WEBVTT - How long can the AI boom last? 0:00:01.160 --> 0:00:10.120 I'm Ruby Jones and you're listening to seven AM. Companies 0:00:10.160 --> 0:00:13.000 are betting big on AI and of burning through cash 0:00:13.039 --> 0:00:16.320 to do it. They're pouring hundreds of billions of dollars 0:00:16.360 --> 0:00:19.800 into building data centers and developing new models such as 0:00:19.880 --> 0:00:23.520 chat GPT, and so far these projects are yet to 0:00:23.560 --> 0:00:28.560 see a financial return. There is one clear winner, however, Nvidia, 0:00:28.960 --> 0:00:31.560 the company that makes the chips that power this tech. 0:00:32.400 --> 0:00:35.239 But how long can its customers keep spending before they 0:00:35.280 --> 0:00:39.120 have to either turn a profit or scale back. Today 0:00:39.240 --> 0:00:42.880 Reuter's journalist Stephen nellis on how long this AI boom 0:00:42.920 --> 0:00:52.920 can last and what a crash might look like. It's Thursday, 0:00:53.040 --> 0:01:02.400 November twenty seven, So Stephen, Nvidia had its latest earnings call, 0:01:02.520 --> 0:01:05.600 and you were listening in on that call, So tell 0:01:05.640 --> 0:01:06.679 me as it happened. 0:01:07.080 --> 0:01:08.920 What will you listening out for. 0:01:09.480 --> 0:01:11.800 I think the biggest thing we're listening for is any 0:01:11.959 --> 0:01:15.760 sign that the AI bubble is near bursting, wif there 0:01:15.840 --> 0:01:18.160 is a bubble at all, So any sign of a 0:01:18.280 --> 0:01:22.920 lack of demand, any sign of faltering interest in videos, 0:01:23.000 --> 0:01:26.160 chips or an AI as a field. And largely we 0:01:26.240 --> 0:01:28.920 didn't hear any of that. What we heard was a 0:01:29.040 --> 0:01:32.600 forecast that was better than Wall Street analysts were expecting, 0:01:33.440 --> 0:01:36.600 revenue that came out better than Wall Street was expecting, 0:01:37.160 --> 0:01:40.039 and a lot of talk about how it's not an 0:01:40.080 --> 0:01:44.720 AI bubble, but instead a tipping point toward an era 0:01:44.800 --> 0:01:48.120 of AI that in Video believes will be much more durable. 0:01:47.800 --> 0:01:51.120 Social intelligence chip maker in Videos latest earnings report came 0:01:51.160 --> 0:01:53.440 in and higher than Wall Street expected. 0:01:53.560 --> 0:01:55.160 The world's most valuable. 0:01:54.760 --> 0:01:57.760 Company reported fifty seven billion dollars in third quarter revenue. 0:01:57.880 --> 0:02:00.559 And so how much is Nvideo worth right now? 0:02:00.800 --> 0:02:03.440 Well, it depends on the day, but they're fairly consistently 0:02:03.560 --> 0:02:07.320 over four point five trillion dollars, and at various times 0:02:07.480 --> 0:02:10.560 during the year they've been not just the most valuable 0:02:10.639 --> 0:02:15.880 publicly traded company currently, but the most valuable publicly traded company. 0:02:15.600 --> 0:02:17.320 In the history of business. 0:02:17.840 --> 0:02:20.799 And that's largely due to the unique place that they 0:02:20.840 --> 0:02:25.280 occupy in the AI ecosystem, which is that some eighty 0:02:25.320 --> 0:02:29.000 to ninety percent of the market is their chips training 0:02:29.360 --> 0:02:32.680 these systems that many people use every day. So whether 0:02:32.720 --> 0:02:36.920 that's chat, GPT, whether that's some other app that you 0:02:37.040 --> 0:02:40.480 use that has AI working in the background, the overwhelming 0:02:40.560 --> 0:02:43.960 odds are that it exists because it was trained on 0:02:44.120 --> 0:02:46.680 data that was run through in Video's chips. 0:02:46.960 --> 0:02:50.320 Okay, so in Vidia has the monopoly on these chips. 0:02:50.360 --> 0:02:51.399 Who is buying them? 0:02:51.639 --> 0:02:54.320 So the biggest buyers and this was an interesting thing 0:02:54.360 --> 0:02:57.360 to come out of the earning support. There are four customers. 0:02:57.360 --> 0:02:59.560 We don't know precisely who they are, but they make 0:02:59.639 --> 0:03:02.920 up sick nexty one percent of in Video's revenue. And 0:03:03.000 --> 0:03:06.480 the most likely candidates for that are some of the 0:03:06.560 --> 0:03:09.360 names that you're probably familiar with. Open Ai, which is 0:03:09.400 --> 0:03:15.000 the creator of chat, GPT, Meta which owns Instagram, Facebook, WhatsApp, 0:03:15.040 --> 0:03:19.000 and other properties. And then other firms like Microsoft, which 0:03:19.080 --> 0:03:22.720 is deploying artificial intelligence. 0:03:22.200 --> 0:03:24.040 Across its apps like Office. 0:03:24.360 --> 0:03:26.720 But they're all snapping them up as quickly as possible 0:03:26.720 --> 0:03:30.280 and trying to string them together into huge piles of 0:03:30.480 --> 0:03:33.280 tens of thousands. Some are even talking about hundreds of 0:03:33.280 --> 0:03:36.640 thousands of chips at once, so they are buying in bulk. 0:03:36.760 --> 0:03:40.280 Okay, So you've got those companies spending huge sums on 0:03:40.600 --> 0:03:43.560 chips that they're buying from Video. But many of these 0:03:43.640 --> 0:03:46.760 AI companies are not actually making a profit yet, So 0:03:46.880 --> 0:03:48.080 tell me about this gamble. 0:03:48.400 --> 0:03:50.880 So right now we're in an interesting time where a 0:03:50.960 --> 0:03:53.560 lot of people are trying these things. I mean, it'd 0:03:53.560 --> 0:03:55.960 be hard to find a listener who hasn't chatted with 0:03:56.040 --> 0:03:59.680 some kind of a chatbot, clicked an image generation button somewhere, 0:03:59.800 --> 0:04:02.640 or even just had something read over a passage or 0:04:02.680 --> 0:04:04.760 a sentence and see if it can improve it. Now, 0:04:04.800 --> 0:04:07.200 the question is, when you did that, how much did 0:04:07.240 --> 0:04:09.320 you pay for it? And if you didn't pay for 0:04:09.400 --> 0:04:13.320 it directly, how many advertisements did you sit through before 0:04:13.360 --> 0:04:15.800 you use that product. One of the issues that we 0:04:15.880 --> 0:04:18.320 have right now is there's a big question all over 0:04:18.360 --> 0:04:20.560 the tech world about how much will people pay for 0:04:20.600 --> 0:04:24.880 these things directly based on how much they perceive it's worth. Now, 0:04:24.960 --> 0:04:27.200 let's talk about the cost side, though, because that's I 0:04:27.240 --> 0:04:29.560 think what's really driving a lot of the lack of 0:04:29.560 --> 0:04:33.200 profitability for the time being, and that is that these 0:04:33.240 --> 0:04:36.240 chips are very, very expensive. They tend to sell them 0:04:36.240 --> 0:04:38.719 in a big server rack that will have something like 0:04:39.160 --> 0:04:42.400 seventy two are coming soon more than one hundred of 0:04:42.400 --> 0:04:46.520 these chips, and those are many hundreds of thousands of 0:04:46.600 --> 0:04:49.640 US dollars, if not into the millions of dollars for 0:04:49.680 --> 0:04:53.000 a single server. So they are expensive and the electricity 0:04:53.040 --> 0:04:55.920 did to operate them is very expensive. You've probably read 0:04:56.000 --> 0:04:58.680 that many of these data centers take as much electricity 0:04:58.720 --> 0:05:01.279 as a small city. So what you have is a 0:05:01.320 --> 0:05:04.359 really simple situation where it costs a lot of money 0:05:04.400 --> 0:05:07.520 to make the thing, and it's unclear exactly how much 0:05:07.560 --> 0:05:09.200 people are willing to pay for that thing in the 0:05:09.240 --> 0:05:09.880 short term. 0:05:09.960 --> 0:05:12.080 What is the path to profitability? When do you see 0:05:12.080 --> 0:05:13.360 the lines cross? If you will? 0:05:13.560 --> 0:05:16.520 I mean, I think it could have happened sooner than 0:05:16.600 --> 0:05:18.520 I originally thought, if we wanted it too. 0:05:18.880 --> 0:05:21.279 But it seems to me like the right thing to 0:05:21.320 --> 0:05:23.240 do is to just keep investing in computing. 0:05:23.279 --> 0:05:25.920 So when we look at companies like open ai, there 0:05:25.920 --> 0:05:28.799 have been some reports that they generated about four point 0:05:28.839 --> 0:05:31.360 three billion dollars in revenue for the first half of 0:05:31.360 --> 0:05:34.160 the year and on on track to hit their revenue 0:05:34.200 --> 0:05:37.240 target of thirteen billion dollars for twenty twenty five. But 0:05:37.839 --> 0:05:39.960 that's with a cash burn of about eight and a 0:05:40.040 --> 0:05:43.160 half billion, and that doesn't even take into account all 0:05:43.200 --> 0:05:45.080 of the other operational costs. 0:05:45.160 --> 0:05:47.960 You know, they have what ninety percent of search queries Okay, 0:05:48.000 --> 0:05:49.720 they already have that was never going to get better 0:05:49.720 --> 0:05:52.920 than that. They have seventy percent of like search ad revenue. Right, 0:05:52.960 --> 0:05:54.880 you can see how this company is deemed to be 0:05:55.040 --> 0:05:55.680 a monopoly. 0:05:56.040 --> 0:05:58.800 And then we can contrast that with companies like Microsoft 0:05:58.880 --> 0:06:02.119 or Google, which are spending lots and lots of money 0:06:02.120 --> 0:06:05.480 their capital lot ways, you know, are are north in 0:06:05.560 --> 0:06:08.040 the tens of billions of dollars and cumulatively in the 0:06:08.160 --> 0:06:10.599 hundreds of billions of dollars. But in the case of 0:06:10.680 --> 0:06:15.159 those broader companies, they're suffusing that AI into other products 0:06:15.160 --> 0:06:17.640 that already make money and just charging more for it. 0:06:17.720 --> 0:06:21.360 So we have seen companies like Google and Microsoft remain profitable, 0:06:21.600 --> 0:06:24.719 and then they also just have very large businesses renting 0:06:24.720 --> 0:06:27.839 out their data centers as sort of picks and shovels 0:06:27.880 --> 0:06:30.839 to other companies trying to cash in on this AI 0:06:30.920 --> 0:06:33.599 gold rush, and that's still a pretty profitable business for 0:06:33.640 --> 0:06:36.800 those companies as well. So I think the profitability question 0:06:36.920 --> 0:06:39.560 ultimately comes down to are you trying to make money 0:06:40.120 --> 0:06:44.279 just generating the technology and monetizing it, or do you 0:06:44.480 --> 0:06:47.200 spread it out over other stuff that's already making money. 0:06:48.000 --> 0:06:51.400 Okay, but Nvidia is doing great of fall of this, 0:06:51.480 --> 0:06:54.800 as you say, the most valuable listed company in human 0:06:54.880 --> 0:06:55.640 history at the moment. 0:06:56.320 --> 0:06:57.480 But will that last? 0:06:57.760 --> 0:07:01.320 That is the four point five billion dollar question, and 0:07:01.400 --> 0:07:04.200 I think the answer is a complicated one. So in 0:07:04.279 --> 0:07:07.320 Video's chips are still in high, high demand, and what 0:07:07.360 --> 0:07:10.920 that means is that in Vidia has a lot of leeway. 0:07:10.520 --> 0:07:11.920 In how they price those things. 0:07:12.320 --> 0:07:14.120 Now, the flip side of this is, you know, you 0:07:14.200 --> 0:07:17.520 think about other big semiconductor companies over time. One that 0:07:17.560 --> 0:07:20.120 people might have heard of from a previous generation would 0:07:20.160 --> 0:07:23.559 be Intel. So back in the nineteen eighties, Intel really 0:07:23.680 --> 0:07:27.600 jumped into the personal computer market. They had a very 0:07:27.680 --> 0:07:30.400 very durable run for thirty or forty years. Now, is 0:07:30.400 --> 0:07:33.320 in Vidia going to last that long? It depends on 0:07:33.360 --> 0:07:36.200 whether the kind of computing that we do the most 0:07:36.240 --> 0:07:38.760 of in the next few decades is the kind that 0:07:38.800 --> 0:07:41.200 in Vidia is good at or some other kind. I 0:07:41.200 --> 0:07:43.800 think that question is far from settled. You know, right 0:07:43.800 --> 0:07:46.040 now we're in the phase we're building the AI, we're 0:07:46.120 --> 0:07:49.720 training the AI. In Vidia has a monopoly on that market. 0:07:49.800 --> 0:07:52.559 But in terms of actually deploying the AI, putting AI 0:07:52.600 --> 0:07:55.200 in your computer at home, putting AI in some kind 0:07:55.240 --> 0:07:58.560 of widget that sits on your countertop, those chips are 0:07:59.080 --> 0:08:01.400 not all made by and Video, and that market is 0:08:01.560 --> 0:08:03.280 much more competitive in the long. 0:08:03.160 --> 0:08:08.840 Term coming up, could the AI bubble trigger a financial crisis? 0:08:18.040 --> 0:08:20.160 So Stephen, we have these companies. 0:08:19.720 --> 0:08:24.280 Open AI, Microsoft, Google, They obviously see the value in 0:08:24.280 --> 0:08:27.960 investing in this technology, even if it isn't making them 0:08:28.000 --> 0:08:31.400 a return at this stage. So how are these companies 0:08:31.440 --> 0:08:34.400 trying to crack this problem of profitability? 0:08:34.679 --> 0:08:37.960 So the biggest thing is going to things that cost 0:08:38.120 --> 0:08:41.120 a lot of money to companies right now. And one 0:08:41.160 --> 0:08:43.120 of the things these companies spend the most money on 0:08:43.240 --> 0:08:46.480 is engineers who write software code right, And one of 0:08:46.520 --> 0:08:48.280 the things that they're trying to make money off of 0:08:48.400 --> 0:08:51.920 is going to other companies in the software business and saying, listen, 0:08:52.280 --> 0:08:54.640 you can write in the case of Microsoft, as much 0:08:54.640 --> 0:08:57.320 as thirty percent of your code using some of these 0:08:57.360 --> 0:09:01.240 AI tools and make your team seems much more efficient. 0:09:01.600 --> 0:09:04.480 So in the short term, I think a lot of 0:09:04.520 --> 0:09:10.320 it is automating white collar work, professional work that companies 0:09:10.360 --> 0:09:12.719 currently spend a lot of money paying humans to do. 0:09:13.640 --> 0:09:16.600 In the longer term, I think it's a much more 0:09:16.640 --> 0:09:20.920 mixed bag where some of the interesting scientific use cases, 0:09:20.960 --> 0:09:24.360 so developing new drugs and things like that, I think 0:09:24.400 --> 0:09:27.520 we'll have to see for those but the longer horizon 0:09:27.640 --> 0:09:30.120 is making new things that we really didn't think we 0:09:30.120 --> 0:09:32.960 can make or wouldn't be possible to make without this technology. 0:09:33.360 --> 0:09:35.600 And I think most of these companies think that you're 0:09:35.600 --> 0:09:39.880 going to have some kind of personal assistant, that we'll 0:09:39.880 --> 0:09:41.439 be able to take care of a lot of things 0:09:41.520 --> 0:09:45.200 in your behalf, and that eventually, if in video's grandest 0:09:45.320 --> 0:09:48.960 visions ever come to pass, that virtual assistant would have 0:09:49.040 --> 0:09:51.960 some kind of physical manifestation, a robot that can do 0:09:52.000 --> 0:09:56.040 the things around your house that you want to be done. Now, 0:09:56.240 --> 0:09:58.800 we should say we're really far off from that right now. 0:09:59.000 --> 0:10:02.040 I don't think you should and plan on having your 0:10:02.080 --> 0:10:06.160 floor scrubbed or your toilets cleaned by robots anytime soon, 0:10:06.200 --> 0:10:07.800 but it is something they're thinking about. 0:10:08.120 --> 0:10:10.800 And there is so much speculation about the growth of 0:10:10.880 --> 0:10:14.760 these companies and the future of AI and fears that 0:10:14.800 --> 0:10:17.720 this could all be a bubble that is set to bursts. 0:10:17.800 --> 0:10:20.160 So what are your thoughts on that and where this 0:10:20.200 --> 0:10:20.800 is all headed. 0:10:21.400 --> 0:10:25.440 One thing the markets are very good at is making 0:10:25.480 --> 0:10:28.640 a judgment about profitability over some kind of term that's 0:10:28.640 --> 0:10:30.800 not that far out. So I would say we probably 0:10:30.840 --> 0:10:32.599 have a few years for some of these companies to 0:10:32.640 --> 0:10:35.760 really start showing a profit, and if they don't, the 0:10:36.200 --> 0:10:39.480 question is whether all the investment and sustaining that will 0:10:39.520 --> 0:10:41.920 be worth it. You know, if you sort of look 0:10:41.960 --> 0:10:45.720 at two previous global financial catastrophes, the dot com bubble 0:10:46.400 --> 0:10:49.280 and then the financial crisis that followed it, the dot 0:10:49.320 --> 0:10:53.200 com bubble was bad, but because it was largely limited 0:10:53.240 --> 0:10:56.559 to the stock market, the societal fallout wasn't that bad, 0:10:56.600 --> 0:10:59.080 at least in the United States. It's a very different 0:10:59.160 --> 0:11:02.240 case in the financial crisis, where that really wrapped up 0:11:02.240 --> 0:11:04.760 a lot of people's homes and their personal livelihood in 0:11:04.760 --> 0:11:07.199 a much more direct way. It's a tricky one to 0:11:07.240 --> 0:11:09.640 try to understand what the AI bubble will because, on 0:11:09.679 --> 0:11:11.960 the one hand, most of the wild wealth creation we've 0:11:11.960 --> 0:11:14.920 seen has been, for better or worse, limited to a 0:11:15.080 --> 0:11:19.600 very small upper strata of US investors. On the other hand, 0:11:20.120 --> 0:11:22.000 if you look at the economic data in the United 0:11:22.040 --> 0:11:24.880 States right now, a lot of what is propping up 0:11:25.200 --> 0:11:29.040 the economy is spending on this boom, and that spending 0:11:29.120 --> 0:11:31.360 is going to people who work in the trades, who 0:11:31.440 --> 0:11:34.960 work in pipe fitting, who are electricians. There's an electrician 0:11:34.960 --> 0:11:37.280 shortage right now. If you are an electrician. You should 0:11:37.280 --> 0:11:39.880 probably get on a plane and come to the United 0:11:39.920 --> 0:11:42.600 States if your codes in your country are similar, because 0:11:42.600 --> 0:11:43.720 you will probably find work. 0:11:44.120 --> 0:11:47.120 So the crash, if there is one, could affect ordinary people. 0:11:47.840 --> 0:11:50.480 The one thing that is apparent is that if you're 0:11:50.559 --> 0:11:54.640 just your average school teacher, firefighter, police officer who has 0:11:54.679 --> 0:11:58.120 a pension fund that is funded broadly by the US 0:11:58.160 --> 0:12:01.840 stock market, there is no from this because most of 0:12:01.880 --> 0:12:05.240 the gains in the stock market have been coming from 0:12:05.679 --> 0:12:09.280 this handful of firms that are really concentrated in the 0:12:09.280 --> 0:12:12.200 AI market. But I don't think it's quite like the 0:12:12.280 --> 0:12:14.240 nineteen twenties where a lot of people went out and 0:12:14.280 --> 0:12:17.520 borrowed money to buy stock in these companies. But the 0:12:17.640 --> 0:12:20.320 question is how much of it is paper money versus 0:12:20.400 --> 0:12:22.920 you know, people's homes and livelihoods every day, and that's 0:12:22.960 --> 0:12:24.079 just much harder to predict. 0:12:27.960 --> 0:12:29.800 Well, Stephen, thank you so much for your time. 0:12:30.000 --> 0:12:30.880 Thank you for having me. 0:12:45.320 --> 0:12:45.920 Also in the. 0:12:45.880 --> 0:12:49.400 News, economists predict a December interest rate cut is looking 0:12:49.520 --> 0:12:54.360 increasingly unlikely following the latest inflation data. The Bureau of 0:12:54.400 --> 0:12:57.880 Statistics consumer price index rose from three point six percent 0:12:57.920 --> 0:13:01.360 in September to three point eight percent in October. The 0:13:01.440 --> 0:13:05.640 largest contributors to inflation were housing, food and non alcoholic beverages. 0:13:06.720 --> 0:13:09.440 And for the first time ever, Australian voters see Labor 0:13:09.480 --> 0:13:12.680 as a better economic manager than the Coalition, according to 0:13:12.720 --> 0:13:16.199 the results of the Australian Election Study. The study, led 0:13:16.240 --> 0:13:19.840 by the Australian National University and Griffith University, is a 0:13:19.880 --> 0:13:23.600 post election survey of voters' attitudes and behavior and found 0:13:23.640 --> 0:13:26.680 that voters preferred Labour's policies on nine out of ten 0:13:26.760 --> 0:13:30.040 issues looked at in the study. The results also found 0:13:30.040 --> 0:13:32.560 that Peter Dutton was the least popular leader to go 0:13:32.600 --> 0:13:35.920 to a federal election since the survey began in nineteen 0:13:35.960 --> 0:13:39.520 eighty seven. I'm Ruby Jones. This is seven am. Thanks 0:13:39.559 --> 0:13:40.040 for listening.