- 15 hours ago
Category
🗞
NewsTranscript
00:00As Kadam Akunda writes, innovation doesn't only need the technology, but also the physical infrastructure to support the product.
00:05Kadam is a columnist for Bloomberg Opinion and a lecturer at the Yale School of Management.
00:09He joins us now on set here in New York. Wonderful to see you.
00:13Let's talk about this, the effervescence, the enthusiasm surrounding AI, meeting the reality of having to pay for it.
00:18And there are extreme things that I think of, the amount of money, for instance, that Meta has poured into
00:23this.
00:23It's begun to retrench some of that here.
00:25How widely spread is that?
00:27Are a lot of companies kind of reckoning with, what is this going to be, and is it worth the
00:31investment at this point in time?
00:33Great to be here.
00:34So, yes, but there's a nuance here, right?
00:37So, there's a company I was talking to, the CEO yesterday.
00:40He said, if Claude charged us 10 times what it does right now, he said, I would pay it, and
00:46I wouldn't hesitate.
00:47He's like, I wouldn't even think twice.
00:48But what kind of company is that, and what is the application?
00:51It's a small health care company.
00:52So, there's a big difference when you have 10,000 engineers who are running, you know, $2,000 per head.
00:59If you go 10x, right, things change when you're at that scale.
01:03But I think what's really, what's driving this is, it's really interesting that the debate on AI is kind of
01:08missing something pretty big.
01:09And that, I think, is what this cost thing is starting to drive us towards.
01:13Because, you see, there are the boosters, and, you know, I used to code.
01:17The ability to talk to my computer and have it do something without is genuinely, it feels miraculous.
01:22Fantastic, yes.
01:23It's astonishing.
01:24And then there are people, some of them will be people with enormous scientific and technical credibility who say, this
01:28is not going to come together the way you think.
01:30Like, this technology doesn't have the potential you think.
01:32There's hype.
01:32Those are also serious people.
01:34You shouldn't dismiss them.
01:35Not doomers per se, but skeptics of this, yeah.
01:37There's a third scenario, which it just strikes me no one talks about, which is you can have a revolutionary,
01:43world-changing, astonishing technology that doesn't end up making anyone any money.
01:48And this is the thing that people, I don't see people talking about.
01:52And I'll give you two examples.
01:54The airline, no one would deny that the airline industry, right, just changed the world, right?
01:59You can go anywhere, perfect safety, and eat peanuts on the way.
02:02Just remarkable.
02:04More of a pretzels gal myself, but I'm going with you.
02:06Continue.
02:06But, you know, Warren Buffett famously said that the airlines were such a machine for incinerating capital that if, you
02:13know, if an enterprising capitalist had been at Kitty Hawk when Orville took his flight,
02:17he should have shot the thing down, right?
02:19Like, no, you can change the world and not make money.
02:23In biotech, you saw sort of, you know, the biotech revolution starting in 1975.
02:28So from 1975 to 2004, Gary Pisano at Harvard Business School showed that the biotech industry might maybe have made
02:36money in one year, right?
02:39And for that entire time, it was essentially just a furnace for money.
02:42You poured money in and, you know, drugs came out, lives came saved, but people didn't make money, right?
02:47And you've got huge capital investments, lots of uncertainty, long delays before payoff, enormous costs.
02:55Oh, and the problem that, you know, that AI is dealing with that biotech didn't is the open-weights labs
03:02out of China in the last benchmarks are four to eight months behind.
03:07Wow.
03:09That's not much of a moat.
03:10That's right.
03:11I also want to ask you, you talk a lot about the physical end of this and the scarcity of
03:16compute and these data centers that have to go in.
03:19And you make a comparison to Edison when they were starting.
03:22Please explain that.
03:23Right.
03:23So when Edison built the light bulb, you know, this is so iconic, then we have a bright idea.
03:28A cartoonist sketches a light bulb over our head.
03:29David gets more of those than I do.
03:31It's fine.
03:31So fantastic, you know, my wife says that mine are very dim, but that's fine, like, yeah.
03:36You know what, it's there, and that's what matters.
03:38That's the key.
03:39But he did, when the first Edison generator plant rolled out in New York City, they were serving something like
03:46400 light bulbs.
03:47Wow.
03:48Right?
03:48So you can have an astonishing technology, but you need a network, an ecosystem of other enabling technologies that have
03:54to build up to do that.
03:56In his case, it was generators and wiring and, you know, everything involved in a power transmission system.
04:01For AI, so we are used to, and I think a lot of the sort of, especially the VCs who
04:06went into AI is so heavy, they're kind of used to investing in SaaS companies, right, where scaling costs zero.
04:12And the marginal cost of production is zero.
04:14But AI is profoundly quite different, right, because you don't see, as users go up, the marginal cost going down
04:21that much.
04:22It stays the same.
04:22I don't want to say it stays the same, but it doesn't drop.
04:24It's pretty flat.
04:25It doesn't drop to zero, right?
04:26Like, just these things are really expensive to run.
04:28And so the scaling of this involves capital expenditures and physical construction and, you know, simple stuff.
04:35Pouring concrete, wiring buildings, those things are hard, right?
04:40They're not like dumping code onto the Internet.
04:43Yeah.
04:43And you need plumbers and electricians.
04:45And, you know, my joke about it was it's the return of the jocks, right?
04:47It's the guys who actually, and the women, of course, who do physical labor are becoming a constraint here.
04:54That is unlike most things that the software, you know, that this world has seen before.
05:00And a lot of those trades are understaffed as well at the moment.
05:02Yeah.
05:02Let me look at the other side of that coin, though.
05:04And I think you're talking about how this might not make money, might not lead to the kind of changes
05:07a lot of proponents are talking about here.
05:09It does strike me the fact that this does have this whole complementary infrastructure component to it may make us
05:15less willing to see the fact that it might not make money,
05:18that this is seen as not just something that's going to improve the economy of information technology, financial services, but
05:24could have this wider spread effect on the U.S. economy.
05:26Do you think that's true?
05:27Is it clouding our sense?
05:28Are we willing to pony up and pay for all of this fiscal construction because it seems like it's providing
05:33these wider benefits to the country?
05:35So I think that cuts both ways, right?
05:37So it is, you know, if you are an electrician right now, this is awesome.
05:40Like, this is fantastic.
05:42This is the best thing that's happened to you in a long time.
05:44And since the United States has sort of apparently decided as a matter of national policy we're not going to
05:48build housing, at least we're building something.
05:50But we'll build data centers.
05:51At least we'll build something.
05:52That's great.
05:53But, of course, the flip side of that is data centers are incredibly unpopular.
05:57They cause political pushback everywhere they go.
06:01Large portions of Silicon Valley appear to have adopted the PR strategy of James Bond villains.
06:07And there is this kind of – I think, you know, in some cases I'm not sure this is about
06:12the data center as much as just an incohate rage of pushback.
06:15This is a thing we can go after, right?
06:17And so when you're building this kind of infrastructure, you need the people there to say yes.
06:24And right now, America – despite all these benefits, I mean, Loudoun County, Virginia –
06:28Oh, my gosh.
06:28Just said that they're –
06:30You can't throw a stone without hitting a data center.
06:32I grew up in Rockville, Maryland.
06:33So this is like home for me, right?
06:35Sixty percent of their tax revenues are going to be coming from data centers.
06:38So you would think this would be great.
06:40And yet instead you're seeing this huge pushback.
06:42I think this is another case where the industry's norms that have evolved over the last generation kind of need
06:48to yield to the fact that saying – having a public relations campaign aware, we're going to take away your
06:56job and possibly end humanity.
06:59Not that last part.
07:01Not a great joke, Adam.
07:01Yeah.
07:02But there's another PR element that you highlight in your latest column, and I wasn't even that aware of this,
07:09and that's – you talk about where they're choosing to put some of this infrastructure, and it's in places that
07:15feels a bit exploitive.
07:16Like especially there's this Colossus One is going in largely without air permits in a largely black neighborhood of Memphis.
07:23You talk about – talk us through why those places and what that's doing.
07:27I mean those places have less political power to push back when the pollution gets so bad, right?
07:32And so if you are under enormous pressure to scale up your compute and you decide that the way to
07:37do that is to cut corners, and given all the obstacles that we've just talked about, you can see why
07:43the way you might decide to do that is to cut corners.
07:45It's easiest to do it in places where people have the least power to complain.
07:49If you do that in Malibu, it's not going to go down.
07:54A vociferous NIMBY crew in Malibu, yes.
07:57We've all spent a lot of time in D.C.
07:58Even in Loudoun County, if Loudoun County didn't want those data centers, you can bet they wouldn't be going in
08:02there.
08:02Yeah, that's right.
08:03I want to ask how you're thinking about the effect this is having on the labor market.
08:07So you mentioned that pitch, this is going to take away all the jobs.
08:08Opened up my computer on Friday, Torsten Slock of Apollo had this note.
08:12There is no negative effect on the labor market because of AI.
08:14We're not seeing jobs lost as a result of it yet.
08:17How are you thinking about that, putting aside, again, all of the kind of ideological perspectives on where this might
08:22lead to eventually?
08:24Do you think it is having an effect?
08:25Is it changing how companies, the CEOs of whom you talk regularly, the way that they hire, the way that
08:31they fire, the way that they staff their companies yet?
08:34Yeah, but what I'd say is what it's showing up in is two ways, right?
08:39It's less firing, although we're seeing some of that.
08:41And a lot of the firing that is happening is less, I think, about AI enabling productivity and more people
08:45kind of scrambling for cash to sort of try and chip away at these gigantic capital expenditures.
08:51But it's making it less – I think people are becoming more reluctant to hire just because of the uncertainty.
08:56But I think the deeper and more profound effect is about fear, right?
09:01You saw, especially during COVID, you saw the balance of power between companies and developers and people who had these
09:07skill sets swung really hard in favor of labor.
09:12What this does is give – is it creates fear.
09:16Even if I'm not being replaced, I might be replaced, which makes me a lot more reluctant to ask for
09:20a raise or say I should have a say in the way the companies are governed.
09:23And I think a lot of CEOs, especially founder CEOs for whom their company is their identity, are like, I'm
09:30not sure I want to give these people a say.
09:33We've only got about a minute left, and I'm going to ask the question that I always ask.
09:38I'm a bit of a skeptic about all this.
09:39One of the reasons is you talk about all this building, all this infrastructure, and a lot of that is
09:44being funded by credit and a lot of credit and a lot of circular credit.
09:47And for millennials who've lived through multiple crashes, I just feel like this is deja vu all over again.
09:53And what is your take?
09:54Is this a bubble?
09:54Is this sustainable?
09:55Is it durable?
09:56Is it rippling, if not popping?
09:58Where are we?
09:59You can have a bubble and still have a real technology.
10:03And that, I think, is – that's sort of to go back to where we were.
10:06That's the point that people are missing, right?
10:08You can say these valuations are – you can look at – I mean, SpaceX, which is apparently now an
10:12AI company, and say,
10:14you're $1.75 trillion for these financials and say, if that's not a bubble, I don't know what is.
10:23I don't know what the GDP of Mars is going to be.
10:25Well, no, we've been talking about there's a meme going around that NVIDIA is worth more than all the arable
10:29land in Australia, right?
10:31I mean, at some point, that just seems ridiculous.
10:32Yeah, and so – but at the same time, you can say that 20, 30 years from now, this was
10:38a bubble, but something came out of it that was really powerful.
10:42That won't be a lot of consolation if you lose your shirt on – in the meantime.
Comments