ACQUIRED · EXTRACTED

NVIDIA

7 principles from the most improbable hardware company in history — on surviving near-death, betting everything on a vision, and building the infrastructure of the future.

▶5.0M views on YouTube

"We were 30 days from going out of business. I had to make the hardest decision of my career — kill our current product and bet everything on something that didn't exist yet." — Jensen Huang
Ben Gilbert & David Rosenthal, on the episode

NVIDIA was founded in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem. It nearly died multiple times in its first decade. It pivoted from a failed graphics chip architecture, survived the dot-com bust, and then made a bet on parallel computing that nobody else believed in. That bet — CUDA, launched in 2006 — is what turned NVIDIA into the most valuable semiconductor company in history and the backbone of the AI revolution. Ben and David tell the full story across multiple episodes.

Tactic 01

Kill the Product Before the Market Does

In 1995, NVIDIA shipped its first product — the NV1. It failed completely. The architecture was wrong, the market rejected it, and the company was months from bankruptcy. Jensen's decision: kill the NV1 entirely, abandon the existing codebase, and start over from scratch with a DirectX-compatible architecture that didn't exist yet. He laid off half the staff to fund the pivot. The NV3 (Riva 128) shipped 18 months later and sold 1 million units in 4 months. If Jensen had protected the sunk cost of NV1, NVIDIA would be a footnote.

The play

When a product is fundamentally wrong for the market, the fastest path forward is a clean kill — not iteration. Identify the one architectural assumption that's broken. Abandon it completely. Rebuild around what the market actually needs.

Tactic 02

Ship Every Six Months or Die

After the NV1 disaster, Jensen imposed a rule: NVIDIA ships a new GPU every six months. Not when it's ready — every six months. This was insane by semiconductor standards where product cycles were measured in years. But it meant NVIDIA was always one generation ahead. Intel, ATI, and 3dfx couldn't keep up. By the time competitors shipped a response to one NVIDIA product, two new NVIDIA generations had already shipped. The cadence itself became a competitive moat.

The play

Set a shipping cadence that your competitors can't match and hold it religiously. The goal is not perfection on any single release — it's compounding improvement over dozens of releases. Cadence creates momentum that individual products can't.

Tactic 03

Bet on the Platform, Not the Product

In 2006, NVIDIA launched CUDA — a programming framework that let developers use GPUs for general computation, not just graphics. Nobody asked for it. The gaming customers didn't care. The financial return was invisible for years. Jensen launched it anyway because he believed GPUs would eventually be used for everything that required parallel computation. When deep learning exploded in 2012 — with AlexNet running on NVIDIA GPUs — CUDA was already there, already mature, already the standard. The platform bet paid off a decade later.

The play

The best platform investments look like costs for years before they look like moats. Identify what your core technology could enable beyond its current use case. Build the platform layer now, before you need it.

Tactic 04

The Intellect Deserves the Truth

Jensen's management philosophy — which he calls 'the intellect deserves the truth' — is ruthlessly direct. Bad news travels up at NVIDIA faster than anywhere else because Jensen has made it psychologically safe to surface problems early. He does not shoot messengers. He does not reward comfortable reporting. He has a saying: 'Every email I send, I CC everyone.' Radical transparency creates a culture where problems get solved instead of hidden, which is the only culture that can survive in a market that moves as fast as semiconductors.

The play

Create explicit systems for bad news to travel fast. CC people on uncomfortable information. Reward the person who surfaces a problem early over the person who hides it. Speed of internal information flow determines speed of problem-solving.

Tactic 05

Vertical Integration at the Right Moment

NVIDIA is fabless — it designs chips but doesn't manufacture them. For years this was seen as a weakness. But Jensen treated the relationship with TSMC as a strategic partnership, not a vendor relationship. NVIDIA invested heavily in co-developing manufacturing processes with TSMC, got preferential capacity allocation, and built deep technical integration. When AI demand exploded, NVIDIA had supply chain priority that competitors couldn't replicate overnight. The fabless model combined with deep manufacturing partnerships gave NVIDIA flexibility AND scale.

The play

Being fabless doesn't mean being weak in manufacturing. Build deep technical partnerships with your key suppliers — invest in their processes, share roadmaps, create mutual dependency. Supply chain priority is a competitive advantage.

Tactic 06

Survive Long Enough to Be Right

Jensen's bet on parallel computing and AI was correct — but it took 15 years to pay off. In 2006 when CUDA launched, NVIDIA's stock barely moved. The gaming business subsidized the platform investment for over a decade. Most companies would have killed CUDA in year 3 when the ROI wasn't visible. NVIDIA didn't. The lesson: being right about a long-term trend is necessary but not sufficient. You also have to build a business that generates enough cash to fund the conviction while you wait.

The play

For any long-term bet you believe in: build a cash-generative core business that funds the patience required to be right. Identify what your 'gaming business' is — the near-term revenue engine that buys time for the long-term vision.

Tactic 07

The Chip Is the Roadmap

Every NVIDIA chip generation embeds assumptions about where computing is going 5-10 years out. The Tensor Core (introduced in Volta, 2017) was designed specifically for matrix multiplication — the core operation in neural networks — before neural networks were the dominant workload. By the time transformers took over AI in 2020-2022, NVIDIA's hardware was already purpose-built for them. Jensen designs chips for the future he believes in, then waits for the market to arrive.

The play

Build your product roadmap around where your market will be in 5 years, not where it is today. The gap between 'what customers are asking for' and 'what they'll need' is where durable competitive advantage lives.

Your action plan

the plays, one checklist.

01
Kill the Product Before the Market Does
Sunk cost is the enemy. Kill the wrong product fast and rebuild around what the market actually needs.
02
Ship Every Six Months or Die
Set a release cadence your competitors can't match. Ship on schedule, not when it's perfect.
03
Bet on the Platform, Not the Product
Build the platform before the market demands it. The best infrastructure bets look like waste until suddenly they're everything.
04
The Intellect Deserves the Truth
Make bad news travel fast internally. The culture that surfaces problems earliest solves them fastest.
05
Vertical Integration at the Right Moment
Treat key suppliers as strategic partners, not vendors. Deep integration creates supply chain advantages that money alone can't buy.
06
Survive Long Enough to Be Right
Being right too early is the same as being wrong. Build the cash engine that lets you wait for the market to catch up.
07
The Chip Is the Roadmap
Design for the customer you'll have in 5 years, not the one you have today. The roadmap is the strategy.