Here's a puzzle: NVIDIA's chips are powerful, but competitors make chips too. So why is NVIDIA so far ahead? The answer isn't the silicon — it's the platform and ecosystem wrapped around it. This lesson is the strategy lesson.
CUDA-X: turning tools into a platform
For 20 years, NVIDIA built software libraries (called CUDA-X) that solve hard problems in science, math, and engineering. In the agent era, those libraries become something new — tools that agents can pick up and use:
"CUDA-X libraries are tools for agents."
— Jensen Huang, NVIDIA GTC Taipei 2026And each one ships with "skills" — instruction manuals the AI can read:
"The CUDA-X library comes with skills — basically a manual — the AI reads it and goes, 'Aha, that's how you use it.'"
— Jensen HuangWhy does this matter strategically? Because the more tools agents can use on NVIDIA's platform, the more reasons there are to build on NVIDIA — and the harder it becomes to leave. That's a network effect, and it's the real moat.
The ecosystem: upstream to downstream
"NVIDIA's ecosystem spans all the way upstream to all of our supply chain… and downstream all the way to data centers, and eventually to end users."
— Jensen HuangA platform creates value by connecting many players — suppliers, partners, developers, customers — not by doing everything itself. NVIDIA's strength is that thousands of companies build with it, which makes it more valuable, which attracts more companies. The product is almost the smallest part of the story.
Giving it away on purpose: open models
Now the move that confuses people. NVIDIA released powerful AI models — Nemotron (language) and Cosmos (the physical world) — and gave them away, including the recipe:
"We open the model, we open the data, and we even open how we trained it so that you can enhance it for yourself."
— Jensen HuangWhy give away something so valuable? Because if everyone builds on your open models, everyone ends up running on your platform and buying your compute. Strategic openness trades control for adoption. It's the opposite of the closed, proprietary approach — and both are legitimate strategies with real trade-offs (open = reach and goodwill; proprietary = control and margins).
"Every company will be an agent company"
To help every business build agents, NVIDIA offers an Enterprise AI Toolkit with the four ingredients from Lesson 2: models, a harness, tools & skills, and a runtime. And Jensen pushes back hard on the fear that agents will wipe out software companies:
"People said the agents are going to disrupt these markets. I said: completely the opposite… this is an incredible time to be a software company."
— Jensen HuangHis reasoning: agents use tools — lots of them. More agents means more demand for software tools, not less. (Worth holding lightly: he's partly reassuring his partners. But the core point — that new platforms create new winners as well as losers — is solid history.)
This lesson is a tour of core IS strategy: platform strategy, network effects, ecosystems, complementary assets, vendor lock-in, and open vs. proprietary. The big professional skill is seeing past the product to the system of value around it. When you evaluate any vendor — for your company or your career — ask: How big and healthy is the ecosystem? How locked in would we become? Is their model open or closed, and what does that mean for our flexibility and cost? Those questions often matter more than a feature checklist.
Why would NVIDIA give away powerful AI models for free?
Open vs. proprietary: which AI platform?
Your company is choosing an AI platform. Option A is a closed, proprietary system that's polished but locks you in. Option B is built on open models you can run and modify yourself, but needs more in-house skill. How do you frame the decision?
There's no universal "right" answer — it's a trade-off you make explicit. Proprietary (A): faster to start, less to maintain, but you're exposed to price hikes and lock-in, and your data and flexibility live on someone else's terms. Open (B): more control, portability, and long-run cost leverage, but you need the talent to run it. You'd weigh: How core is this to our strategy? How sensitive is our data? Do we have (or want) the skills? How painful would switching be later? The translator move is turning "which is better?" into "which trade-offs can we live with, given our strategy?"
Name a platform you rely on (an app store, a game platform, a social network). What's its ecosystem, and how "locked in" are you? Could you realistically leave?
Do you buy Jensen's claim that agents are good news for software companies? Who do you think wins and loses as agents spread?