2024 state of cloud - emerging into a brave new world

In this episode to kick off 2024, our CEO Vinay Joosery and SanjMo principal analyst Sanjeev Mojhan get together once again to look back at 2023 and ahead to 2024.

They discuss the events and trends that made the greatest impact to 2023, from inflation, rate increases, data sovereignty washing, security breaches and market exits to generative AI's buoying of an otherwise uncertain, lackluster economic year.

They then pivot to 2024, considering the hybrid operating model's second act through the lenses of a shift from cloud first to cloud smart thinking and how bringing AI to your data is superior to the alternative. They also discuss whether committed spend is really OpEx instead of the worst of CapEx, i.e. you're locked into spend but don't own an asset, as well as the need to return to an innovation mindset to gain an edge.

But these are just the highlights, 2024's inaugural episode promises to give you plenty of insights and thought provoking material to help you make the most of 2024!  

Find out more about Sovereign DBaaS at https://severalnines.com/sovereign-dbaas

Transcript

0:00

One lesson that we are learning very, very quickly is taking your data to AI is not a good idea. What is a better idea is to bring your AI to data. This is Sovereign DBaaS Decoded. A podcast for IT leaders and implementers looking to reliably scale database ops while maintaining control of their data stack. In each episode, we join industry experts to discuss the what and why of sovereignty and how you can implement the Sovereign DBaaS concept of your own using open-source databases, deployment models and tooling. Let's get started! Hello and welcome to 2024’s first episode of Sovereign DBaaS Decoded. I'm Vinay Joosery and this episode is brought to you by Severalnines. Our guest today is Sanjeev Mohan, Principal Analyst at SanjMo and former Gartner Research VP. Thanks for joining us today. So, Sanjeev, what's new? So Vinay, thank you so much. It seems like an annual pilgrimage for me to start my year by being on your podcast. For the second year in a row, I'm honored to be on it. To answer your question of what's new and what's going on, I have to say 2024 just the number seems a bit magical. It's a leap year, it feels like this is the year we will really leap forward. And although I have to say that I've entered the year with a lot of excitement and momentum, but I'm also noticing there's a lot of trepidation in the markets, if I may say so. What I mean by that is while most of the people I talk to are quite upbeat about 2024, I also feel there's a little bit of hesitation. There’s a lot going on in our world right now, which we will talk about. But there is a little bit of caution. So companies are saying that we are excited about 2024 but, and the feeling that I'm getting is that companies want to take it cautiously, Q1 and Q2. So the first half of 2024. And then they think that the rest of the year should be fine after that. But basically they're being very careful with how much money they spend and where they spend it, and the investments. They may hoping that, you know, if things don't improve, then they should be able to last the whole year with whatever their budget is. Yes. And in a way, it kind of sets the stage for, you know, the recap we will do for 2023. Right. I mean, you know, before we get there, actually, let me say this is this is our 10th episode of the Sovereign DBaaS Decoded podcast. Right. So I never thought we'd get past three or four. So, you know, we are happy that, people are listening. We're happy that we've had a bunch of guests. Right. And, in a way our hunch was correct when we launched the podcast. People want to hear about data sovereignty. And during that time, like last year, even the top two hyperscalers, AWS and Microsoft, they came out with their own sovereign cloud programs. So, that's interesting. I remember when we spoke, you were like, “Well, from the US point of view, sovereignty, you know, maybe it's more of a EU thing.” You know, we found it interesting that, all the hyperscalers actually they have these sovereign cloud programs. And probably, thinking like, you know, India has their own law now on data protection. A bunch of other countries as well. So that was a good move actually to start this podcast. And our episodes have been downloaded more than a thousand times. So now we are back with you where we started in a way, right. One year ago. And, so what can we expect for today? Well, we'll talk a bit about the state of cloud, right, Sanjeev; you gave us your your hunch about 2024. We'll actually first recap the cloud in the DB space in ‘23. Right. And we'll do it in three stages, influences, events and surprises. Right. So we'll see what's going on today after that and then we'll look forward to ‘24. So taking the influences, we saw more instability in the world, right? I mean, from the geopolitical scene, it's just getting mad right? And we, we also see a more restrictive business environment, which you are guessing that it will, it will actually, you know, continue and I mean, if you thought the end of ‘22 was bad, ‘23 got worse in terms of layoffs, inflation, central banks hiking interest rates. So, Sanjeev, how do you think these macro influences have affected the cloud market? You know, I'm reminded of something I I heard from some of my my peers, that the CEO of AWS was saying that he’s seen good times, he's seen bad times But these are neither, they’re uncertain times. So that uncertainty is coming this year because first of all, in the US where I'm based, we have a major election coming up every four years, but we're not the only ones. There are over 50 countries. More than half of the population is is having elections this year. So that… although, you know, I don't let these things come in my way, you know, I think a lot of these things are noise. But still, you know, people are affected. Elections. We have a couple of wars going on in the world which every day you read, they just seem to be…there’s a flashpoint and they're expanding. So these things caused companies to reconsider and think,“You know, what should we be doing this year?” I know we're talking about politics and the… some of these are non-technical topics. But then there's this whole uncertainty, if you may, which has been introduced by AI. And, you know, it's like how transformative is AI? I see articles every day saying that AI will add every year,, McKinsey says that 4 trillion dollars a year to the global GDP which will increase by 7% to 20% on an annual basis. Which is a lot of potential. But then you also see very next to it is an article that says“Is the generative AI hype over?” So what is it? Is it overhyped, is it underhyped or is it normal? So these are both issues that every company has to grapple with before they they know where to make the right bets. So that's something which I feel is is adding to a little bit of uncertainty around us. Yeah. And actually, in terms of events, we can certainly say that, I mean 2023 was probably the year of generative AI, right? And it, helped quite a few companies offset losses in value. Right. If you think about, you know, some companies like you know Adobe, NVIDIA, you know, it kind of pushed up all the valuations. So from that point of view that has driven a lot. Another event which is actually, which in the EU, it had quite a big impact you would say. There was some updates on the Schrems II. So last year there was an adequacy decision, right. For EU to US data flows. And basically it means that data transfers from the EU to the U.S.A. they are not objectionable from a data protection perspective and that’s based on an Executive Order of President Biden in 2022, which is called Enhancing Safeguards for the United States Signals Intelligence Activities. So what it means is, you know, we've done it two times now, right? We've had some kind of framework to transfer data. Then there was Schrems I. It was canceled, then there was a new framework. Schrems II happened. It was canceled. Now we're kind of getting into this,“Well, it's kind of okay, anyway.” And now the bets are on for Schrems III, right? It's kind of like, you know, can an enterprise really trust that they can follow this? Because it’s been overruled two times and the basic facts haven't changed. Right. In Europe, we have a very different way of looking at privacy compared to the US. There are surveillance laws in the US and they won't change those laws. And then there was one final thing that thought was was interesting. There was some high profile outages and security breaches, including MongoDB, which is quite a, you know, major DBaaS provider. And then obviously, VMware acquired by Broadcom, you know, we talk about private cloud. I mean, that's just huge. We saw that, you know, Broadcom ended the, the partner program for VMware. So a lot of partners actually are wondering“What's going to happen, you know, to us?” What's your take on these events Sanjeev? You already mentioned AI… I want to actually go back to AI, but you know, another very important from a political point of view or economic point of view is this whole inflation, interest rates. To answer all your questions about what's happening, on the business side, a lot of companies got heavily funded in 2021 when we had zero interest, almost zero interest rate going on. So now that there’s inflation and all these other things we talked about, they don't have a lot of runway. A lot of companies this year are going to have a rude awakening and find out that, you know, whatever valuation they had two years ago is no longer going to be the case. So they need to raise more money. That will be interesting. I don't think there's a shortage of money by the way, but I think IPs have a lot of money to spend, they just don’t know where to spend it. And they, like I said, they're being cautious where to spend to, it's not the shortage of money. It is that these companies have to swallow a bitter pill and, you know, go in with a down round. We might see some mergers and acquisitions going on, although some companies may be just a fire sale. So it may be an M&A on paper, but it's really, they just couldn't they just shut down. You know. So there's a lot of trepidation, but now to go back once again to AI and the regulations. So while I know Schrems I Schrems II and now getting into Schrems III, the EU AI Act at least the new version because there was already a version for many years, but the new one came into effect in 2023, literally a few weeks ago. My personal feeling is that Europeans over index on privacy and it's like you said, you know, we've tried a couple of different versions and we haven't really succeeded. Now we are trying again, but the problem is that AI is morphing and changing faster than we can pin it down. So I just feel that, you know, in a traditional database environment where things are very static or well-known or well-defined, like with a data warehouse, you could still have some rules and regulations in place. But with AI, with the speed at which it's moving, I don't know how the privacy acts can keep up with that. Even in the U.S. We already you saw just a few weeks ago New York Times sued OpenAI. OpenAI is like “No, I'm sorry, this is derivative work. This is not your original work.” And all kinds of arguments that have been made on both sides of the pond as to what we should be doing. So I just feel that these acts are going to be in the process of keeping, trying to keep up with the advancements in AI. So I don't know if that answers your question, but I just feel that because the generative AI piece, not AI itself, but the generative AI piece is relatively new. Europe needs to decide what side of the battle we need to be on. And I'm being very honest here, a lot of your followers are in Europe. They may not want to hear me say these things, but I feel Europe lost badly in the cloud business. Like literally, they don't even play in that space. It's all American companies except in China. They have a thriving cloud service providers. But Europe, with all this emphasis on privacy, never really managed to create an alternative to AWS, Google, Microsoft Azure of the world. And they should have. Because privacy is so important. So let's not make the same mistake for AI is my point. Yeah, and I will totally agree with you. I think, you know, looking at market shares and everything. Hyperscalers are, you know, number one, number two, number three. You know, they have like, most of the market. And I don't know, I mean, I don't know how we get there, how we got there. But but for sure, AI is another space. But again, the same the usual suspects are they are very well positioned to take that market. Right. I mean, looking at, in terms of surprises, you know, we looked at, like generative AI, how much it grew. It had a huge impact on, for example, people like, you know, like NVIDIA. I mean, they do the picks and shovels of AI, Their market value tripled right? Then the other thing is that you don't do AI without databases, right? So there are these vector databases which we saw a lot of. There are pure databases like Pinecone and Milvus Then you have existing databases that have their own vector search. For example, PostgreSQL with pgvector, right? Speaking of which, we added to the list of extensions that we support. Oh, very nice! So for me, that was a surprise. You know, for last year, all these vector databases getting funded because, you know, it seems that the database market keeps growing and growing every two years. There's something new that comes in and kind of expands. Right. Right! You know, the other thing was, sovereignty. Right? So we made a bet in 2022 that sovereignty would become a concern. Right. But we could not predict the spate of sovereignty initiatives from the hyperscalers. And AWS and Microsoft, they announced their sovereign cloud programs just last year, three or four months ago. And in a way it kind of brings confusion to the EU market, right? Because these products are only sovereign insofar as the agreements allow. Right. I mean, if there's a Schrems III and it invalidates the sort of, agreement that the EU made with Mr. Biden, then customers will be back to square one. And I think for an enterprise it's kind of strange to you know, to navigate. But again, there's not like as you mentioned, we don't have those superpower companies, tech companies in Europe where you can say, you know, they can get the technology from Europe itself. Finally there's one thing, you know, we've partnered with a partner for many years with a company, MariaDB. And that one, it’s sad to see. MariaDB Inc. is in big trouble. They were listed on the New York Stock Exchange in December ‘22. They were trading at $11.50. Right? The stock is now down to $0.18. Right? They've had layoffs, they've ditched Xpand, which used to be Clustrix, which was this kind of a clustered database running on multiple nodes. And DBaaS. They don’t do DBaaS anymore. Correct! They had their SkySQL DBaaS. So not sure what's next for these guys. But does any of these surprise you Sanjeev? Yeah actually I, I was quite a fan of MariaDB and the work they were doing. But as in Gartner, I'd been to their event. They used to do a very big event and I was upbeat on MariaDB because MariaDB’s deployment is actually pretty big. ServiceNow embeds MariaDB and there were over 100,000 deployments. So it came as a surprise to me that, just the thing that they went out so quickly and now they're a small shell and I still talk to them because some of my good friend are still there. But a lot of people I know have since been let go. I wonder if they made the mistake of going public. They went public through the SPAC SPAC yeah. And I think that was a big mistake. They should have stayed private because then they could have navigated this. Once you're a public company, you’re exposed. Everybody knows what's going on. So maybe that was not the smartest move. And I think going public using the SPAC vehicle to me is a bit of a red herring anyway. The thing is, when they went public with the SPAC, there was a lot of bad writing about SPAC in general, about that mechanism. It's sad to see the company almost imploding. Maybe it will be picked up by some acquirer. Yeah. And very interesting to see where MySQL is going. Oracle for many many years trade was, again I'm saying these things, I hope people don’t take it in the wrong way. But I felt Oracle was in denial with their MySQL existence. Because Oracle has such a flourishing DBMS market. And by the way, I started my career at Oracle, I think very highly of Oracle, but MySQL now has taken on a new form with HeatWave, and it's now not just a transactional database, but it's got transactional and analytical capabilities. And one more point I want to make and then I want to hand it over to you, is you mentioned vector databases. What started as a very exciting new use case of a database, which is to store the vector embeddings so you can then do a vector search has now become table stakes. So having vector search in a database is not all a difference, a differentiation. The only differentiation that remains is how can you take the vector search capability and have that workload coexist with your traditional workloads? And this is why so many different databases now have vector search capability because they're trying to say,“You know what you could already do an OLTP, you could do OLAP. Then we introduce full text search using Lucene, and now you've got vector search.” So it's very exciting in my opinion. And in a way, I think it it follows a little bit the trends that we've seen, right? I mean, we talked about polyglot persistence. Right? Like we've talked about that for and since ten years ago. Right? And then what we saw was that suddenly, you know, people started adding maybe a key value interface. Maybe they had graph capabilities. Right. And then the question is,“What does the enterprise do?” Do they continue maybe selecting specialized databases? Because then they're going to end up with a lot of databases. And that maybe doesn't work too well. Right. Or do they go multimodal? Because like PostgreSQL that they're using or their Oracle, already has all these capabilities. I think that's something that even we've seen with the graphs and other use cases. I will add two points. So first of all, I completely agree with you. I have a huge insight, because if you read between the lines of what I just said, multimodal databases, in my opinion, are the way to go. But the second point I also want to make is that, the same is true for graph database, but there are a couple of people like Neo4j for example have done really well in graph. So maybe there's an opportunity for just a handful of vector database standalone products to do really well, because they do things so well. For example, Pinecone is a leader in vector databases in my mind, from the commercial point of view. There are many others, like open-source, you know Milvus, Qdrant… There are a lot of really good options. Some are even embedded. But let's talk about Pinecone. I saw their announcement from literally a day ago where they talked about some new capabilities that could reduce hallucinations down to zero when you build your RAG (Retrieval-Augmented Generation) pipelines. So if they do that, then that could be a game changer. So having a standalone database actually may not be such a bad idea if you get to use capabilities that are far beyond what you would find in a highly integrated database. Correct. Another point I want to make, by the way, talking about Pinecone, is last year Snowflake introduced this whole new engine called Snowpark Container Services. So on the same data, you can now, you don't have to use Snowflake's native engine. You can use a containerized engine where you can even pick and say,“I don't want to run it on CPUs, I want to run it on GPUs.” In that container you can run Pinecone. So interestingly, what happens is that now I can take my data in Snowflake, turn that data into vector embedding, and store it in a containerized Pinecone all within my VPC. So that's another interesting point I wanted to raise. Yeah. No, for sure. I totally agree with you there. There are capabilities that obviously if you are specialized on doing vector search, then you'll have more time to, to add more things. Right? Whereas the big Oracles or even Postgres, you know, they have so much to work on. So that's tough. And looking a little bit at these, you know, influences and events that we talked about. Right? For last year. I mean, the question is, you mentioned Q1, Q2, maybe companies will be in a little bit of a kind of waiting mode because, you know, there's so much insecurity. Now, my question to you is, in terms of cloud strategy, I mean, are companies transitioning from cloud first to cloud smart? And will hybrid become more of a viable option? Because we know that, public cloud is very expensive. And as you mentioned in the previous podcast, it's not like, you have very, very variable kind of workloads. Right? And a lot of workloads are just, you know, kind of normal. I mean what are your thoughts there in terms of this hybrid or..? Yeah, I think hybrid is coming back to some extent. I will be careful by saying that does not mean things are moving to on-premises. We will still see cloud migrations, but it's not going to be predicated on everything must move to the cloud. On-premises deployments are actually starting to look pretty attractive because some of these things that I talked about like containerized services, some of the things that we we put a lot of cloud native are now available on-premises. So I now have an option where I can take the latest and greatest binary from the cloud, but run it on the data that sits within my safe and secure, or seemingly, safe and secure on-premises environment. So if I happen to be in an environment where I am very skeptical about moving to the cloud for various reasons, security being one of them, but it could be latency, it could be regulations. I can now take advantage of of the cloud operating model on-premises. Which, you even see that in announcements on Dell for instance. Dell with RedHat, OpenShift is now saying that,“You know, we give you the entire AI ecosystem, so you don't have to move your data to the cloud.” The second point I also want to make is that, one lesson that we are learning very, very quickly, is taking your data to AI is not a good idea. What is a better idea is bring your AI to data. If you are going to train a model, which we are still in the early stages of an end user training their own AI model, we are still doing RAG and some fine tuning. But if NVIDIA Grace Hopper chips become really cheap and Intel, AMD, Qualcomm, ARM-based, all these become commodity for training an inference, then a lot of these workloads will move to the edge or to on-premises because the cost has gone down. In such a case, we want to move AI to where the data already exists. You know to train an AI model we need a lot of data for it to, to be of value. Where is that data? If it's in the cloud, ok fine, train it there. But if it never moved to the cloud for for whatever reasons, let's bring AI to where the data is, train it there; and once a model is ready, if we can have a smaller model, I can even put it on an edge device. Like Android’s, I saw another announcement just yesterday how Samsung and Google Cloud are now collaborating to run some of Google's models on these Android phones. Yeah. Now, let me get your spin on these. I mean, we talked about costs, right? Cost is a, is kind of the big driver in this, right? One way that those typical hyperscalers are kind of optimizing the cost is through these multi-year contracts. Right. So this means committed spend. Which, seems closer to the CapEx model than the OpEx one, right? But you don't own the asset, right? So if you have the committed spend, because that's kind of what the enterprise can say is like,“Okay, let's do a contract with hyperscaler A and then we’ll commit to using, X amount of million dollars or whatever, right, in the next three years.” And then you gotta pay it even if you don't use it. But then, it feels a little bit like it's a bit of a CapEx model. But then you don't own it. You don't own the asset. So because it's not an operational expense that you can actually just turn down. Right. If your company's trying to cut costs. I mean, I don't know if you have any experience in this, but, you know, what are the implications for companies who got themselves in these contracts? Are we saying like, “Okay, we chose cloud first as our direction and we'll just continue down that road?” Right. Understanding that it costs more; or, as you mentioned, since we can do the cloud model through our own environments, we have the cloud operating model, we have all the toolsets today. To do that. Are the enterprises going to look at that and say, “Okay, yes we signed that contract, but actually we got to start looking at, at alternatives?” Like, you mentioned edge right on-prem, co-location you know... managed service providers or any other alternative environments? What do you think? Reserved capacity, the thing is a conundrum. When you don't have too many choices, then okay fine, reserved capacity is how we saved cost. If you reserved that capacity for one year or three years, then we got a favorable pricing. But now I'm also starting to hear how difficult it is to estimate what that reserve capacity should be. For example, many of my clients will tell me that,“I had expected to spend X amount of money over three years. But with these new use cases, I've already in six months, I've already spent that money, so I'm running hot on my budget. I've exceeded it.” Some companies tell me that, “We had excess capacity available.” And if I'm not mistaken, I don't track this space very well. So Vinay remind me. I thought if you had excess capacity, you could make it available on Marketplace. But I think AWS has shut that down recently. Yeah, I'm not sure. What I know is, I just read in the news that some large company had bought a lot of capacity and then they didn't use it all. But then they actually managed to get somehow a discount into their next contract so that unused credit was kind of, you know, moved into the next one. What really amazes me about this reserved capacity, a very interesting case, slightly tangential to what we are discussing, but still very relevant. I was looking at some benchmark numbers where somebody had compared the cost of running a workload on-premises with the cloud. But then when I dug deeper, I was like the cloud cost that you are comparing it with is a reserved instance cost. So it's actually amortized over three years. So you cannot take that number because you're paying for three years. So you cannot take that number at a point in time and then compare it to on-prem, because that's already discounted in the cloud. And so people are playing all kinds of like interesting tricks with this concept. Yeah. So, in a way, if people are moving more towards, you know, diversified environments, right, that will take care of vendor lock-in. Because, I mean, most of the analysts are kind of, talking about the blast radius of potential incidents. Right? There's more vendor lock-in. You have potential large scale events, right. That can affect a lot of companies. And also to improve navigability of the regulatory landscape. Right. Because if you are in control of your environment, it's not just cloud first, but you can decide where to deploy things based on new regulation, then you have more control, right? Because I mean, we've seen new data protection laws. And like last year, India, Indonesia, Saudi Arabia, Vietnam. Have you seen any effect on the, on the data protection law in India, for example? On all these enterprises doing business? I mean, maybe it's too early to say. Yeah, I think it is too early. If I'm not mistaken, in India, it's still an act. I haven't kept up. So I'm not sure if it's already a law... Yes, it’s a law. It's a law since last year, I think from October. I think they tried for several years and they didn’t manage, but then last year it became a law. It became a law right, I see. Okay, I'll have to look into that. Okay. All right. So looking forward to 2024. What can we expect? You know, we talked about more data protection regulations. We can expect governments expanding, you know, privacy regulations. And based on the fact that the three main hyperscalers are now offering, you know, quote unquote, sovereign clouds, I mean, is that evidence that regulations have some teeth to them? What do you think? Yeah, I see GDPR finally is starting show its teeth after many years of not being that proactive. But, now I see that is happening quite a bit. Talking about what's going to happen in the near future, I have to say, we are seeing a very interesting tradeoff between having control over your workloads by bringing in the best-of-breed tools versus going to a cloud provider and taking the integrated offering, where you don't have full control, but you get a few things in your favor. For example, Microsoft last year introduced this concept of Microsoft Fabric. I feel that Microsoft Fabric is made up of a lot of these parts. It's got the, the database, the application environment, ingestion, ETL, data transformation TAC log… a lot of these things are cloud pins. What I'm starting to see is that a lot of customers in the current economic times where there's a lot of pressure on containing cost and accountability, are saying that,“How do we simplify our technical infrastructure? Do we really need to integrate the best-of-breed products, which actually provide very deep technology advantages but I have to pay to integrate these? And then if something breaks, I have no idea. Did it break in Fivetran, did it break in dbt, in Looker, in Snowflake... why is my dashboard showing me the wrong information?” But if I go to Microsoft Fabric, then maybe even from a regulation point of view, it is one neck to choke as opposed to five different ones. So you see, it's a trade off. But then, welcome to vendor lock-in. Yes. Correct. I mean I totally agree. But I think, you know, in general in times of tougher, you know, tougher economic, conditions, you would say that… people wouldn’t, they wouldn't be too brave and try to like, let me do this and that and this. I mean, it's probably simpler, as you say, to get a platform where you don't have to take the cost of integration. It's the vendor that has already taken the the cost of the integration. And you're getting it packaged, so to speak. And then you have this one neck to squeeze. Looking at the economic environment, you mentioned it earlier on this call, right? We had big investments in tech during the COVID years. Right? Then after that, companies started to cut down on costs. 2023 was a tough year. And the question is, “Will ‘24 be better?” You sort of mentioned Q1, Q2. It's kind of wait and see and then we'll see what happens. But let's say if you would look further for the whole of 2024, right? Are we looking at a repeat of ‘23? Because I guess we will continue to suffer the effects of high inflation. How does this affect cloud spend? Right. And will, by the end of the year, will enterprises have diversified their cloud diet? Right. Being cloud smart as opposed to cloud first? I don't see a repeat of 2023. I have very high hopes. But then my wife calls me an insufferable optimist. And there's a reason why I feel 2024 will be different. 2023 was driven by all these geopolitical things we talked about and financial market. At some point we have to put a kibosh on optimization and start innovating. We cannot just keep optimizing forever. I feel in 2024, we’ll see a lot of consolidation happen. Some companies are going to suffer. You talked about job losses. Yes, a lot of job losses took place. But then they also over-hired and we had unrealistic expectations. So there had to be some rationalization of workforce. And you know, I feel bad for people when that happened. It's even happening in 2024. We're only two weeks into this year and we've seen job losses at Google Cloud, at Amazon, Cloudflare a bunch of companies lately. So it's still happening, but that's because of the uncertainty. This is the year when we have to say, “Okay, we've optimized, we've cut it down to the bone. Now how do I get more of a competitive advantage? I have to invest and maybe AI is a vehicle that I use to invest.” But it will turn around is my sincere hope that this year we will see more investments because you cannot stop technology from marching. To all those people who are saying,“That AI is going to take over our jobs, AI is going to take over the mankind and EGI is going to subjugate us,” I am like, please stop. We don't really know. Let’s just use AI for the good it can do to our organization. There'll always be bad actors out there. No matter what is a new technology we cannot stop that. But this is the year to take a hard look at AI and see what are the use cases that I can put it to make my business more efficient or to improve customer experience or to improve developer productivity. There are just so many. There is no dearth of places where AI can help. Yes, it's not perfect. Generative AI may hallucinate Things will get better just like any technology. I’m telling you 1994, when I first got into World Wide Web, the whole thing was just coming into existence. People were complaining,“What the heck is World Wide Web? They are just static web pages, it's just documentation.” And look what we can do now on the internet, pretty much everything. Yeah. So, yeah we'll get there. And by the way, we know that with every new technology, the runway to that level gets shorter and shorter, from radio to television to the internet, World Wide Web to now AI. So another 3 to 4 years, some of the things would be embedded with AI…it’ll be table stakes. We wouldn't even think about it. Like we don't think about going to a bank in middle of the night, which you cannot, to withdraw money. We just go to an ATM machine and it's just no brainer. So I think you see more of that starting 2024 onwards. And actually based on what you said earlier, if we're looking at AI being a driver for new investments, then you take AI to where the data is, right? It's not like you're going to shift and lift to somewhere else, right? So basically you have your workloads in your on-prem data centers or whatever. Then probably you're looking at augmenting that environment with AI capabilities and getting the benefits of it. Yes, Right. Okay. Well, Sanjeev, this is great. You know, time to wrap up. And I mean, always great to have you with us. So, you know, we've looked at 2023, how it has been tough economically. Right. And probably much worse, was it not for AI-driven workloads. And then it remains to be seen whether enterprises will avoid concentrating their workloads. Right. On the hyperscalers, or maybe adopt a more sovereign approach. Right. I mean, maybe this taking AI to the data would also match the sovereign approach because, you know, I think, I saw somewhere that only 50% of workloads are on the cloud. So there's still, most of it is not in the cloud. I think it’s less than 50. It’s reaching equilibrium of 50%. But I think it's…again it's very hard to pin it down. But I think, from what I see it’s like in 40s, 40% range 46% yes, one of the surveys I saw. Unless you talk to AWS, they'll tell you it’s only 15%. Yes! Yeah. All right. Well, thank you for joining us, Sanjeev. This was, you know, very appreciated. And thank you to all our listeners, for joining us and have a great 2024. Thank you so much. Bye!