Nvidia Corp says the companies building data centres need to do so in ways that appeal to local communities.
"The way we think about AI and data centres in particular is that data centres can be – and should be – an asset to communities,” Josh Parker, head of sustainability at Nvidia, said on a Bloomberg Green panel on Nov 11 at COP31 in Türkiye.
"There are bad ways to build data centres, but there are many, many good ways to build data centres as well,” he said.
There’s growing pushback in the US and Europe against the build-out of infrastructure needed to power artificial intelligence (AI). The strength of community anger has even shifted the political landscape.
Last week’s tech-stock selloff was fuelled by anticipation that the results of the US midterms will lead to more regulations for AI, including tougher energy and environmental standards for data centres.
Nvidia, which dominates the supply of chips used to train and run AI models, has "been focusing on innovation,” Parker said. That means focusing on things that drive the "sustainability of data centres, energy efficiency in particular, but also things like water conservation.”
Investors in AI are having to navigate intense patches of volatility. The market is still trying to digest an admission from Nvidia chief executive officer Jensen Huang that he had an affair with a woman identified in several reports as a possible Chinese spy. Parker declined to comment on the matter, beyond saying Nvidia’s sustainability commitment is unchanged.
In the US, hyperscalers accounted for at least 75% of corporate clean energy purchases last year, according to Parker.
"Those investments are going to pay off,” he said. And the "demand signal that they’re sending right now” is that "we need more clean tech, we need more clean energy.”
Scaling AI will require huge investments in expanding and upgrading electric grids. Türkiye has made it a COP31 goal to get nations to agree to have 35% of global final energy consumption come from electricity by 2035. That compares with about 23% currently, according to a September estimate by the International Energy Agency (IEA).
Luke Nelson, head of international sustainability at JPMorgan Chase & Co, said the key to achieving that goal lies in piling investments into the grid, something the bank has already declared as a goal.
"The number one enabler that we see is actually the grid,” he said, speaking on the same panel as Parker. "You can’t achieve electrification goals without having a functioning grid. And we see that as a priority for achieving goals around energy security, affordability, reliability and ultimately sustainability as well.”
According to Parker, the end effect of such investments will be a better climate as AI opens the door to more efficient power systems.
"I think it’s inevitable, actually, that AI will help us drive net emissions reductions,” he said.
For now, there’s no definitive estimate of the climate footprint of AI. The technology’s rapidly growing need for electricity, however, is uncontested. In April, the IEA said electricity consumed by AI-focused data centres is set to triple by the end of the decade. By then, all data centres – including those that aren’t focused on AI – will account for about three per cent of global electricity demand, according to the IEA. – Bloomberg
