Tom Riley, head of global thematic strategies at Axa IM, was interviewed by Delano on 6 March 2024 to share his views on the technology sector. Photo: Axa IM

Tom Riley, head of global thematic strategies at Axa IM, was interviewed by Delano on 6 March 2024 to share his views on the technology sector. Photo: Axa IM

Delano reviewed AI progress with Axa IM, the elephant in the room whenever one is discussing with a technology expert. In addition to reviewing Nvidia prowess, the article--the first part of a two-instalment series--covers the companies contributing to the “AI revolution” and companies in other sectors displaying remarkable technologies.

Growth reported are not projections, they are real revenues

“We have seen meaningful revenue growth associated with AI,” said Tom Riley, head of global thematic strategies at Axa IM during an interview in early March. He noted that 4Q2023 data centre sales at Nvidia were $18.4bn and grew 409%, year-on-year. “Normally, when you’re seeing growth rates in the hundreds of percent, these are on tens, or maybe hundreds of millions of dollars.”

High capex backing up spending on AI chips

Riley thinks that companies such as Amazon, Microsoft and Meta are contributing to the growth of the industry. Delano questioned Riley on the sustainability of that growth as some of Nvidia’s large clients may have order in a panic mode to avoid ending too far in the order queue.

Training is the concept of building the models and teaching them how to respond; inferencing is when they’re deployed for their end use case in the real world

Tom Riley, head of global thematic strategies , Axa IM

“In the semiconductor industry, this would be the concept of double ordering.” He does not think that these companies are inflating their orders above market demand “to get their place in the queue to secure the chips or maybe even to deny competitors the chips.”

Riley observed that Nvidia’s clients reported “elevated levels of capex intentions” (capital expenditures) which “are not trivial amounts… running in the billions of dollars.” Yet he admitted that “you don’t see that until sometimes it’s too late.”

Could the next leg of growth come from inferencing?

Riley explained that AI is split into two main concepts. “Training is the concept of building the models and teaching them how to respond; inferencing is when they’re deployed for their end use case in the real world.” He noted that Nvidia broke down their result for the first time to report that 60% of its sales is for training and 40% for inferencing.

Riley thinks that “we’re at the quite early stages of moving from developing these models to actually using them as business cases.” He gave the example of Microsoft Copilot application, an Office assistant launched recently.

Companies poised to benefit from the success of Nvidia

The ecosystem starts with their competitor AMD which is also surfing the current AI wave. TMSC from Taiwan is the chip manufacturing partner of Nvidia. Broadcom and Marvell are making “applications specific integrated circuits,” also known as ASICs, which are used for certain applications of AI.

Elsewhere, Riley noted the electronic design automation companies--Cadence and Synopsis--are beneficiaries of proliferation of semiconductors for artificial intelligence. They develop software and tools to design semiconductors. Outside Nvidia, ASML and Applied Materials are supporting the industry providing the tools to build chips.

Riley commented that ASML has some competition in older-generation legacy node chips, but ASML has a dominant position in the small node transistor which support AI development.

Reverberations further away

In the industrial space, Riley believes that companies such as Schneider Electric and Siemens are also poised to benefit from the spending on data centres. The same is also true for Daifuku from Japan which specialises in automation solutions for clean rooms with pure air to avoid the contamination of wafers in chip-building sites.

He thinks that a trend that is as significant as artificial intelligence impacts a wide range of applications and with the passage of time, it will continue to broaden the number companies benefiting from it or pivot their business models to embrace that opportunity.

After the unsuccessful and expensive venture of Meta, is Metaverse dead?

Riley explained that Metaverse was about “content creation in a virtual world and visualisation technologies, which were around kind of graphic rendering that are very compute intensive.” These technologies are relevant for the developments of artificial intelligence.

He expects the Metaverse to continue to gain traction in the gaming space. However, the industrial sector rather applies the concept of “digital twin” to digitally design a new factory or aircraft engines. He sees its relevance in real time monitoring and predictive maintenance among other possibilities.

In the healthcare space, a VR headset enables to track motion to improve rehabilitation. Following an operation, the technology tracks someone’s recovery more accurately and in real-time instead of going to the doctors once a fortnight.

Riley commented that these healthcare and industrial technologies have been around for more than 10 years but “the convergence of some of the visualisation and graphics technologies and artificial intelligence has driven use cases into new areas.”

This article was published for the Delano Finance newsletter, the weekly source for financial news in Luxembourg.  Subscribe using this link.