Navin Chaddha on the $3T AI Opportunity and What to Build

Software is a $60-billion-a-year industry. The money the world spends on knowledge workers is $30 trillion. Navin Chaddha did the arithmetic out loud: if AI agents do just 10% of that work, "10% of 30 trillion is three trillion — is five times the size of the enterprise software market."

That number reframes the whole 2026 debate. We sat down with Chaddha, managing partner at Mayfield and a 17-time Forbes Midas List investor whose bets include Lyft, HashiCorp, and SolarCity, to pressure-test where the durable AI opportunities are, what's overhyped, and what founders should build right now. He's been in Silicon Valley since the early '90s, through PC, internet, social, mobile, and cloud. His read on this one is blunt: "If those prior waves were 10x disruptions, AI is a 100x disruption."

Why this wave is a 100x disruption

Chaddha's case for "once in a lifetime, not once in a decade" rests on two forces multiplying. The first is that the relationship between humans and machines just inverted. "Our whole life in the computing industry we have been fighting to understand machine language," he said. "But for the first time, machines can understand our language." Where it once took ten or twenty million people who could program to talk to computers, now, in his framing, seven billion people can. That's not a productivity bump. It's a market expansion.

The second force: compute and GPUs crossed a threshold where machines can "think, plan, and most importantly, take action." Stack a 10x making computing conversational on top of 10x intelligence-as-a-service and you get his 100x. The conclusion he keeps returning to is that this is the moment to participate, because a wave like it "may not come for the next twenty, thirty, forty years."

The headline he's least interested in is AGI. Asked directly about the various paths to it, Chaddha declined to predict a date and pushed back on the idea that one architecture wins. "You will need transformer-based models, these physical models, and spatial models are just starting. You will need multi-models." Robotics, he noted, is still "an unsolved problem" because the open-internet data that trained today's LLMs doesn't exist for the physical world. His phrase for where this lands isn't artificial general intelligence. It's collaborative intelligence. "AI is the horse, we are the jockey." Can AI do 70% of the tasks in a given use case today? Yes. Everything, in two years? "I don't think that's gonna happen."

The biggest market is professional services, reimagined

If the spend is in knowledge work, the most interesting target is the multi-trillion-dollar professional services industry, and Chaddha thinks it's about to be rebuilt the way e-commerce rebuilt retail. His logic starts from the constraint that actually blocks AI adoption in the enterprise, and it isn't the model. "Surprisingly, the number one is enterprises don't have talent." Services firms exist to supply that talent, billed per hour or per month.

What changes is who does the work and how it's priced. "The front end of a services company will still be a human which talks to a human," he said, with "a lot of the work" done by coding agents and humans still in the back end. The business model shifts from perpetual license to subscription to outcome-based: "You will get paid for the outcome you create," not per hour or per contractor. He pointed to Mayfield portfolio companies Groove and Supio as going after exactly this. Services don't disappear. They get "reimagined in how they run and how they price."

Underneath it all, Chaddha's map of the stack is where his investing advice gets sharp. The GPU is the brain, models are the operating system, and value gets created first in the plumbing, then moves up. Today it sits at the GPU and model layers. Over the next decade it migrates to AI-native applications, and "every industry, every function is going to have an AI teammate." Which is also why he's emphatic about what not to build: "You can't go say I'm building the next model which will take Anthropic down. That's looking in the rear-view mirror." Same for trying to topple Nvidia. "Don't try to be a copycat. Don't try to do incremental innovation. Do things which were not possible before."

Vibe coding, tiny teams, and the unicorn that isn't real

The democratization of building is real to Chaddha. He sees the developer population going from roughly 20 million who like writing code to 100 million creating apps with tools like Lovable, Cursor, and Claude Code. But he draws a hard line between making software and making a business. A lot of what gets built will be hobbyist apps. "The smart entrepreneurs will go create things which haven't been created before and serve new markets." He's also watching a stranger frontier: a portfolio company, Sekai, betting that "software is the new video" — a TikTok-for-software where people generate content, greeting cards, and wishes the way they make Reels. Their metrics, he said, are "through the roof."

On the viral claim that one person can build a billion-dollar company, Chaddha is precise and a little contrarian. He believes tiny teams will hit unicorn valuations, but he calls that out for what it is. "Unicorns are imaginary creatures. It's paper valuation." The real test is a billion dollars of revenue, "and that's gonna be hard. If the buyer of your product is a human, they don't buy from agents." His version is the tiny-team phenomenon: maybe 50 people where you'd once have needed 500, sometimes five, but rarely one, and always humans and AI working together.

For founders raising now, his Inception-stage checklist is people-first and refreshingly indifferent to TAM. "We don't care about market size — markets expand." He looks for founder-market fit, an X-factor, and IEQ over IQ ("everybody has IQ"). The problem has to be a painkiller, not a vitamin, a "hair on fire" problem where users are "climbing buildings" to tell others it exists. And he wants blue oceans: "Big companies get created in emerging markets," the way Uber, Lyft, and Airbnb made markets that didn't exist before. The line founders should tape to the wall: "Companies die of indigestion, not starvation. You have to do one thing and be the best in the world at it."

The takeaway

Chaddha's boldest 2026 prediction isn't about a model. It's about the wall in front of all of them. Every AI data center, he estimated, needs roughly 500 megawatts to a gigawatt, "equivalent to half a million to one million homes." So his call is the "golden era of nuclear coming back," specifically small modular reactors. If you're a hardware engineer, that's the company he'd build: solve cooling and power "so that I don't need to build more energy plants."

For everyone else, the advice points one way. Solve net-new problems 100x better, price for outcomes, build a focused team of humans and AI teammates, and stop trying to out-Anthropic Anthropic. The two books he hands founders, Inside the Tornado and Built to Last, bracket the same idea: survive the chaos first, then build something with a mission, vision, and values worth lasting for.

That focus on customers, code, and founder-market fit over funding-as-finish-line is the thesis behind QAI Ventures — capital, customers, and code for AI-native founders. Watch the full conversation with Navin Chaddha on The QAI Podcast, and find every episode at /podcast.