THE AI INTERVIEW
Michael Kagan has witnessed three of computing’ s great upheavals from the inside. Today, as CTO at NVIDIA, the US company whose computing platforms power much of the world’ s AI, he sits at the centre of a fourth.
Michael’ s career began at semiconductor giant Intel, where he helped develop the processors that drove the personal computer revolution. He left in 1999, after managing the design of the Pentium MMX, a breakthrough chip of its era.
Michael went on to co-found Mellanox, a high-speed data centre networking company, with eight colleagues. Its founding vision was that computing would become a service rather than a server. He reflects:“ Nine of us started the company, asking‘ how hard can it be?’ It was very hard.”
Mellanox’ s technology became fundamental to data centres, helping to make the cloud revolution possible. That same high-speed networking later enabled AI, enabling vast numbers of processors to work together on complex computations.
NVIDIA began working with Mellanox around 2012 and completed its acquisition to the tune of US $ 7bn in 2020. Michael became NVIDIA’ s CTO that year, after 21 years at Mellanox.
He recalls:“ We were joining forces with a spaceship and getting into its cockpit, with the opportunity to make a much more significant impact on the world.”
“A traditional data centre wasn’ t designed as a production facility”
Michael Kagan CTO NVIDIA
Why the AI factory is different The term‘ AI factory’ is now widely used, not least within NVIDIA’ s four walls and across a number of product announcements – but also among its global partner network.
While the AI factory evolved from the data centre, Michael explains that they are“ fundamentally different”.
He says:“ A traditional data centre wasn’ t designed as a production facility, but an AI factory actually manufactures intelligence. It is one big machine running one big service. It is not just multiple computers wired together with a network, but one machine designed from the ground up.”
That shift is being driven by the limits of Moore’ s Law, the observation that the number of transistors on a chip doubles roughly every two years. AI has pushed demand for computing far beyond that pace.
“ Since around 2015, demand for computing has grown at a rate of about two times every quarter,” continues Michael.“ Even if Moore’ s Law continued,
22 October 2026