Jensen Huang’s 30-Year AI Crusade

In February 2025 NVIDIA posted record quarterly revenue of $39.3 billion, nearly triple the year before, and CEO Jensen Huang opened the call with a grin: “The GPU, once for games, is now the factory of the 21st century.” Investors cheered, governments courted him and every AI start‑up begged for more H100 chips. None of it was luck. It was the payoff from three decades of patiently building the right technology long before the world cared.

Origin Moment: Napkins, Coffee, and a Crazy Idea

Huang was born in Taiwan, moved to rural Kentucky at nine, and washed dishes at a local Denny’s through high school. In 1993, he met two chip‑designer friends at another Denny’s in Silicon Valley. Over refills of bottomless coffee they sketched a plan on a napkin: build a new kind of processor that could make 3‑D graphics feel real. Each co-founder put in $40 000, rented a tiny office, and named the company NVIDIA.

Those diner nights set his leadership compass: start scrappy, stay close to builders, and never stop iterating. By 1999, the team had shipped GeForce 256, the first GPU to handle 10 million polygons per second, and it had seized the PC‑gaming crown.

First Turning Point: Betting on CUDA Before Anyone Asked

Gaming was good business, yet Huang worried that graphics alone could not outrun commoditization. In 2006, against board skepticism, he funded an internal skunk‑works project called CUDA that let scientists program GPUs for general computation. Analysts called it a distraction; rivals laughed.

The payoff arrived six years later when researchers training AlexNet discovered GPUs could slash AI training time from weeks to days. CUDA was already mature, so Huang could ship developer tools overnight while competitors scrambled to catch up. That first‑mover advantage still underpins more than 90 percent of deep‑learning workloads that run on NVIDIA silicon today.

Why it matters: Investing in a platform that solves tomorrow’s problems lets you capture network effects that capital alone cannot buy later.

Cultural Reset: Turning a Chip Vendor into a Platform Partner

Early NVIDIA acted like a classic parts supplier: sell chips and move on. Huang flipped the script. He formed a Developer Relations army, published open‑source libraries, and held free training labs called Deep Learning Institutes. Inside, he banned PowerPoint-only product pitches—engineers must demo code or hardware.

The culture shift created a flywheel: more tools attracted more developers, whose feedback improved next‑gen silicon, which lured even more developers. That ecosystem now includes 4 million registered engineers and thousands of start‑ups building on NVIDIA SDKs.

Second Turning Point: The Data‑Center Moonshot

Consumer GPU sales flattened in 2014. Rather than chase price wars, Huang poured capex into data‑center chips purpose‑built for AI, then bundled them with networking, software, and support. Critics called it “over‑integration.”

Fast‑forward to 2025: data‑center products delivered $35.6 billion in one quarter, dwarfing gaming revenue and turning NVIDIA into the arms dealer of generative AI.

Key insight: when you own the full stack—silicon, firmware, drivers, libraries, even customer training—you turn a component sale into an annuity.

Mindset & Habits: Four Practices You Can Steal

Lessons for Readers

1. Build Before the Market Arrives

Huang funded CUDA when no customer asked for it. Early investment let NVIDIA set standards rather than chase them once AI exploded. Pioneers capture platforms; followers rent them.

2. Turn Customers into Co-Designers

Developer dinners and open libraries blur the line between vendor and user. Shared roadmaps create loyalty deeper than contracts and make competitors’ offers feel risky.

3. Control the Stack to Control Margins

By selling entire AI systems—from chips to networking to software—NVIDIA moves upstream and locks in premium pricing while shielding itself from single-component commoditization.

4. Maintain Founder‑Level Scrappiness

Unannounced lab visits and handwritten notes remind a 30,000-person company that craft matters. Rituals prevent layers of management from dulling urgency.

5. Write Roadmaps in Pencil, Not Stone

A decade-long vision guides capital spending, yet quarterly reviews let Huang pivot when new science emerges. Long-term thinking plus short feedback loops beat either extreme.

Weekly Challenge

Pick one strategic bet your team keeps postponing. Allocate a small budget this week to build a working prototype, not a slide deck. Share what you learn, and we’ll feature the most daring experiments in next Thursday edition.