What to Improve
6 gaps in Huang's approach that your company can address
"Even when free, not cheap enough" is dangerous complacency. Google TPUs run 75% of Gemini. Amazon Trainium offers 30–40% better price/performance. Some analysts predict 60%+ non-NVIDIA silicon by end of decade.
Study where your "competitor's chips are free" blind spot is. Every dominant platform holder eventually faces viable alternatives. The ones who survive take those alternatives seriously before they have to.
100% TSMC fabrication. 90% data center revenue. Four hyperscalers as primary customers. Any single point of failure (geopolitical disruption, customer consolidation, fab capacity) could be existential.
Identify your TSMC: the single supplier where failure is catastrophic. Build redundancy before you need it. The time to diversify is when you have leverage, not when you're desperate.
55+ direct reports, 100 emails daily. Works because Jensen is Jensen. The entire operating model is built around one person's superhuman capacity for context-switching, pattern recognition, and stamina. It doesn't survive succession.
Take the principles but build systems that don't depend on a single superhuman leader. Flat hierarchies and raw information flow are powerful, but they need to be encoded in process, not personality.
Points to radiologists as evidence AI doesn't kill jobs. But radiologists are high-paid specialists who can adapt. The impact on lower-skilled workers (data entry, customer service, logistics) is different in kind. The analogy doesn't hold.
Build genuine transition plans for affected workers. The company that sells the disruption should also invest in the transition. This is both ethical and strategically smart. It reduces political risk to your core business.
Arena keynotes, leather jacket, rock-star treatment. The showmanship is effective marketing but makes it hard to separate new strategic insight from repackaged messaging. Audiences cheer instead of interrogating.
Separate the framework (valuable, reusable) from the marketing (effective but not actionable for you). When studying Huang, focus on the mental models and decision patterns, not the performance.
33 years, no visible succession plan. Culture, strategy, and operating rhythm are intertwined with his personal style. The company's identity and its founder's identity are nearly indistinguishable.
Start succession planning early. Document frameworks and decision-making processes so they outlive your tenure. The best leaders build organizations that don't need them, even if they never leave.