00:00Today on Forbes, desperate for power, AI companies look to the nuclear option.
00:08For all the hype and billions of dollars swirling around AI, the only certainty is the grid-depleting
00:13amount of additional electricity needed to power the data centers running the AI itself.
00:20And in that, Clayton Scott sees a massive business opportunity.
00:25Scott, chief commercial officer for NuScale Power, which is commercializing small-scale
00:29nuclear systems, told Forbes, quote,
00:40A Goldman Sachs report, which sees AI applications triggering a 160% jump in overall data center
00:47power needs, estimated that chat GPT queries need nearly 10 times as much electricity as
00:53Google searches.
00:54Pair that with the reality that the U.S. grid is at or near peak capacity relying on existing
01:00fossil fuel plants, with a backlog of wind and solar renewable energy projects waiting
01:05to be connected to the grid in an effort to slash carbon emissions.
01:08And you've got the recipe for an energy crisis.
01:12Scott hopes that NuScale Power provides one answer for data centers and their AI customers.
01:17Each of its small modular nuclear reactors, or SMRs, can generate 77 carbon-free megawatts
01:24continuously.
01:26That could allow them to operate independently of the grid entirely.
01:30The Portland-based startup, which went public in 2022 after over a decade of R&D, is working
01:36with data center developer Standard Power to supply 24 SMRs capable of collectively
01:42generating nearly 2 gigawatts of electricity, enough for a mid-sized city.
01:47Currently, it's navigating the Nuclear Regulatory Commission's lengthy review process and doesn't
01:51expect to deploy reactors until late this decade.
01:55The company reported just $4.6 million in revenue in the first quarter and a loss of
01:59$56.4 million as it prepares for commercial deliveries.
02:05NuScale is one of many nuclear and battery startups aiming to capitalize on AI's energy
02:10intensity.
02:11Typical data centers use about 32 megawatts of electricity, enough to power 6,000 homes,
02:17compared with about 80 megawatts for AI-oriented data centers.
02:22This according to Doug Vine, director of analysis at the Center for Climate and Energy Solutions,
02:27an environmental policy think tank in Arlington, Virginia.
02:31He said, quote,
02:32And they need energy-dense systems, which, unfortunately, is why natural gas plants are
02:37still popular.
02:38But the same thing is true of nuclear power.
02:41From a very small footprint, you can generate a lot of electricity.
02:46Armen Shahabi, a Lawrence Berkeley lab scientist who's tracked data center power demands since
02:512007, said that NVIDIA's H100 chip, loaded with eight GPUs and two CPUs, is the most
02:59in-demand AI server box for data centers.
03:02Shahabi estimates that each one uses 8 to 10 kilowatts of electricity, versus previous
03:07servers which would use a fraction of that.
03:10They generate so much heat that cooling them uses up to 30% more water than traditional
03:15data centers.
03:17Globally, AI demand may use up to 6.6 billion cubic meters of water in 2027, or according
03:24to a Cornell University study, quote,
03:26More than the total annual water withdrawal of Denmark, or half of the United Kingdom.
03:32Shahabi said, quote,
03:34They're so heat-dense and so power-dense, they're up in the range of 70 to 80 kilowatts
03:39per rack.
03:41And there's talk that's going to go up to 100 kilowatts really soon.
03:44It's basically 10 times higher than servers used in the past.
03:49Companies like Microsoft are also making plans to use SMRs to power their AI data centers.
03:55And company founder Bill Gates is chairman of TerraPower, which is developing a new kind
04:00of sodium reactor that also uses salt to store energy.
04:04OpenAI's Sam Altman is backing Oklo, a developer of small fission reactors that recently went
04:10public, as well as Helion, one of many startups attempting to commercialize nuclear fusion,
04:16a compelling carbon-free power source that's free of problematic nuclear waste.
04:21Still, it's unlikely any fusion-based systems will become commercially viable until the 2030s.
04:29For full coverage, check out Alan Onsman's piece on Forbes.com.
04:35This is Kieran Meadows from Forbes.
04:37Thanks for tuning in.
Comments