- 1 hour ago
Everyone talks about models. Far less attention is paid to where AI actually runs. And while the first generation of cloud computing centralized technology, AI may be pushing in the opposite direction. As governments, industries, and enterprises seek greater control and transparency over their data and digital infrastructure, new questions are emerging about where AI should be built, deployed, and governed. What will the geography of intelligence look like over the next decade? Who will build the systems it depends on? And how will companies ensure AI delivers value to businesses and governments ?
Category
🤖
TechTranscript
00:00So what does the map of AI actually look like in the next decade?
00:07Who builds the system it runs on and how do companies and governments make sure AI delivers real value in
00:16their terms?
00:16Here to unpack this, we have Cormac Waters, EVP and GM of Oracle EMEA, in conversation with Anna Stewart, CNN
00:26correspondent.
00:27So please give them a very warm welcome.
00:52I feel like we should have danced on stage to that.
00:55I know, I was thinking that, but my dancing is worse than my talking.
01:00Welcome, everyone, to this next session.
01:02I'm very excited that we are not talking about sexy models from OpenAI and Anthropic.
01:08We're going to get into the plumbing of artificial intelligence.
01:11We're talking about where it actually is in the cloud.
01:15And I'd like to start on the thorny issue of sovereignty.
01:19And I'll ask you the most uncomfortable question I can, which is what happens, Cormac, when you get a phone
01:25call in the middle of the night from Washington saying, you know what?
01:29Switch off Germany.
01:31Switch off Paris.
01:32We don't want you operating with them anymore, which is, of course, similar to what we're seeing in terms of
01:38the Trump administration with Anthropic.
01:40No, actually, it's a question that almost every one of our customers asks, to be honest, right?
01:46And the reality of it happening is highly unlikely, of course.
01:49But to imagine what would it be like and therefore what can I do as a business, as a customer
01:55of Oracle to protect myself from this?
01:58So we think sovereignty is probably the biggest topic in our world right now.
02:03So we address it in several ways.
02:05So we offer a cloud.
02:09We feel we're a hyperscaler.
02:11But we have our AI, our cloud, is subtly different.
02:14We have the infrastructure, we have the data, and we have the applications.
02:17So we have what we call a full stack.
02:19So it's quite a comprehensive entity.
02:22We can give that to the customer in pure public cloud.
02:26It's okay for some functions and so on, but more and more people want it to be more sovereign, more
02:31protected.
02:32A government wants it to be highly secure.
02:35So sovereignty to them means where my data is, is in my data center, who operates on the data, my
02:42citizens.
02:43So actually, UK citizens, French citizens, maybe even EU citizens, but more and more national.
02:51And what law is governing the operation of the data center?
02:55That's the most...
02:56This is the tricky bit, because I think a lot of the time we talk about sovereignty with data centers
03:00and things.
03:00And you think, well, as long as the data and the center is housed in my country, operated by, you
03:06know, my people, my nation, it's safe.
03:10But the trickier bit is this legal jurisdiction.
03:12You are an American company.
03:14Yeah.
03:15Yeah, we're actually a global company, but headquartered in America.
03:19But we have, our European data centers are operated to a Dutch legal entity, actually.
03:25But ultimately, yes, it goes to an American parent.
03:28So if that final piece is of the most concern to the customer, which it is in some cases, in
03:35many it's not.
03:36We have two flavors of sovereignty.
03:38We have an EU sovereign cloud, which is operated by European citizens.
03:42All companies in Europe can use it.
03:46It's the same as a public cloud.
03:48And it's controlled by Oracle Netherlands, as it happens.
03:52But there is that ultimate parent thing.
03:53If you want further sovereignty, what we do then is we take our cloud and a third party operates it,
04:01a European entity, like Telecom Italia, for example.
04:04So it's operated under their terms and conditions, under their laws, their European business, and European people.
04:11So there is no American oversight on this one.
04:14So the only thing that needs to be worked out then is how do you keep the infrastructure and the
04:19technology current?
04:20So what are the rules about bringing in patches and upgrades and so on?
04:25And there's a carefully worked out protocol to do that.
04:27So that is the way that's becoming more and more common, actually.
04:31It's incredible.
04:31I mean, we're living in unprecedented times, but you're having to think through all of these different scenarios, which may
04:38be very unlikely.
04:38But in this age, we just don't know what's going to happen next.
04:42Well, the uncertainty is real, right?
04:44I mean, whether it would ever come to fruition is a question.
04:47But the uncertainty is real.
04:48And so companies need to make sure they have the option of going fully, fully sovereign or European sovereign or
04:56just make it private.
04:58And there are different flavors.
04:59The trick for us or the opportunity for us is that it's exactly the same cloud technology in all of
05:05those flavors.
05:06So there's no difference.
05:07It's not a scaled-down version.
05:09You don't get reduced functionality.
05:10You get exactly the same cloud, all the 200 cloud services and the same pricing models.
05:15But it's where the data resides, who operates on it, and what legal jurisdiction.
05:21Backstage, you mentioned multinationals because that's quite an interesting area.
05:25You're a massive multinational.
05:27You want to use Oracle, say, for your cloud services.
05:30But what jurisdiction are you looking for here?
05:32Yeah.
05:33So they tend to go where they're headquartered as well.
05:36So they will decide whether they want their data to be resided in Europe in this case, which is fine.
05:43And then most of the time, for business reasons, they prefer to have a single instance.
05:47So all of their subsidiary data in one place so they can get better analytics, better use of actual use
05:54of AI.
05:55So the plumbing of AI is one thing.
05:57But to actually use it, you need data.
05:59So if your data is fragmented and all over the place, it's much more complex to actually use it.
06:05So having it all in one place is a more sensible approach.
06:08What is the biggest concern when you speak to your existing clients when it comes to the geopolitics around sovereignty?
06:15The uncertainty.
06:16Just the uncertainty.
06:18And I think it's, I mean, I'm a very proudly European person, as you could hear from my accent.
06:23And who would have thought we'd be having these conversations, right?
06:27But it's true.
06:28And it's here.
06:29I think it's good for Europe because I think it's forcing us to actually look at our regulation and see
06:35we need regulations.
06:38Fundamentally, we need regulations.
06:39We need to be a bit faster with regulations.
06:41I was going to say, I don't think Europe is short on regulations.
06:43Yeah, but we're not exactly the world's fastest in getting them.
06:46We're not quick.
06:46Right.
06:47So we do need them.
06:48And we need to get them more, in a more expedient way.
06:53I think we need the standards to be global and open.
06:56So the interoperability between clouds needs to be more transparent.
07:00And I think we'll get the best if we have an open public-private partnership.
07:06I was really interested to see that Brussels has proposed new rules, you know, to speak about regulation around sovereignty,
07:15the tech sovereignty package it's called.
07:16And that could shut out providers who are under American jurisdiction, I imagine, like Oracle, from operating in Europe.
07:25How concerning is that?
07:27And do we think this is just a reaction to President Trump's administration?
07:30Or do we think this is, you know, genuinely something they're going to use at some stage?
07:34So, again, to be a little bit soft about it, I'm Irish, right?
07:39So my mother always told me, never talk about politics or religion, right?
07:43So I can't get specific on that.
07:46I promise not to ask about religion.
07:47Okay, but I think the reality is a lot of amazing technology exists all over the world.
07:57A lot of it comes from the U.S.
08:00If we limit ourselves to just homegrown stuff, we will go backwards rather than forwards.
08:05So we have to find a way to govern the use of global software.
08:12I think most companies, if you're just a, governments are national by definition.
08:19Companies tend to be multinational or want to aspire to be.
08:23So we need to keep ourselves open to the best technologies, but that are governed in a sensible way to
08:30protect us.
08:31And that they can't be just flicked off with a switch.
08:35So if you're running, so our AI to us, it's the infrastructure for sure.
08:39Then it's the data, and then it's the applications.
08:42So we think the best way to surface AI, to actually use it, is through the business application.
08:48So if you're doing your, you can do some sexy stuff, of course, you know, with all sorts of cancer
08:52research.
08:53That's fantastic.
08:54But normal day-to-day business, it's in how you match purchase orders, how you process payments, how you look
09:00at your supply chain.
09:01The software that does that, there are a handful of awesome vendors in the world.
09:07But very few of them are located in just one country.
09:12Most of them are multinational and have businesses.
09:15Oracle's business, 50% of our business is outside of America.
09:19Our competitors are all the same.
09:21We don't want to switch off and lose those, that technology, that innovation, by being too protected.
09:28But we do need protection.
09:30There's clearly a big talking point at the moment, and it's a bit of a trend, this issue of sovereignty.
09:34And do you worry that you'll see some companies in Europe wanting to find a European provider just to completely
09:40limit the risks that could be there?
09:43100%, they will.
09:43Well, so actually, again, not to be a party political broadcast, but what we're trying to do as Oracle to
09:50serve that market is that take our technology, our cloud, and actually quite label it as it were.
09:56So we give it to a third-party operator, and they fully run it.
10:01And they run it independently of Oracle.
10:03It's not connected to the Oracle network.
10:04It's just European or Italian or Norwegian or English or Irish.
10:11We've been doing it in the Middle East for about four or five years.
10:14They've been ahead of us on the sovereignty stuff, actually, which is interesting.
10:18So we have sovereign clouds in UAE, in Qatar, in Saudi Arabia, all operated by entities legally owned by those
10:28countries, but using our technology.
10:31And they found a way to make that work.
10:33I think we can do the same thing here.
10:35Wow.
10:36Let's move the conversation on to data centers.
10:40Oracle is spending billions of dollars on data centers.
10:44It's hugely ambitious.
10:46Some would say maybe quite an aggressive strategy.
10:48I believe there have been 30,000 layoffs.
10:51Will there be enough data centers to meet the demand of the future?
10:56Yeah.
10:57So we're spending a lot of money on data centers, like lots of people, right?
11:00So that's for sure.
11:01I think there's, yes, there's enough demands to satisfy the capital expenditure that's going on.
11:08We have $600 billion of orders taken that we haven't fulfilled yet.
11:13It's massive, right?
11:14So that's there.
11:15I think one of the things that will happen is that in order for the gigafactories or the massive centers,
11:20they're needed to train the large language models.
11:24And when you get to inferencing or reading from them, let's call it, when you get to the inferencing thing,
11:28maybe we see a trend towards having that inferencing engine closer to the customer's data.
11:35So customers will want to use large language models to take advantage of AI.
11:39It makes perfect sense.
11:41It's even better if it's flavored with their own private data.
11:45So they get a private version of it.
11:47But you don't want to move their private data to the public cloud or the public model.
11:50So therefore, the data center needs to move closer to their data.
11:53So we're seeing large data centers for the training and smaller setups for the inferencing, multiple smaller inferencing.
12:03So I think there'll be a lot of really big ones, and there'll be even more very small ones.
12:08Okay, this is a question to all of you.
12:10Hands up if you would like a data center near your home.
12:14Oh, we've got a handful.
12:16Okay, well, I'm kind of with you.
12:18They're noisy.
12:20They often use a huge amount of water.
12:22They can use as much as the town's worth, I think.
12:26And they use a lot of electricity.
12:28So how are you going to fix that so that we can all have the data centers closer to home?
12:32They're useful, but they're not these energy-guzzling, misery facilities that ruin our communities.
12:40Yeah.
12:40So actually, the two big topics for us right now, sovereignty and sustainability, they're the two topics.
12:49On the sustainability of data centers, it's a huge...
12:51I mean, I live in Ireland.
12:52As I say, we have a lot of data centers in Ireland, and it is a big topic.
12:56So we have to get better with the technology.
12:59So the latest ones we've built have a very unique cooling system.
13:04So the cooling is happening in the closed system closer to the chip.
13:08This is a wee bit techie.
13:11So it's like, get the cooling happen closer to where the heat is generated, but have it in a sealed
13:15system so the water is not...
13:16So it's a closed loop.
13:17The water has not been...
13:19Or the cooling liquid has not been lost.
13:20The heat will evaporate.
13:22So you lose less water.
13:23That's for sure.
13:24The second part, then, is the powering.
13:26The power of it.
13:27So actually, we have about 25 data centers across Europe.
13:30We've been fully using renewable energy since 2019 already.
13:37We'll be carbon zero net neutral as a total value chain by 2050.
13:43But we're actually using fully renewable energy since 2019.
13:47So what we're looking at now is how do we...
13:50We did an interesting partnership in America, actually, with a company called Bloom Energy, where they're putting a self-contained
13:56energy plant right beside the data center, which doesn't draw on the public grid.
14:02Okay?
14:02So it's kind of...
14:03It's avoiding that issue.
14:05And the power plant itself is actually more eco-friendly.
14:10It's not eco-friendly.
14:11It's more eco-friendly than would be generally used for power generation.
14:16The other piece that's interesting is it would appear across Europe there are pockets of power excess or energy excess
14:26that's not being utilized.
14:27So we could find a way to see where they are and then maybe plan the data center deployments closer
14:35to where the excess power is.
14:37Do you plan to do anything with the excess energy that results from your...
14:42Yeah, yeah.
14:42We have...
14:43In several jurisdictions, we've fed it back into the grid, right?
14:47And then it can heat homes or...
14:49Then it can heat homes and all of that.
14:50When I say things like that, people don't think, oh, that's not real.
14:53It is real.
14:54But it doesn't...
14:55It's not the primary...
14:56We build a data center to do data processing.
14:58We have excess power.
14:59We put it back in.
15:02We're primarily a drawer of power initially, but we do have excess.
15:05We don't waste the excess.
15:06And will there be enough data centers in the world to meet AI demand in five years' time?
15:11I think that's probably the biggest question, actually.
15:14So we can build a full cloud data center with all the same 200 cloud services that we have on
15:21our biggest public cloud data centers on three racks.
15:24So it'll fit in the closet.
15:26Because we can put that cloud data center into a customer's premises or a customer's data center.
15:30So will there be enough?
15:33Yeah, I think there will be.
15:35But we do need to be more cognizant of the planning of where we put them and making sure that
15:42we have adequate power.
15:44And they don't always have to be in populated areas.
15:47What's government's role here in making sure that...
15:50And I'm not just talking about data centers, really, but making sure the infrastructure is there to build and shape
15:56an AI future?
15:59So if data centers are becoming a reality, and I think they already are...
16:03They're like a utility, really.
16:04They're like a utility.
16:05I think every government should be looking to see, do they want to participate in that wave, as it were?
16:11Most should say yes.
16:12What are you doing to plan your power grids to accommodate?
16:17And how can we do that?
16:19And I think it shouldn't be done in isolation.
16:22I think there's a public-private conversation that's happening in most jurisdictions, actually.
16:27So I think it's the new reality.
16:30Data centers, it's a utility.
16:31It's like cloud computing.
16:34You just plug it in there, and it works.
16:36So you'd like to see more investment from governments, I imagine?
16:39I'd like to see more conversation and more investment, yeah.
16:42I think we can explain what's happening, where we see the demands.
16:46There's a whole variety of companies like Oracle and other hyperscalers and software companies
16:52and AI companies and power companies.
16:54Why don't we go back to the old days and have a conversation?
16:58Right.
16:58Now, many in the audience will be business leaders, I'm sure.
17:02And I was reading a survey about how businesses are adopting AI and what's one of the biggest drawbacks.
17:08Nearly a third of businesses surveyed by McKinsey last year felt held back because adopting AI,
17:15because data had to stay in a certain country.
17:17Yep.
17:18What are you telling companies that are in that position?
17:21And what are the other issues from the clients you meet?
17:24So there's a lot of AI pilots, hundreds and thousands, and maybe millions of them.
17:29Very few continue on to production.
17:32So we took a different approach to it, very deliberately.
17:38So we feel you can develop your own AI, and that's great.
17:42We give you the tools to do that.
17:43We also have the business applications.
17:45So we surface AI through those.
17:47So we give you, at the moment, there's over 1,000 AI agents in our ERP system, for example.
17:53And that means that every customer gets to use AI in the system they use every day.
17:59They're not trying to do it over here and then figure out how to connect it all up.
18:03The real problem is actually the data.
18:07To have great AI, you need great data.
18:10If the data is dispersed, if it's in different systems, on different technologies, you do need some way to get
18:17a grip on your data.
18:18So what we've seen is the need for a data platform, if you want to call it.
18:23It's like a tool to manage multiple databases.
18:27It's a very simple way of putting it.
18:29But we give all our customers the data platform that lets them manage their Oracle data and their non-Oracle
18:35data,
18:35whether it's in the Oracle cloud or it's on-premise.
18:38But that's sort of problem number one, when you have a new customer coming to you wanting to implement AI.
18:43This is one of the biggest issues.
18:44The data is just all over the place.
18:46If we can get a data strategy first, then you will be able to deploy much better AI.
18:51No question about it.
18:52Otherwise, you're working in pockets.
18:54And it's like the olden days of we all had different systems and every department had a little back office
19:00IT system running.
19:01And then someone puts in global ERP and they go, that was a good idea.
19:05All right.
19:06So now the idea is, why don't we plan this properly and say, AI is moving so fast.
19:12Very few companies can keep pace.
19:14You need a partner that can give you innovation every 90 days.
19:18So you don't know what you're getting yet, but you're going to get more 90 days time.
19:21How do you adopt that?
19:22So how do you take that on board?
19:24So if you have a coherent data strategy with the applications to the surfaces and an innovation partner to help
19:32you adopt it.
19:33Actually, we have about 15,000, we call it Fusion ERP and 50,000 NetSuite ERP.
19:41So 65,000 ERP companies in the world.
19:4465% of them are now using AI on a daily basis.
19:4865%.
19:48It's huge.
19:5015 months ago, it was 10%.
19:52Wow, that's massive.
19:54So because they're getting to use it in the business application, they're not doing it on the site.
19:58They're doing it in what they do every day.
20:00Processing purchase orders, reading a P&L, looking at the supply chain.
20:05Practical use.
20:07You only have 40 seconds, but I'm going to ask it anyway.
20:10Of all the businesses you work with, where have you seen the best transformation due to AI?
20:17Ironically, the biggest uptake has been in the regulated industries of banking, ironically.
20:24I would have thought the opposite, right?
20:26I think it's because they're regulated.
20:29They're very structured.
20:30They're very controlled.
20:31They know exactly what.
20:32And they can use it for fraud detection, looking for data patterns.
20:38The tools are becoming so much better and faster now.
20:40They have the structures in place that they can readily adopt.
20:43So it's really funny that the least, frequently the industries of network, not often thought of as the fastest moving.
20:51They're adopting it the fastest.
20:53There you go.
20:54We are out of time.
20:56It was such a pleasure to speak to you.
20:57Thank you very much for making the plumbing of AI so interesting.
21:01And thank you to the audience.
Comments