NVIDIA has taken AI performance to unprecedented levels, and according to CEO Jensen Huang, this was achieved with minimal reliance on Moore’s Law. Instead, NVIDIA has leveraged its own technologies to push boundaries even further.
For those unfamiliar, Moore’s Law traditionally notes that computing power doubles approximately every two years through node shrinking, meaning noticeable performance boosts happen every 12 months or so. However, in the fast-evolving AI landscape, we’re experiencing groundbreaking advances every six months, largely due to NVIDIA’s innovative strategies. Jensen asserts that Moore’s Law is no longer applicable in today’s world and believes there are virtually no limits to progress.
During a keynote at Computex, Jensen addressed whether the development pace of AI chips might slow down. He pointed out that Moore’s Law has lost its relevance because of increasing physical constraints, particularly since node size reductions no longer yield the performance leaps they once did. Instead, NVIDIA has achieved significant advancements through alternative methods.
Notably, Jensen credits advanced packaging technologies, such as TSMC’s CoWoS, for boosting performance. NVIDIA’s own NVLink interconnect technology further enhances chip capabilities by networking multiple platforms together, providing computational power previously thought impossible.
NVIDIA is not slowing its momentum. Jensen hinted that the company could potentially shift to a quarterly product release schedule, which would be a remarkable achievement. Currently, NVIDIA follows a “six-month” product cycle, as seen with the release of Vera Rubin and Blackwell Ultra, showcasing their ongoing commitment to driving AI innovation forward.
