate at Accel, he was the person who initially spotted a tiny, six-person company launched from a Harvard dorm room called The Facebook.

Now, while much of the venture capital world obsesses over underlying models, advanced chips, and raw computing infrastructure, Chien looks at the landscape differently. He argues that history always repeats itself, and the biggest long-term winners of the artificial intelligence era will be the application companies built on top of the technology, not the companies selling the base models or APIs.

Historical data heavily backs up this theory. During the original desktop web cycle, infrastructure companies generated roughly four hundred billion dollars in total market capitalization. Meanwhile, application companies generated a massive 3.1 trillion dollars, which accounts for eighty-eight percent of all new value created during that era. The massive shift repeated itself during the mobile cycle, where infrastructure brought in seven hundred billion dollars compared to a staggering 3.7 trillion dollars generated by consumer-facing applications like Netflix, Spotify, Meta, Uber, and Airbnb. In both previous tech waves, infrastructure valuations peaked incredibly early and never returned to those high points in nominal terms.

Chien points out that the aggressive price wars among foundational AI models have already begun, proving that base infrastructure is turning into a commodity. When a major tech platform drops its monthly subscription rate by half and doubles the included storage capacity, price erosion is no longer a future prediction. It is a live trend happening right now. For purer model providers, this environment means profit margins will shrink rapidly because customers ultimately care about moving and sorting data as cheaply as possible rather than paying a premium for a specific engine.

The real differentiator in this changing landscape is building specialized, highly personalized products that address categories with severe supply shortages. Chien highlights specific examples from his own portfolio to illustrate this point. One company, a women’s healthcare platform, uses automation to scale the availability of specialized doctors who can treat perimenopáusical health issues. This is a massive market that traditional healthcare systems have historically ignored due to a lack of trained medical experts. By using technology to scale doctor workflows, the platform managed to treat hundreds of thousands of active patients while maintaining excellent profit margins.

Similarly, entertainment companies in his portfolio are scaling fast, with some hitting between one hundred million and six hundred million dollars in annual recurring revenue. These businesses do not pitch themselves to the public as AI startups. Instead, they position themselves as platforms for hyper-personalized human experiences. The artificial intelligence serves as a hidden engine that makes the product function smoothly, not the core product itself.

On the technical front, the gap between massive cloud-based models and the chips running locally inside your mobile devices is closing at a staggering speed. Two years ago, smartphones trailed the capabilities of leading frontier models by nearly two full years. Today, that processing delay has shrunk to just six months. Chien predicts that by mid-2027, the gap will drop to a mere three months.

This rapid convergence means that hyper-personalized, local intelligence will move straight into our pockets much faster than industry analysts expect. For software developers and technical leaders, this means architectural decisions that rely heavily on external cloud APIs have a much shorter shelf life than current roadmaps assume. Local processing will quickly shift from an optional premium feature to a standard design expectation due to its massive advantages in user privacy, zero latency, and lower operational costs.

Ultimately, Chien believes the next major consumer gold rush will not happen by selling raw algorithms. The founders who build generation-defining companies will be the ones who focus on using these tools to solve real human problems, scale scarce resources, and help people reconnect in the physical world.