Jensen Huang's AI Bet: How NVIDIA Went from Silence to $5 Trillion

When Jensen Huang launched CUDA in 2006, nobody wanted it. Nobody asked for it. Nobody clapped. NVIDIA's valuation cratered from $12 billion to under $3 billion. Wall Street called it a waste.

Eighteen years later, NVIDIA is worth over $5 trillion — the most valuable company on Earth — and every major AI model on the planet runs on its hardware. This is the story of the most consequential technology bet in modern history.

The Denny's Origin Story (1993)

NVIDIA started at a Denny's restaurant in San Jose in 1993. Jensen Huang, Chris Malachowsky, and Curtis Priem — three engineers — pooled their ideas about building specialized chips for 3D graphics rendering. They launched out of a condo in Fremont, California, with a simple thesis: games would drive demand for dedicated graphics processing.

By the late 1990s, that bet paid off. The GeForce GPU became the gold standard for PC gaming. But Huang saw something most people missed: GPUs weren't just for pixels.

The CUDA Gamble (2006)

In 2006, NVIDIA unveiled CUDA (Compute Unified Device Architecture) — a software platform that allowed developers to use GPUs for general-purpose computing, not just graphics. This meant thousands of GPU cores could run mathematical operations in parallel, making them ideal for scientific simulations, financial modeling, and — eventually — neural network training.

The market's response was brutal. NVIDIA had effectively doubled the cost of every chip to include this new software layer. Customers didn't ask for it. Developers didn't understand it. The stock collapsed.

Huang didn't flinch. He poured hundreds of millions into CUDA development, building compilers, libraries, and developer tools. For nearly a decade, the investment produced minimal returns.

As Huang explained on the Acquired podcast: "When you push your chips in, I know it's going to work. We developed the entire software stack. We ran QA on all the drivers and all the software. When you bet the farm, you're saying: I'm going to take everything in the future, all the risky things, and pull them in advance."

The AlexNet Moment (2012)

Everything changed in 2012. Alex Krizhevsky, a University of Toronto researcher, trained a deep neural network called AlexNet on two NVIDIA GTX 580 GPUs. The system won the ImageNet image recognition competition by a historic margin — reducing the error rate by 10 percentage points over the previous best.

AlexNet proved that GPUs could train neural networks orders of magnitude faster than CPUs. Suddenly, every AI researcher in the world needed NVIDIA hardware. CUDA — the software layer nobody wanted — became the only way to access that hardware efficiently.

Hand-Delivering the Future (2016)

By 2016, deep learning was exploding. Huang was so deeply embedded in the AI research community that he personally knew the researchers who would go on to lead OpenAI, Anthropic, Google Brain, and DeepMind.

In a now-famous moment, Huang hand-delivered the first NVIDIA DGX-1 AI supercomputer to OpenAI's San Francisco lab. That machine — and its successors — would train the models that became GPT-2, GPT-3, and eventually ChatGPT.

The ChatGPT Inflection (2022–2023)

When ChatGPT launched in November 2022, it triggered the largest corporate demand surge in semiconductor history. Every tech company on Earth suddenly needed thousands of NVIDIA GPUs to train and serve AI models.

NVIDIA's trajectory accelerated:

• June 2023: Crossed $1 trillion market cap
• February 2024: Crossed $2 trillion
• June 2024: Crossed $3 trillion
• October 2025: First company to reach $5 trillion

The company's stock price increased roughly 25x from pre-ChatGPT levels.

The Numbers Behind the Dominance (2026)

NVIDIA's financial results are staggering:

• Revenue FY2026: $215.9 billion (up 65% year-over-year)
• Data center revenue: $193.7 billion (90% of total revenue)
• AI chip market share: Approximately 90% of all AI training chips
• Order backlog: Over $500 billion through end of 2026
• Jensen Huang's net worth: $176 billion (7th richest person globally)

The hyperscalers — Meta ($135B capex), Google ($185B capex), Microsoft, Amazon — are collectively spending nearly $700 billion on AI infrastructure in 2026. More than half flows through NVIDIA.

The CUDA Moat

NVIDIA's dominance isn't just about hardware. The real moat is CUDA's software ecosystem. After 20 years of development, CUDA has:

• Millions of developers trained on the platform
• Every major AI framework (PyTorch, TensorFlow, JAX) optimized for CUDA
• Thousands of libraries for specific domains (cuDNN for deep learning, cuBLAS for linear algebra, TensorRT for inference)
• An ecosystem so deep that switching to AMD or custom ASICs requires rewriting entire software stacks

Competitors like AMD (ROCm) and Intel (oneAPI) are trying to break in, and hyperscalers are building custom ASICs. But Meta's decision to commit billions to NVIDIA GPUs in 2026 — despite building its own chips — shows how difficult it is to replicate two decades of software investment.

Next Generation: Rubin (2026)

NVIDIA isn't standing still. The Vera Rubin platform, shipping in the second half of 2026, represents the next leap:

• Six chips designed together using "extreme codesign"
• Up to 10x reduction in inference token costs
• Purpose-built for the agentic AI workloads that will define the next era

The Leadership Lesson

Jensen Huang's story isn't just about technology — it's about conviction under pressure. He bet NVIDIA's future on a technology nobody wanted, endured a 75% stock collapse, and spent a decade building an ecosystem with no clear payoff. When the AI revolution arrived, NVIDIA wasn't just ready — it was the only company that could serve the demand.

As Huang himself put it: "Without GeForce there would be no CUDA. Without CUDA there would be no AI. Without AI there would be no today."

TL;DR

Jensen Huang launched CUDA in 2006 to silence and a stock crash. He spent 18 years building the GPU computing ecosystem. When AI exploded, NVIDIA was the only company ready — now worth $5 trillion with 90% AI chip market share, $216B revenue, and $500B+ in orders.

—

Explore AI infrastructure in our tools directory, learn about transformer architecture, and compare model costs in our LLM pricing calculator.