WEBVTT - Here's Why AI Might Need Nuclear Power to Succeed

0:00:02.440 --> 0:00:09.440
<v Speaker 1>Bloomberg Audio Studios, Podcasts, radio news. I'm Stephen Carol and

0:00:09.560 --> 0:00:12.039
<v Speaker 1>this is Here's Why, where we take one news story

0:00:12.080 --> 0:00:14.159
<v Speaker 1>and explain it in just a few minutes with our

0:00:14.200 --> 0:00:22.000
<v Speaker 1>experts here at Bloomberg. Over the roughly seventy years since

0:00:22.040 --> 0:00:25.520
<v Speaker 1>the first commercial nuclear power plant went into operation, the

0:00:25.560 --> 0:00:30.360
<v Speaker 1>popularity of splitting atoms for electricity has undergone some major swings.

0:00:30.800 --> 0:00:34.320
<v Speaker 1>After the Fukushima disaster in twenty eleven, countries including Japan

0:00:34.360 --> 0:00:37.320
<v Speaker 1>and Germany pulled back from the power source. But since

0:00:37.320 --> 0:00:40.440
<v Speaker 1>the war in Ukraine, it's experiencing a comeback as a

0:00:40.479 --> 0:00:43.960
<v Speaker 1>secure energy source, and some big names in tech are

0:00:44.000 --> 0:00:44.960
<v Speaker 1>getting in on the act.

0:00:45.440 --> 0:00:48.000
<v Speaker 2>Nuclear is a wonderful way for as one of the

0:00:48.040 --> 0:00:50.840
<v Speaker 2>sources of energy, one of the sources of sustainable energy.

0:00:51.040 --> 0:00:54.960
<v Speaker 3>Tech companies wanted the customer for green energy, and at

0:00:55.040 --> 0:00:57.000
<v Speaker 3>least for a period of time, they're willing to pay

0:00:57.040 --> 0:01:00.760
<v Speaker 3>a bit of a premium that is extremely helpful to

0:01:00.880 --> 0:01:02.440
<v Speaker 3>reboot the nuclear industry.

0:01:02.640 --> 0:01:04.640
<v Speaker 2>It's totally fair to say that AI is going to

0:01:04.680 --> 0:01:06.680
<v Speaker 2>need a lot of energy, but it will force us,

0:01:06.720 --> 0:01:09.560
<v Speaker 2>i think, to invest more in the technologies that can

0:01:09.640 --> 0:01:10.119
<v Speaker 2>deliver this.

0:01:10.720 --> 0:01:13.759
<v Speaker 1>That was in videos. CEO Jensen h Wang, Microsoft founder

0:01:13.800 --> 0:01:17.720
<v Speaker 1>Bill Gates, and Open AI's Sam Altman. There, so, here's

0:01:17.760 --> 0:01:24.399
<v Speaker 1>why artificial intelligence might need nuclear power to succeed. Bloomberger

0:01:24.440 --> 0:01:28.400
<v Speaker 1>porter Jonathan Tyrone joins me now to explain. Jonathan, first

0:01:28.440 --> 0:01:31.760
<v Speaker 1>of all, how big is the renewed interest we've seen

0:01:32.040 --> 0:01:33.920
<v Speaker 1>in nuclear power in recent years.

0:01:34.200 --> 0:01:37.760
<v Speaker 2>Oh, there's definitely legitimate growing interest in nuclear power. You

0:01:37.760 --> 0:01:40.000
<v Speaker 2>can see that in opinion polling. You can also see

0:01:40.000 --> 0:01:43.400
<v Speaker 2>that in investment in the Western world that's mostly at

0:01:43.400 --> 0:01:47.320
<v Speaker 2>this point flown into extending the lifetime of existing reactors.

0:01:47.440 --> 0:01:50.680
<v Speaker 2>The biggest builds of new reactors are occurring in China.

0:01:50.760 --> 0:01:52.960
<v Speaker 2>They have twenty new builds going on right now. That's

0:01:53.080 --> 0:01:55.840
<v Speaker 2>unprecedented anywhere else in the world. But then you know

0:01:56.080 --> 0:02:00.720
<v Speaker 2>emerging market countries arranging from Egypt to Bangladesh to India,

0:02:00.880 --> 0:02:04.120
<v Speaker 2>there are also new builds going on. The places that

0:02:04.560 --> 0:02:07.640
<v Speaker 2>really are legging behind right now are Europe and especially

0:02:07.680 --> 0:02:11.280
<v Speaker 2>the US, where there are currently zero new builds happening.

0:02:11.760 --> 0:02:15.400
<v Speaker 1>Okay, it's interesting to understand that global spread. Part of

0:02:15.440 --> 0:02:18.600
<v Speaker 1>the future conversation, or the conversation around the future of

0:02:18.720 --> 0:02:23.280
<v Speaker 1>nuclear power has been around what are called small modular reactors.

0:02:23.440 --> 0:02:25.520
<v Speaker 1>Can you just explain what are they and why are

0:02:25.520 --> 0:02:26.560
<v Speaker 1>they proving popular?

0:02:27.000 --> 0:02:27.280
<v Speaker 3>Sure?

0:02:27.400 --> 0:02:30.280
<v Speaker 2>I mean, small modular reactors are not a new concept.

0:02:30.320 --> 0:02:33.200
<v Speaker 2>People have been talking about small modular reactors for twenty

0:02:33.280 --> 0:02:36.400
<v Speaker 2>years or more. And you know these are factory assembled

0:02:36.639 --> 0:02:40.040
<v Speaker 2>models that would be around three hundred megawatts. That's about

0:02:40.080 --> 0:02:43.000
<v Speaker 2>the third of the size of a traditional reactor, which

0:02:43.200 --> 0:02:46.200
<v Speaker 2>you know is going around one thousands of megawatts or

0:02:46.200 --> 0:02:48.680
<v Speaker 2>one gigawatt. The idea behind them is that they can

0:02:48.720 --> 0:02:52.040
<v Speaker 2>be built in a factory, inter creal production and therefore

0:02:52.240 --> 0:02:56.720
<v Speaker 2>lower cost. Right now, there are more than ninety individual

0:02:56.760 --> 0:03:01.160
<v Speaker 2>designs all competing in the marketplace for acceptance in the market,

0:03:01.280 --> 0:03:03.720
<v Speaker 2>and this is actually one of the big problems. There's

0:03:03.760 --> 0:03:05.919
<v Speaker 2>a lot of investment going in what you would call

0:03:05.960 --> 0:03:09.000
<v Speaker 2>paper engineering. The only countries that have actually built in

0:03:09.160 --> 0:03:12.919
<v Speaker 2>SMR are China and Russia. The US and Europe. Again,

0:03:13.000 --> 0:03:16.560
<v Speaker 2>they have scores of different designs but have been less

0:03:16.560 --> 0:03:18.400
<v Speaker 2>effective in actually implementing them.

0:03:18.600 --> 0:03:22.880
<v Speaker 1>Do SMRs provide a solution for countries who are worried

0:03:22.919 --> 0:03:28.080
<v Speaker 1>about energy security if they're cheaper and faster to build well.

0:03:29.160 --> 0:03:32.240
<v Speaker 2>And in security is an interesting concept. It's true that

0:03:32.520 --> 0:03:36.360
<v Speaker 2>nuclear reactor can run anywhere from eighteen months to a

0:03:36.400 --> 0:03:40.800
<v Speaker 2>couple of years without receiving new fuel. However, nuclear reactors,

0:03:40.840 --> 0:03:43.440
<v Speaker 2>just like gas plants or coal plants, do need to

0:03:43.440 --> 0:03:48.160
<v Speaker 2>be refueled intermittently, and that fuel cycle is very complex

0:03:48.400 --> 0:03:52.600
<v Speaker 2>with very long supply chains. And that's also an issue

0:03:52.640 --> 0:03:54.880
<v Speaker 2>that we see in the US right now, where the

0:03:55.000 --> 0:03:58.080
<v Speaker 2>US continues to get a large share of its nuclear

0:03:58.080 --> 0:04:01.640
<v Speaker 2>fuel from Russia. That's the same problem that we witness

0:04:01.720 --> 0:04:06.240
<v Speaker 2>in Eastern Europe. So yes, SMRs and nuclear could give

0:04:06.400 --> 0:04:11.560
<v Speaker 2>more base in times of crises to change your supply chain. However,

0:04:11.560 --> 0:04:15.000
<v Speaker 2>in terms of energy security in this complex world of trade,

0:04:15.280 --> 0:04:18.880
<v Speaker 2>they are also prone to uncertainty.

0:04:19.160 --> 0:04:22.440
<v Speaker 1>So that's the supply side of the equation. But let's

0:04:22.440 --> 0:04:25.760
<v Speaker 1>talk a bit about the demand. What AI got to

0:04:25.760 --> 0:04:28.839
<v Speaker 1>do with this renewed interest in nuclear power.

0:04:29.160 --> 0:04:33.200
<v Speaker 2>Well, the AI companies, to their credit, are creating demand

0:04:33.200 --> 0:04:37.440
<v Speaker 2>for clean energy, and undoubtedly nuclear power is a much

0:04:37.520 --> 0:04:41.839
<v Speaker 2>cleaner energy source than alternative fossil fuels. And so what

0:04:42.000 --> 0:04:46.000
<v Speaker 2>the tech companies want to do is called demand aggregation,

0:04:46.279 --> 0:04:49.800
<v Speaker 2>and they are looking at putting together all of their

0:04:50.000 --> 0:04:55.400
<v Speaker 2>increased electricity demands and working with utilities to create power

0:04:55.440 --> 0:04:59.080
<v Speaker 2>price agreements. These are packages that can be financialized and

0:04:59.120 --> 0:05:04.280
<v Speaker 2>then used to finance new builds power. The issue is

0:05:04.320 --> 0:05:07.480
<v Speaker 2>that when you look at long term or middle term

0:05:07.680 --> 0:05:13.200
<v Speaker 2>electricity demand forecasts, you're still looking at data centers coming

0:05:13.240 --> 0:05:15.440
<v Speaker 2>in and around one to one point three percent of

0:05:15.520 --> 0:05:20.120
<v Speaker 2>total global electricity consumption. And even with the onset of

0:05:20.200 --> 0:05:24.839
<v Speaker 2>artificial intelligence, the growth of electricity demand for data centers

0:05:24.920 --> 0:05:27.800
<v Speaker 2>is projected to be rapid, but the level set to

0:05:27.839 --> 0:05:31.120
<v Speaker 2>remain relatively small. That's coming from the IEA. And then

0:05:31.160 --> 0:05:34.760
<v Speaker 2>you put that into comparison with other demand sources and

0:05:34.800 --> 0:05:39.960
<v Speaker 2>we're looking at desalination electric vehicles and it becomes more

0:05:40.000 --> 0:05:44.719
<v Speaker 2>of a story about Silicon Valley and Western markets trying

0:05:44.760 --> 0:05:48.960
<v Speaker 2>to reboot a clean energy source nuclear power that has

0:05:49.000 --> 0:05:51.080
<v Speaker 2>fallen by the wayside in recent decades.

0:05:51.440 --> 0:05:55.160
<v Speaker 1>And are the investments that we're seeing companies like Microsoft

0:05:55.160 --> 0:05:57.919
<v Speaker 1>and Google making going to make a difference in this.

0:05:58.320 --> 0:06:01.680
<v Speaker 2>Well, the nuclear injurt you certainly hope so, but again

0:06:01.760 --> 0:06:05.279
<v Speaker 2>we're looking at a demand deficit. And I'll cite again

0:06:05.600 --> 0:06:09.560
<v Speaker 2>next Terra Energy. This is a pro nuclear utility. They

0:06:09.720 --> 0:06:13.520
<v Speaker 2>operate nuclear power plants in the US, but even its

0:06:13.720 --> 0:06:17.039
<v Speaker 2>chief executive officer, John Ketchum, came out last week during

0:06:17.160 --> 0:06:20.040
<v Speaker 2>his quarterly earnings conference call and said that he's simply

0:06:20.040 --> 0:06:26.400
<v Speaker 2>not bullish on SMRs because while they conceivably could provide

0:06:26.680 --> 0:06:29.960
<v Speaker 2>a service of clean energy or these AI data centers,

0:06:30.080 --> 0:06:33.520
<v Speaker 2>their actual implementation and deployment is so far out he

0:06:33.720 --> 0:06:37.599
<v Speaker 2>estimates towards the second half of the twenty thirties that

0:06:37.640 --> 0:06:42.200
<v Speaker 2>they're not practically placed to serve the increased data needs

0:06:42.600 --> 0:06:45.960
<v Speaker 2>that Silicon Valley needs right now, and so he's looking

0:06:46.080 --> 0:06:49.880
<v Speaker 2>at more traditional sources of nuclear energy, such as extending

0:06:49.880 --> 0:06:53.760
<v Speaker 2>the lifetime of existing plants and then filling in generation

0:06:53.920 --> 0:06:58.760
<v Speaker 2>capacity of variable wind and solar with peak gas generation plants.

0:06:59.160 --> 0:07:02.560
<v Speaker 2>So power there's a lot of interest, there's a lot

0:07:02.600 --> 0:07:05.880
<v Speaker 2>of great ideas out there, but it still is facing

0:07:06.080 --> 0:07:08.000
<v Speaker 2>a big challenge when it comes to implementation.

0:07:08.400 --> 0:07:12.160
<v Speaker 1>Thanks to Bloomberg reporter Jonathan Tyrone. For more explanations like

0:07:12.200 --> 0:07:14.480
<v Speaker 1>this from our team of twenty seven hundred journalists and

0:07:14.520 --> 0:07:16.720
<v Speaker 1>analysts around the world. Search for quick Take on the

0:07:16.760 --> 0:07:21.800
<v Speaker 1>Bloomberg website or Bloomberg Business app. I'm Stephen Carol. This

0:07:21.880 --> 0:07:24.360
<v Speaker 1>is Here's why. I'll be back next week with more.

0:07:24.560 --> 0:07:25.400
<v Speaker 1>Thanks for listening.