FidelityConnects: Outlook – AI, U.S. tech and software

[00:04:13] Pamela Ritchie: Hello, and welcome to Fidelity Connects. I'm Pamela Ritchie. Our next guest reminds us that the version of AI we use today is the worst version of it. Rapid innovation in AI means the next iteration could be issued in just a few weeks' time and could be a vast improvement on today's capabilities. The regulatory frameworks for AI are still also to come. How do enterprises in the interim trust an AI provider with its valuable data? We're very lucky to have joining us today the person who can weave context around a number of these questions for investors. Please welcome here in studio Ben Holton. He's a Fidelity research analyst and portfolio manager. Ben, great to see you.

[00:04:53] Ben Holton: Great to see you. Thanks for having me.

[00:04:54] Pamela Ritchie: Delighted to have you here. We'll invite everyone to send their questions in over the next half hour or so. Let's just go back to that rapid innovation discussion. It has to do with the stocks that you follow. You're taking a look at the valuations. They can go from dogs to stars pretty quick.

[00:05:10] Ben Holton: Yeah, as fast as the technology is moving. We are seeing such rapid improvement. The line about the AI you use today is the worst version of it that you'll ever use, that's purely because tomorrow's is better and next week's is even better. We used to talk about what things look like in five years, that's now what things look like in 5 months.

[00:05:31] Pamela Ritchie: Okay, give us an example that we've watched. There are examples where they've been sort of, oh, well, that stock's not going to make it, they're behind.

[00:05:39] Ben Holton: Absolutely. The market has been divided into AI winners and AI losers but that debate is far from settled. You see things bounce back and forth from one bucket to the other. That's where the opportunity lies. Google is a perfect example of this. From February until April this year, Google was dead. No one was going to search Google anymore, the stock was in the dumpster.

[00:06:00] Pamela Ritchie: Search was over. Search as we knew it.

[00:06:02] Ben Holton: Search was over. ChatGPT or Perplexity or pick your favourite AI poison was going to win. Since April Google stock is up 70% and touched all-time highs last month. You can swing back and forth very quickly from winner to loser.

[00:06:17] Pamela Ritchie: Because Gemini is good and people like it.

[00:06:20] Ben Holton: Numbers aren't really changing. Google's earnings estimates did not change by 70% since April. This is all multiple and it's all sentiment driven. Of course, that's discounting the future outlook for Google.

[00:06:46] Ben Holton: I think with Google there was a perception they weren't moving fast enough. You have seen all of the companies move fast and that can change perception.

[00:06:56] Pamela Ritchie: Tell us a little bit about the other companies that we know as the Mag Seven, the household names that we've all been talking about, some people investing in and that are holding up the S&P 500 because there's so much of it. Take us into the story of sort of the froth, the discussion, the fear, the risk and so on. Do you see that? You look at these companies all day, you see what they're doing on an AI front, is it worth it, is it not, the CapEx story?

[00:07:21] Ben Holton: Yeah, a lot there. Of the Mag Seven, most of them, call it Mag Six, are not crazily valued relative to the market, relative to their history, despite being such a big piece of the market and performing quite well year-to-date for the most part. The froth is in the perception of what the future looks like and I think that's more down cap where you're seeing some crazy multiples, crazy meaning you're either priced as you're dead or priced as you've already won. Again, debate's not settled there.

[00:08:00] Pamela Ritchie: These are companies, for instance, that have had blast-off success because they're doing something in the realm. These are the disruptors, essentially. We've got the stalwarts to an extent, the incumbents, and then the disruptors. How vast is that field?

[00:08:16] Ben Holton: Most of the disruptors on the software layer are still in the private markets. Most of the incumbents have been put into the AI loser bucket. There's a debate on that whether it's right or wrong, that's the way the market is sitting today. Now, in the picks and shovels there you're seeing more public market success. The hardware providers, the cloud providers, that's where you're definitely seeing more AI winner status applied to public companies.

[00:08:49] Pamela Ritchie: When you think about ... we talk a lot about the data centres and what it takes to make them work properly. There's the building of them, the construction side of things but then there's the cloud, which we've been talking about for years. Everyone's going to need more capacity but this is a whole new realm of capacity that's going to be needed.

[00:09:07] Ben Holton: Microsoft's CTO was talking this week at a conference and he said a massive crunch of capacity is an understatement. What they are seeing is just insatiable demand. You're seeing this across every cloud player. This is why when we're talking about CapEx numbers we're having to move from billions to tens of billions to hundreds of billions and now some are talking trillions. Let's just be clear, those are insane numbers that you need to start having value out the other side for something not to go really, really wrong.

[00:09:45] Pamela Ritchie: This is being funded by the big companies to a large extent, the debt origination, concerns there? I mean, these are big companies that have killer cash flow so they're generally seen as fine. That said, there's a lot of debt being issued for this.

[00:09:59] Ben Holton: GPU financing has become a thing. That should scare some people. If we bring it back right to the start the pace of AI is so fast and what it can do today versus what it could do a month ago is changing perception on how valuable this could be. Right now it is full steam ahead.

[00:10:19] Pamela Ritchie: Okay, so there's sort of the retail story, what you and I and others will be playing around with, trying to figure out, get it acclimated into this world. Then there's the enterprise question which is always the question of when all the other companies using it within. This becomes accretive, this becomes something that you can say, yep, see that, we saved and it's helped with our costs and productivity and so on. How fast is that moving, the enterprise piece.

[00:10:43] Ben Holton: Slower than the consumer. Obviously, anyone can go out and download ChatGPT or Gemini or Grok. Enterprise, you have IT oversight, you have risk oversight, regulatory oversight, that has to happen slower.

[00:10:55] Pamela Ritchie: And they're worried about?

[00:10:57] Ben Holton: Security is with any software.

[00:10:59] Pamela Ritchie: The data.

[00:11:00] Ben Holton: With any software the number one concern is security. Does this do what it says it's going to do and is the data we're putting in there safe? We can't have something go wrong with our client data, that is paramount, so enterprises will always take a slower role to make sure things are moving in the right direction from the security perspective. Now, enterprises are not moving slow on this. We, as a collective like regulated industry and finance even are moving faster than we have in prior tech transitions because you're getting so much pull from the everyday employee. You see it in your personal life, you want it at your work life. We're seeing it.

[00:11:41] Pamela Ritchie: Your research, you're an analyst, you're using it how? Are you using it?

[00:11:45] Ben Holton: Yeah, we have multiple AI tools internally that we use. It's a great back and forth discussion on topics. Makes you smarter, makes you faster. What you're seeing, I think, in the enterprise is a lot of that. You're seeing behind the scenes usage. Number one example right now would be anyone who writes software in their company is probably using an AI assistant to help write that software. As a customer of that enterprise or a client of that enterprises you won't see that. You're seeing it maybe a little bit more visibly in search results. Pick your favourite e-commerce website. The search of the items it's showing you is now increasingly developed by AI, or determined by AI. It makes it more relevant. Behind the scenes but as an end user you don't see or understand that it's AI but it's helping you.

[00:12:39] Pamela Ritchie: Can it replace the data issues the U.S. government is having with everything to do with collection of jobs data to ... I mean, there's a whole raft of discussions of how we do everything, collect things for all kinds of purposes, including the government story, but it does kind of go to this disruption of how we collect things and we can do it faster.

[00:13:01] Ben Holton: I think AI right now is more on the analytics side. If you have a bunch of data what does that data mean, or what can you infer from that data rather than the collection of it. But that whole data estate is something you'll hear enterprises talk a lot about, getting your data estate in order. That means getting all of your data ready for AI. That's a big undertaking.

[00:13:23] Pamela Ritchie: What does that roughly involve? I mean, apart from maybe sectioning it off, putting it in proper folders or boxes or whatever it is online that you're doing but what else does that mean?

[00:13:31] Ben Holton: It means making sure it's clean so the AI can understand it. This could have been data that you had but not necessarily using. Now all of a sudden you're able to use it. Access and governance, who should be able to see that data. You can imagine enterprise has lots of sensitive data, not everyone in the enterprise should have access to every piece of data. You need to have that locked down, especially when AI makes it easier to search over that data. Maybe you had some errors in access before that weren't uncovered, AI makes it easier to, oops, uncover this.

[00:14:07] Pamela Ritchie: This seemed to go wrong three years ago, or something like that.

[00:14:09] Ben Holton: Yeah, and I shouldn't be able to see someone else's salary. Right now there's no way to search that if you have very powerful enterprise search that just searches over all of your company's data. It might surface that if the access controls aren't correct.

[00:14:25] Pamela Ritchie: So companies are getting ready, you won't go too far into the regulatory side because it's kind of up to government, but what do enterprises need, perhaps, from a regulatory framework side of things? Or actually is it much more internal, getting themselves ready?

[00:14:41] Ben Holton: I think from a government standpoint, stability, just knowing what rules we're playing against, that is always helpful.

[00:14:49] Pamela Ritchie: Is Canada going to work with the U.S.? The EU has done some work to get some regulatory frameworks up and running. We already all use ... you've mentioned this ... we all use Microsoft Word. We all use certain things so how do you put up digital borders if you need to?

[00:15:06] Ben Holton: Digital divides for an open economy are pretty tough. It's pretty tough, especially when most countries don't have their own industry to support it. The Microsoft Word example's great. We don't a UK word processor, we don't have a Canadian word processor. This is a global economy and right now most of the innovation's out of the U.S.

[00:15:30] Pamela Ritchie: Okay, and is shared around the world for a price.

[00:15:33] Ben Holton: Absolutely.

[00:15:33] Pamela Ritchie: And that's how that works. Take us inside the companies that are trying to transform. This goes a little bit to maybe the DeepSeek moment, which it'd be interesting to know what you think in hindsight, but it was a lot about sort of training getting up to speed more quickly. How would you talk about the training and companies getting up to speed? How quickly can that be done?

[00:15:55] Ben Holton: The AI training is interesting. It runs on a few different kind of axes. One is just dollars in the ground.

[00:16:03] Pamela Ritchie: Invest in it.

[00:16:03] Ben Holton: Well, it's how much equipment do you have? The other is how good is that equipment? Then the third is how efficient are you at using that equipment? The DeepSeek moment was basically a way to be more efficient using a given dollar or a given number of GPUs. What that did, and I remember talking with our clients at the time around that, of just it made things better and faster and cheaper. Ideally, that means we get more of it.

[00:16:34] Pamela Ritchie: That's really interesting when you take it kind of inside and how that works within the companies themselves, the fact is they're kind of making it work faster within. Do we see that first movers within this are going to show us how to do this? I'm sort of thinking big companies, the banks, for instance, would be making this work internally. Are they going to help everyone else get there? Or again, are these disruptors, smaller companies coming along to sort of say, look, we can move faster than a great big bank, for instance.

[00:17:05] Ben Holton: This will be topical this week. OpenAI showed a bunch of internal use cases of their software. This is things like a sales agent, a customer service agent, an inbound lead agent. They're showing companies, look what you can do with this software.

[00:17:26] Pamela Ritchie: What do they show us?

[00:17:27] Ben Holton: They showed us inbound sales, for example. They get millions of inbound leads that it's very hard to ignore and it's very hard to have a person to respond to them. What they found was a hyper-personalized with AI response that comes very quickly is better than either a generic response or a response that come weeks later from a person.

[00:17:52] Pamela Ritchie: So how is it different from generic, for instance?

[00:17:55] Ben Holton: It would know something about the questions I've asked rather than just saying, here is our price list.

[00:18:01] Pamela Ritchie: And we'll get back to you.

[00:18:03] Ben Holton: Or please visit this website. It will be a pitch to you around some of the context that you gave it in your inbound question.

[00:18:10] Pamela Ritchie: How is some of the competition setting up within the industry? You can take sort of the Mag Seven or the Mag Six but also OpenAI and other big companies, they're all competing, they're all moving towards and writing the course of history, to an extent, together, what kind of nuance can you offer to the relationships between some of companies?

[00:18:32] Ben Holton: You can look at the relationship even, we'll narrow it down, between Microsoft and OpenAI is very interdependent and depending which press you're looking at, potentially, pretty volatile. Now, I think the companies would both say that's the TMZification of the story but it's coopetition.

[00:18:59] Pamela Ritchie: There's some personalities in there.

[00:19:00] Ben Holton: There's personalities for sure. I think most are collaborative. There's a few that stick out as more combative, you've probably guessed which of the Mag Seven are more combative, but you see most software companies today, if they're building an AI functionality probably runs either on a Anthropic or OpenAI model.

[00:19:21] Pamela Ritchie: Neither of which are public.

[00:19:22] Ben Holton: Neither of which are public but both of which could go from model provider to application provider. Maybe they go from a supplier to a competitor to another software company.

[00:19:31] Pamela Ritchie: That's really interesting. Is OpenAI ever going to become public? It's getting valued at crazy amounts.

[00:19:37] Ben Holton: Their capital needs would suggest at some point in the future they probably do. Now, no need to go right now because they're still able to raise lots of capital in the private markets.

[00:19:49] Pamela Ritchie: And they have interesting sort of the way they're set up. Some companies have come up through the ranks with their innovations set to IPO and go straight into a company that would be public and some aren't. Maybe just discuss that a little bit. It's kind of, it's just an interesting piece of the story.

[00:20:06] Ben Holton: This is kind of a relic of history of how, say, OpenAI was founded as a research lab and as a nonprofit. They're very much in talks and in negotiations and in lawsuits right now to change that structure to allow...

[00:20:19] Pamela Ritchie: Like a nonprofit is sort of like a charity?

[00:20:21] Ben Holton: I think a bit different.

[00:20:22] Pamela Ritchie: A bit different.

[00:20:23] Ben Holton: But it's different fiduciary responsibilities, different responsibilities and, of course, different share structures.

[00:20:30] Pamela Ritchie: Right, and so to get out of that and just switch things around a bit is complicated.

[00:20:34] Ben Holton: Exactly.

[00:20:35] Pamela Ritchie: I mean, it's fascinating. Let's go to the energy side of things and kind of back to the shell of the data centres themselves, how they get plugged in. This is always the question. Nuclear does seem to be the future for a lot of these companies that are interested in it. It's not going to happen tomorrow. How do you take a look at that investment side of things? Maybe you're not directly taking a look at the companies but what are we going to be using for power?

[00:21:01] Ben Holton: Today the answer is anything and everything.

[00:21:04] Pamela Ritchie: It could be coal if needed.

[00:21:05] Ben Holton: Today you are taking power wherever it comes from. That includes natural gas turbines, there's a ton of them going in. There's some diesel engines and diesel powered data centres which doesn't sound great but...

[00:21:19] Pamela Ritchie: Makes you cough just thinking...

[00:21:20] Ben Holton: Well, the demand moves much faster than we can move the real world in supplying it with the energy that we would like to eventually have it supplied with. I think long term you are seeing a lot of interest in nuclear. Microsoft signed a deal to restart Three Mile Island.

[00:21:38] Pamela Ritchie: But that's a few years out.

[00:21:40] Ben Holton: That can't move fast.

[00:21:41] Pamela Ritchie: That can't move fast. What about going to the source of where energy comes from which sometimes is not inside a city. It's coming over power lines and so on from a different place. That said, is there going to be an argument at some point for putting data centres and even companies or whatever closer to the energy and where it's being pumped, created, whatever.

[00:22:05] Ben Holton:  I think there's a couple different axes where you need to consider. One is just how much land do you need. Some of these data centres are massive. You're not going to build one downtown Manhattan. How much power you need, and that speaks to where the power is today, where the power will be in the future, and then latency. For AI training...

[00:22:27] Pamela Ritchie: Latency is how good the connection is.

[00:22:29] Ben Holton: How long it takes for a round trip is the way to think about it. For AI training it is less important because everything kind of happens inside the data centre. You or me pinging an AI tool if that has to run to the other side of the world and back that introduces disruption and latency. Something for a ChatGPT, maybe that's okay for a Chat answer, as we move--

[00:22:51] Pamela Ritchie: 'Cause it's just text.

[00:22:51] Ben Holton: --as we move into the AI of the future, video, things like that, more real time applications, latency becomes a much bigger issue which means data centres have to start looking more where people are rather than where land and energy are.

[00:23:09] Pamela Ritchie: So how does that happen, or do people move closer to where the land and energy are?

[00:23:14] Ben Holton: I don't think it's the latter.

[00:23:15] Pamela Ritchie: You don't think it's the latter. They come to the cities.

[00:23:20] Ben Holton: Yeah.

[00:23:20] Pamela Ritchie: I think you had an interesting story about kind of piece of land in the middle of nowhere, actually, set to be set up as a new data centre. It's a good...

[00:23:30] Ben Holton: Part of the research is looking at what all the private companies are doing. OpenAI has become a bit of a kingmaker or destroyer of other stocks. Looking at what job postings open AI has, right now they've got a construction manager job at a very small town in the middle of Texas. The population for that town is sub 100 so you're building...

[00:23:53] Pamela Ritchie: Wow. They probably wouldn't even make a sign to say welcome to —where is it?

[00:23:56] Ben Holton: It's somewhere in the middle of Texas.

[00:23:58] Pamela Ritchie: Okay, somewhere in the middle ... welcome to, we only have less than 100 people.

[00:24:03] Ben Holton: It will be interesting because that will bring a lot of temporary jobs to that region.

[00:24:09] Pamela Ritchie: To build it.

[00:24:10] Ben Holton: One of the kind of downsides of data centres is once they're built you don't need a lot of people running them.

[00:24:15] Pamela Ritchie: That's interesting. So it's not a long term situation in terms of jobs.

[00:24:20] Ben Holton: I think it's a long term situation but not in one spot because the buildout of these data centres, once you build one you move on to the next. As a travelling electrician right now you can probably make a lot of money.

[00:24:33] Pamela Ritchie: Fantastic. That's so interesting. There's one of the questions coming in on data centres here. Are we building data centres here in Canada? Well, we are. I've driven past a bunch. Who is building those in sort of a large scale situation?

[00:24:45] Ben Holton: Anyone and everyone. I think this is where some of the geopolitics comes into it a little bit. For these massive data centres there is some sensitivity as to where your data is being stored. Data governance, data residency, especially in Europe is a big thing. Are you going to build a massive gigawatt scale data centre in Canada to only train data from Canada? Probably not.

[00:25:18] Pamela Ritchie: I see.

[00:25:18] Ben Holton: And how comfortable are you if you build that that you can train the world's data there. That becomes a bit of a debate given the geopolitical changes that are happening.

[00:25:30] Pamela Ritchie: Could, in theory, there be data centres built in North Dakota, I've got to get my geography right, to just sort of plug the line in from across the Canadian border, for instance, get it really close to Canada and then...

[00:25:43] Ben Holton: From a power perspective, maybe.

[00:25:45] Pamela Ritchie: Maybe, okay. Because that solves the residency problem there. That's really interesting. Do you see that being part of a trade discussion?

[00:25:56] Ben Holton: With the current administrations around the world it seems like anything and everything is up for debate.

[00:26:01] Pamela Ritchie: Everything is a trade discussion. No, I guess that's probably right. There's been a very interesting new rising company that has been enormous forever but has now kind of got a new role in the AI drive. This is Oracle. Mostly just curious about how you've watched on the playing field of how this has come into the 21st century in a way that perhaps not everyone thought they would.

[00:26:24] Ben Holton: Well, you talk about companies that maybe at one time were viewed as legacy.

[00:26:29] Pamela Ritchie: Legacy tech, yeah.

[00:26:30] Ben Holton: Melting ice cube, melting iceberg to setting all-time highs. Moving into new ventures. What spiked Oracle the other day was signing a massive deal to power OpenAI. They signed up north of $300 billion--

[00:26:50] Pamela Ritchie: Sort of eye-watering.

[00:26:54] Ben Holton: --[crosstalk] contract.

[00:26:54] Pamela Ritchie: Is it structured in the same way that most deals would? There seemed to be some interesting ways that this came about, some creative finance, maybe.

[00:27:04] Ben Holton: The financing is still TBD. They will have to build this out. It doesn't start for a few years. They'll have to build all this out, they'll have to finance this buildout. The nuance here is because of a complex relationship between Microsoft and OpenAI Microsoft has first right of refusal on all OpenAI's cloud capacity. That means Microsoft would have seen this contract and said, no, we don't want it. That, of course, raises questions.

[00:27:31] Pamela Ritchie: Of how Oracle did it.

[00:27:34] Ben Holton: Well, of why did Oracle want it if Microsoft didn't? There can be risk-reward scenarios that each looks at a bit differently. They can be at different spots on that risk-reward curve. They can at different spots on their future vision. Where do they see AI in 5, 10 years?

[00:27:52] Pamela Ritchie: And what are each providing, ultimately.

[00:27:54] Ben Holton: Each providing a very similar thing. They're providing cloud capacity, in this case probably full of Nvidia GPUs and several gigawatts of size.

[00:28:03] Pamela Ritchie: That is absolutely fascinating. When you see the funding itself, let's talk a little bit about sort of the consultants, the funding that companies have forever when they're getting software used to or built for within an enterprise situation have come in. They're big, again, kind of household names that will come in and help your company get brought into the next decade and so on. This is going to be new with AI because who are you inviting in? Is it still the same sort of auditors/consultants that we know of, the big four, for a long, long time? How do their jobs change? Do their jobs remain?

[00:28:43] Ben Holton: We talk about AI winners and AI losers. IT service this year has firmly been placed in the AI loser bucket. Do we see their role change? Absolutely. I think the consultancy part will still be important. Layering a new software in without changing the workflows feels half-baked. You probably need to rethink your whole workflow. Consultants have generally been a partner in doing that.

[00:29:10] Pamela Ritchie: And they charge a fee to sort of go in, set up a desk and be there.

[00:29:15] Ben Holton: They charge for advice and then they charge for implementation. The implementation side of the business looks, potentially, more challenged in that if AI is making everything faster and cheaper to do, this is a competitive industry, you will have someone pass on those costs in the form of lower price. So you'll see potentially big price compression.

[00:29:46] Pamela Ritchie: The advice and how do we do this, still relevant but the actual implementation might be much faster and maybe cheaper.

[00:29:54] Ben Holton: The AI zealot would say why do you need a consultant to give you the advice? Can't just AI itself give you the advice? That's the AI maximalist. I think there will be a divide between the consultancy side and the implementation side.

[00:30:09] Pamela Ritchie: The discussion of software, which is what you've covered for forever, is software, software's not dead but from an investment perspective what is the view on software at this point, even if it changes in six months?

[00:30:21] Ben Holton: It's interesting. The software index, depending on which index you're looking at, has generally outperformed the market year-to-date. However, that's misleading. There is Palantir, Oracle and Microsoft, three big names, three big returns. Outside of them the picture looks much more dire versus the market. That is, back to this AI losers, AI winners, there's a narrative that software is dead. There's, I think, three challenges right now that come into the debate, whether that's vibe coding, meaning we're all going to have AI to write our own software, we won't need to buy software anymore. There's seat compression, meaning most software today is sold loosely on the number of people in an organization.

[00:31:05] Pamela Ritchie: So you get a contract for the software to work for 24 desks.

[00:31:10] Ben Holton: Correct. If you have 100 people using that software you get charged 100 seat licences. If AI makes it such that you don't need 100 people anymore does that start to deflate the software budget? Then there's the ever present risk of just new competition. Is there AI native competition to the incumbents? I think all of those are real debates with different levels of validity and nuances in each but again, that's where opportunity lies in the market. Understanding which ones are being priced as if they are dead, and they won't be, versus the ones that are priced as if they've already won, which maybe they have, but if you're priced like you already have won nothing can go wrong.

[00:31:55] Pamela Ritchie: That, of course, is never the case. Let's end out just kind of the way we began. If you were to take a bit of a picture, a snapshot of what AI is today knowing it's changing, where are we?

[00:32:08] Ben Holton: AI is a consumer app today that will move into the enterprise, it will change industries.

[00:32:19] Pamela Ritchie: Fantastic. Ben Holton, it's great to have a conversation with you. Thank you for joining us on Friday and explaining so much of what you've researched.

[00:32:26] Ben Holton: Thanks for having me.

[00:32:27] Pamela Ritchie: Coming up on the show, we go to next week because the weekend is in between, on Monday Investment Director live from London, Tom Stevenson, joins us for a discussion on global markets. He'll discuss the latest headlines moving markets in the UK, across Europe and he'll even dive quite significantly into opportunities in China which we have not discussed as often recently. You do want to tune in for that.

[00:32:48] On Tuesday Fidelity Director of Quantitative Market Strategy, Denise Chisholm. She joins us to discuss her latest sector thesis. So Monday and Tuesday both of those will feature live French audio interpretation so do, of course, join us in either official language.

[00:33:02] On Wednesday portfolio manager Jed Weiss, he'll give us an update on the international equities that he is watching, how big market movers like AI and the geopolitical story are affecting companies in Europe and Asia where he is investing. Great to have you join us here today. We wish you a good weekend. We'll see you next week. I'm Pamela Ritchie.