The AI Genie Could Deliver on the Hype

You're listening to Asia Centric from Bloomberg Intelligence, the podcast that pulls back the curtain non global business so you can invest better across the Pacific realm. I'm Tom Corbett in Hong Kong.

And I'm John Lee. Artificial intelligence, its rapid emergence has shaken up the tech world with eye popping potential to change the way we work, live, and do business.

From chips to the cloud, and from data centers to chatbots. AI is already plowing ahead, opening up a world of possibilities, but also prompting some to waive the caution flag.

Some estimates value AI's business opportunity in the billions and trillions, But what are the risks and can the aigenie deliver on the.

Hype As we start moving more towards AI processing worst case scenario by twenty thirty, if things aren't done correctly and there aren't any energy efficiency bottles that are implemented, it could count ten percent of all of global energy production.

Let's bring in Woojin Ho, senior tech analyst with Bloomberg Intelligence.

Woojin Welcome, Hey Tom, thanks for having me on, Hi, John, how are you? Hi?

With Jin which in chat GPT sparked off huge excitement on artificial intelligence. Is this hype real and how big is this market opportunity?

I do think the hype is real. I think this is very much unlike the metaverse. This is something that can increase the productivity for people, and this is why we're starting to see a lot of real investment going in by the cloud providers like Microsoft, Google, Amazon and Meta on the US side.

And how big is a market opportunity do you believe?

It's still unclear right because we can look at it in multiple different facets. There's an infrastructure side of the ledger and there's also a services side the ledger. We do know that hundreds of millions of dollars have already been invested in it right now not seeing much revenue. But if they can start producing revenue or make it into a revenue generating opportunity from a service aside, we're talking about the scopes of hundreds of billions and possibly into trillions.

Wou jin Ho, you talked about how you believe the hype around AI is real. Can we dive down a little deeper into that. What are we overlooking or what are we underestimating in terms of the potential, the power, and the risk.

Sure, So think about how we're using AI today. It's more for efficiency gains, and I think we just started to tap into what we can do with AI. Think about what AI can do, what chat, GPT and other AI engines can do like it. It can tap into multiple different databases and come out with an output that would probably happen and split seconds that could have happened

Is it an overstatement to say that AI represents a quantum leap in the evolution of tech?

Still unclear if that's going to be the case, right, But it does seem as if it's a step function change in terms of tech, much like what we saw on the Internet back in the early nineties. And once people learn how to use AI and the AI tools effectively, you'll see those productivity games potentially in.

Which companies do you believe stand to benefit from this trend?

Well, we're already starting to see some of the early beneficiaries. It's right now, it's mainly hardware and video is the first company that comes to mind.

Clearly they manufacture or.

Produce the GPUs, but there's also the compute processing side, which Intel as well as a MD and then you have some of the networking players like Arrested Networks and networking chip guys like Broadcom that produce chips to help support the AI side.

And how much computer power does AI require? For example, if we ask a question to Google search a traditional search versus a say, an AI powered search like CHATGBT, can you give us some sense of how much data and searching is required.

So if we think about the compute power that's needed right now, you need the highest available processor and processor speed. Now that there are two levels of compute that we need to think about. First is actual on the compute processing side, and that's where Intel and AMD come in, and think about it in the scope of ten to one hundred times in compute power to process all of the information that's coming through those servers. Right now, we're

And woujin ho. What constraints do those power concerns impose upon AI and its potential.

Well, that's a challenge here right when we're starting about high power compute, we're talking about a lot of energy that's being used an dissipated coming out of these servers and these data centers. Let me give you a stat all the data centers owned by the clouds they generate or use up roughly consumer about three percent of all

We're talking about the cloud guys.

Which think this sounds like a potential environmental nightmare.

Well, for some of the Bloomberg listeners who are very ESG conscious it is really a potential nightmare for the cloud guys.

Could you draw analogies to say cryptocurrency. I read this stat that bitcoin usage consumes more power than some mid sized countries in the world.

Yeah, if you think about the bitcoin ecosystem, it's still a fairly.

Concentrated ecosystem if you think about it.

There are some bitcoin farms that try to build out their own servers, but now we're talking about larger cloud guys that are trying to build AI farms at a larger and grander scale than than a bigcoin server farm. We're not talking about thousands of people that are trying to mine bitcoin. We're talking about millions and hundreds of millions of people to billions of people trying to access AI data coming out of the cloud, and the cloud

But that's going to come over time.

And wou Jinho a lot of those questions about AI and chat GPT and the power that they require, the US is not the only country that wants to pursue a piece of this technology. China is also very interested, and you have enough power constraints in China right now where that could be an issue for them if they go down this road.

And China has two challenges. One for us is the access to the latest generation of chips. And one of the advantages of getting the latest generation of chips is less power consumption because the smaller geometry chips doesn't consume.

As much power and it's much more efficient.

And two you need fewer of those boxes as the AI engine becomes more efficient. The issue is is that China will not have access to the latest generations of chips. There actually, and Video offers a China version of their GPUs, whether it's as an Apple eight hundred or the new hs In Henry eight hundred, which is the China version of those chips, They're just not going to be as

You can see a world of clashing imperatives emerging because upwards of two thirds of China's power is generated today by coal, which is of course a fossil fuel, so you have that factor coming in as well.

Right, And that's exactly the point, right, It's like you have to pick the better of two evils. Do you want to really start pushing the AI engine just to compete with the rest of the world or the US in this case, or do you want to manage and conserve energy to support factories on a day to day basis.

Would you just continue from what you were saying. US regulations are prohibiting certain chips to be exported to China, correct.

That is correct.

So there are some sanctions that the US imposed on US technologies that they cannot ship to China, and Nvidia is one of those companies and they actually line itemed a couple of Nvidia chips, Nvidia AI chips, the A one hundred and above. So what the US is essentially trying to do is slow down the progress of China's AI development.

So China can still use certain invidio chips. I think you mentioned A eight hundred. Can you compare the eight eight hundred to say the most advanced of the A one hundred or the H one hundred.

Sure, So some of the notes that I looked at in preparation for this, and it just seems as if a lot of it is the interconnect communications between the graphics cards, right, that's one. So speed and latency is very key for FURAI and also with the processing power of those chips as well, so it's not going to be as powerful, and that's one of the reasons why China may need to have more servers and more GPUs to compensate for the less processing power.

So how many years of generations behind would China be in terms of, you know, the usage of these Nvidia chips.

Well, if you want to add capacity, they'll try to keep up as close as possible my senses, and if we think about the power dynamic as well, the energy consumption dynamic as well, they could be anywhere between three to five years behind in terms of the chips itself.

You're listening to Asia Centric from Bloomberg Intelligence. By the way, if you like what you hear, and we hope you do, please rate us on Apple Podcasts, Spotify, Google Podcasts, or wherever you may be listening to us. Of course, more stars are better. Your feedback matters, and we love hearing from our listeners. Our guest is wu Jin Ho, senior tech analyst with Bloomberg Intelligence. Wujin The Financial Times ran an opinion piece in early May which rather boldly declared

That's a fantastic question, tom So, if we think about what's happening with Apple in particular, as well as some of the geopolitical pressure that the US has started to press US manufacturers, it seems as if Apple has.

Been moving along quite rapidly.

We actually did a lot of work on the Chinese supply chain prior to the ft, and initially we thought it was about over ninety nine percent of Apple smartphones with manufacturing in China. Now I'm going to give a nod to one of my colleagues, Stephen Saying, based out of Taiwan. We did another layer of analysis on the China supply chain and it seems as if the Chinese manufacturers and the assemblers have really started to move out of their operations to de risk their geographical exposure.

It's just as if it's more like close to thirty.

Five percent of the factories are in China now and they've started to move out to the Asian right, whether it's Vietnam and Malaysia and as well as a little bit more in India. Now, China is still going to be a mainstay. I still argue that there's still a lot of manufacturing expectise in China because of their multiple

So, currently ninety eight percent of iPhones are assembled in China. What's your expectations for say, like five years out.

Well, ninety eight percent of the smartphones have actually shifted more, has accelerated a lot faster, and it could be closer to forty percent, you know, in twenty thirty or even thirty percent in twenty thirty in China as these assembly

Is India You're ready for that kind of a challenge technologically?

Technologically there is engineering expertise. The one thing that India still needs to build up that that China has and India lacks is an ecosystem, right and that will take multiple years to build up. The other thing is the infrastructure. Where they lag in infrastructure is.

On the communication side.

If I look at the speed test, dot net or UKULA data, they rank eighty eighth in the world in terms terms of high speed fiber optic communications. But if we compare it to where China and the US, they're fourth and eighth in the world. And I want to go back in history where how did China get to

So the fiber optic network's being laid down, that's the tipping point for where China stands today. Would you say that's the case or is that an overstatement?

No?

I think that China made the playbook. I think Korea made the playboat as well. India started to recognize that and I would even say that US lags from a fiber rollout, and they're really started to lay out fiber as well over the next three to five years to really build up their digital infrastructure.

And I do view fiber.

I know that five G is touted as the communication but you cannot have wireless unless you have a wire to connect it back to a network. So fiber is the key tipping point for digital transformation success.

Well, jin can Apple effectively detangle from the China supply chain. Is the certain parts of the manufacturing supply chain that you just cannot get away from China.

In terms of manufacturing, well, there are several parts. Quite frankly, if you look at some of the components where China leads in, they lead in LCD panel, screen panel. We start to see that ship from Samsung and LG going into.

BOE and it's mainly manufactured in China.

If you see the display glass that's manufactured in China,

So the idea of Apple decoupling completely from China, you're saying, that's probably not going to happen.

I don't think that's not only an Apple statement. I think that's a global electronics statement. I think the aspirations are there, but I do think that China is going to be a last thing manufacturing partner over the long term.

Was you if I could circle back to artificial intelligence or AI, which Asian companies stand to benefit from this trend?

Well, let's start from the bottom up. It's going to be the server companies predominantly. They will be getting the large GPU line cards and chips from Nvidia and Intel an AMD. But it will be a companies someone like an Insper or a Lenovo or an H three C those other companies. But if we start moving up the chain on the services side, it's going to be someone like a buy Do or Tencent as they build out those AI services and try to offer it to the local economies and local regions.

And what about on the hardware side, Well, on.

The hardware side, it's going to be inspur Lenovo H three C, as much as I want to say someone like a Huawei.

On the networking side.

There's somewhat negatively impacted because of the US sanctions as well. They just don't have the access to the technologies on the networking side as they hope. So if anyone's going to benefit from the networking side, it's probably going to be H three C.

And what about the Asian cheap makers like Taiwan Semi, TSMC or Samson Electronics or esk Heinex.

So if we think about it from the chip maker front, AI needs a lot of advanced process notes and TSMC is going to be the leading beneficiary right now because they make the five nanometers seven animeter and in line for the threes and the two animeters, and for AI workloads, you actually ead a lot, a lot, a lot of memory. And if we think about ske Heinex and Samsung not only the DRAM side, but also on the nan side, they are going to be the biggest beneficiary among the Asian technology.

Providers, wujin Hole. Could it be that we don't know what we don't know about AI and chat GPT and its potential and how would you compare its emergence today with the way the Internet emerged in the early to mid nineteen nineties.

So there's a lot that we don't know, to your point, that we don't know going out twenty thirty years from now, sure in terms of what chet chipee can do, how it'll evolve. If we think about what the early Internet was in the nineteen nineties, it was basic email, right.

And then it evolved.

To very simple graphics to virtual.

Realities, a clunkier version then of what it is today.

Exactly much clunkiler version.

But if you think about what the Internet actually provided back in the early nineteen nineties, it created a whole vast of news services. We could imagine a service like an Amazon or an Uber, or food delivery services like delivery. Let's just say back.

Then, it's not just the technology, it's the way it's applied exactly.

I don't know how people are going to apply AI going forward. There are some scary aspects of it in terms of the productivity gains of AI will lead to potential job losses, but it might be productivity gains that we can really leverage and there might be more jobs created as a result of that.

So it's just I don't know.

What's going to happen in the next five, ten, twenty thirty years from now from the evolution of AI.

But those productivity gains could also bring with them investment opportunity exactly, economic growth, and all the things that go on with that exactly.

It can actually be seeds to new industry. It could be seeds to new technologies that AI was able to discover in the speed of years which would have been decades. So it depends on how companies implement AI and how they leverage the technology to really drive growth going forward.

They do with it.

Which there seems to be a lot of press on how AI can help with the search functions like Google Microsoft. Do you see a point where people actually pay to have subscription to these services.

Yes, but not for a search function.

I do think that again, going back to how AI evolves, I think there will be new services that may blossom out of AI that will.

Be paying for GO going forward.

Right, That's one of the I guess unknowns over the next three or five years on how to monetize AI because there's a lot of investment that's going in. We're talking about billions of dollars of investment, whether it's on the infrastructure side, of the software side, or development time. But if it's going to create the efficiencies for the end user, they want to really monetize that investment. And sooner or later it's going to come. I just don't know what shape or form it's going to come in.

It sounds like you need a lot of investment to build out AI, and a lot of these investments being made by cloud providers. Can corporates get into the action as well?

That's a fantastic question, Johnny.

The way AI is structured, you just need a bunch of servers, a lot of memory, some GPU, Right, that's a fairly and software that sits on top of it. What the cloud guys are investing in right now is a large scale AI engine that could support millions of people. I do think that the corporates can do this. I think there's a lot of interest in corporates building in house. Traditional hardware vendors will be beneficiaries of it, like a Dell or an HPE on the US side, someone like

Would it be cost effective?

It may not be right, but the risks could be far greater than the events smith itself, and then the reward could be far greater by building it inside as well.

Woo Jen All this raises questions about regulation. The AI genie is out of the bottle. How do we rain it in? How do we channel its power to constructive purposes? How do you see that unfolding?

I think it's already started to happen, especially on the US side. We're starting to see the regulatory environment try to work with the cloud providers in terms of trying to rein it in. First of all, it's going to take some time for the regulators to understand what it is and what it may mean to do, and the guidelines continue to change over time, primarily because the definition of what AI can do continuous change daily as well.

Woo Jen Ho. It's been remarkable having you here. We covered a lot of ground talking AI chips, China investment, and these are just the opening chapters in a story of exciting change and remarkable business opportunity. We look forward to staying in touch and hearing more of your insights.

Thanks for having me, guys, Thanks for.

Jin I'm Tom Corbett in Hong Kong.

And I'm John Lee. This podcast was edited by Clara Chen and you've been listening to the Age Eccentric podcast