Harnessing the Potential of Gen AI Investment

[Jeffrey Teh, Webinar Host] Good morning ladies and gentlemen it's also good afternoon to you too. 

Welcome to today's webinar on reimagine infrastructure: harnessing the potential of Gen AI Investments. Now we are delighted to have you all, and can't wait to dive into insightful presentations and discussions from our speaker today but first let's take a moment to introduce our speakers who will be sharing their insights with us today. 

First up we have Truman Crotty Data Center and Hybrid Cloud lead for Southeast Asia region at Telstra International.

Truman, I know that your expertise lies within hybrid Cloud adoptions. Can you walk us through what has changed or updates with the current environments at the moment?

[Truman Crotty, Data Centre and Hybrid Cloud Lead SEA at Telstra International] Yeah absolutely, so what we're seeing is more of a focus on where data is sitting as part of this gen AI transformation so whether it's coming from the hyperscalers or being generated without The Edge we're seeing all these components come together and I think this webinar will cover off how we see the Gen AI future ever progressing. Okay good, great. 

Second speaker we have Patrick Lastennet, the director of platform and enterprise at Digital Realty. 

Patrick, I know your expertise lies within modern hybrid architecture can you walk us through a little bit the latest developments or changes you've seen in the market.

[Patrick Lastennet, Director, Business Development – Enterprise at Digital Realty] Sure, Jeff um I think what we're seeing today really is a a compelling business case for Gen AI, we certainly see that the CTOs at Enterprise level are now considering moving away from what we call Shadow AI where basically the business undertakes AI use cases strictly in the cloud and they're really looking at architecting specifically for AI taking into account data center challenges and connectivity challenges, so I'm very excited to delve more into those challenges today with Truman.

Great, thank you so much so we're very fortunate to have Truman and Patrick with us today. they will cover a wide range of topic but the following are the most important topic that we cover  first thing we're going to talk about modern hybrid architecture, leveraging hybrid Cloud agility, optimizing infrastructure return of investment really, maintaining strategic control over data. 

Now, to ensure that this is an engaging and comprehensive discussion, we have structured the webinar into four segments or five segments but we also need your participation at any given time.

If you think of any questions or any you know lightbulb moment please feel free to type it on the chat box or questions and I'll be able to relay back later during the Q&A time.

Okay so the four segments we're going to talk about is adoptions, AI adoptions model, infrastructure, partnership between Telstra and Digital Realty

Each segment we will dissect and provide you a bit more detail on how we get into the world of Gen AI investments.

Now after the presentation we'll open the floor for questions like I said, from all of you and once again i'd like to extend a very warm welcome to our participants dialing in from Malaysia, Indonesia, Singapore, Hong Kong, I also believe and also New Zealand. 

Today's webinar promises to deliver very very good knowledge I've seen the ppt really to inspire you how you imagine your infrastructure to harness potential in Gen AI investments.

Now let's begin with segment number one: AI adoptions.

Okay, both gentlemen AI is at the forefront of many minds today and all our executive thoughts today. 

Patrick and Truman how do you envision, I mean the the impact of AI on businesses and industry?

Let's start with Patrick first to sort of give us an overview, and then we follow with Truman, go ahead.

Thank you Jeffrey, so I think you know obviously there's a lot of talk about AI. We've all heard the iPhone

moment reference from Jensen at Nvidia, we can see all the hype the buzz around AI but it's important to see that it's part of a wider picture. I think it's a trend that we've been following at Digital Realty. 

We see us move from a digital economy to a data economy. So imagine all of the industries, all of the vertical sectors, imagine that digital transformation if you push it to the limit and really digitize, all of the workflow the processes, then arises economic opportunity and additional revenue opportunity and once really all of this digital transformation is achieved, and you can leverage the data which is associated to it. We measure the overall opportunity to 100 trillion dollars. 

So this is a huge opportunity and hence really, you want the urgency and transforming in order to leverage this opportunity.

Your thoughts, Truman?

Completely agree with Patrick on the huge potential of the AI technologies that you know disrupts competition, you know the industry is now entered what Telstra is defining as the hyperconnected AI ERA.

This era is categorized by a landscape of distributed work workforces, decentralized applications, external vertical ecosystems, and the proliferation of machine to machine and human to machine digital interactions.

In 2024 and beyond, companies will initially explore and then widely adopt AI technologies and platforms.

Telstra International has partnered with MIT technology review insights on a global report on generative AI  which we released in a little over a month ago, so it's really hot off the press. 

So while there is emergent areas of AI including edge-AI, intelligent applications and model authorizations to name a few, the focus for us with MIT was just on Gen AI and we did this for one reason, oh sorry, two reasons. 

We did it to make sure that we weren't a mile wide, we were an inch narrow and the other one is that we wanted to focus on Gen AI which at the moment is top of mind for all of our customers and companies.

The report garnered knowledgeable perspectives from enterprise business and IT leaders around the world on macro industry trends, their AI adoption journey, and their areas of focus in 2024 in corresponding challenges.

So you know the first thing I want to highlight in regarding to the report is the survey respondents.

Only 6% indicated they had neither adopted Gen AI or have no plans to adopt Gen AI in 2024.

Secondly, besides the minor inevitable variations that you see, the trends and insights for all these people were consistents across every industry.

So by and large every firm had ambitious plans for Gen AI adoption and in fact based on our report the vast majority 78% sees Gen AI as an opportunity to grow their business and disrupt their competitors and those views are likely reflected across everyone that's sitting in the in attendance right now. 

And thirdly I would say driving into some of these key findings of this graph that you are seeing in front of you as we asked the respondents about their existing and upcoming Gen AI deployments to re-emphasize the graph represent since 94% of the respondents have either begun adopting Gen AI in 2023 or have plans to move into 2024.

And based on our survey results, the initial or subsequent Gen AI deployments are focused on cost saving and efficiency gains. So for example, over 50% of the respondents say they have already applied Gen AI to the automation of repetitive tasks and will continue to do so in 2024.

So it makes sense you know the nascent technology that organizations use typically start with POCs (Proof of Concepts) where there's a high tolerance of business risk the low value repetitive tasks are often less uncertainty or complexity. So therefore Gen AI on this level has that less risk of generating incorrect or fabricated information. And then similarly through the recent maturing of tools and technologies, Gen AI can create and apply easy service scenarios to reducing resolution times, reducing customers dissatisfaction, and improving customers experience. 

So for those customers who are typically more advanced in their Gen AI journey, we explore more sophisticated user cases. And so let me share a couple of examples. You know AI models help create prototypes of products by generating 3D models, simulations and renderings. These capabilities have been available in some form and another over the  vast number of years. However, Gen AI is now being leveraged to also quickly analyze consumer buying behaviors and predict market trends. And the ability to spin that around can now start shortening time for purchases but then also bring new products to service markets. 

So it's that sort of inflection point that we want to get in place to foster a rapid cycle of innovation going forward.

Okay great, Truman you actually mentioned in your response also the research itself, a lot of people are not Gen AI ready I would say. And during our invitations and our audience as well I can't agree more, we have a lot of audience are now gearing up to it, what shall we do and that leads me to segment number two. 

Okay, now there is a lot of a a lot of different AI models out there. Last time we spoke, you also demonstrated to me there are this lowend there are this high end, and so we understand that there are different AI models and different characters use cases as yours we'll share later. Infrastructure and data requirement, can you provide some of the insight into various types of Gen AI model? 

Yeah yeah, absolutely. I think broadly speaking Gen AI as you can see on the screen is broken down into three aspects.

You have the general purpose way out on the left hand side, you have the fine tuned in the center and then finally you have the domain specific,  and I'll take you through very very quickly through that.

The general purpose one these models are designed to perform for a wide range of tasks so these are the chat gpts of the world, so they trade extremely on large data sets and diverse resources enabling to generate text as you see, images, software code and other components and they handle diverse queries and perform multiple functions making them highly suitable for general use where specific expertise is not required. 

Again as we said, chat GPT, image generators, language translators, general purpose AI is designed to perform well across a wide large set of data, then you go to the next point and the next layer which is the fine tuned, and it's in essence general purpose but now it goes through a further training that it adds additional data sets that are specific to your company or your business. These data sets augment or complement the data that general purpose has, and the process allows the AI to perform better than the original general purpose model for those tailored infrastructure. 

So for instance, a chat bot fine-tuned for medical literature aims to provide more and relevant answers and healthcare related queries than a general purpose environment. And then finally we have the domain specific and I mean this is the big end of town you know, it's specifically created for a specific task. These models use data sets that are domain specific, they are superior performance, they have superior accuracy. So what you see is you know you run those simulations for drug discovery, predictive maintenance, in manufacturing or fraud detection in finance, so you know if you're asking specific questions they will absolutely knock it out of the park. 

However, if you start asking him for larger more broader terms on your data then it sort of disappears without a trace. 

So I suppose the integration of everyday operations and culture referred to the AI and the business dynamics underscores  the employment of aligning AI technologies with business processes, so you need to find the right place at the right time for your environment. So again that's where we need to pull this all together and so I would ask the team and and the people on the webinar  to start looking at that fine tune and how that fits in your strategy, because if you want to remain competitive in the hyperconnected AI era, your organization must begin with Gen AI across all the functionalities of its business. 

Because most organizations typically build up on the Gen AI and think that they've got it made, the problem is as you start adding that fine tune data in that's where you start seeing the better results, and you start gaining that critical  momentum for your business as they move through the Gen AI process. 

Very well said,  Patrick your thoughts?

Oh sure, I couldn't agree more with Truman. This is really critical. The interesting thing is we're starting to see enterprises forming an opinion on this, and the the emerging trend that we see is is really around the fine-tuning or what we call private AI which is effectively importing a public model and using it against private data within a private environment. But I think you know, everything is possible, I think it really depends of your business cases, typically if you were to do the first scenario where you basically just a user this a OPEX only model, that can be useful as well, appropriate then fine-tuning there's a little bit of investment which is necessary obviously to actually procure the infrastructure to support the additional GPUs that you're going to use for fine-tuning the model. And obviously once you go really proprietary you want  to have your own GPT like for instance Bloomberg GPT than this you're talking like you know tens and hundreds of million of investment which could be justified if you want to maintain your sort of leadership within a specific sector. 

So I think it it really depends on the maturity of the of the of the set enterprise, but the majority of our customers or the conversations we have  definitely are around fine-tuning and private AI.

Okay, Truman and Patrick, you both mentioned a lot on data and also just now touch by on architecture. Now from what I understand I think the audience may agree with me, data management is highly important if we want to get this right. Am I correct? 

Yeah 100%. Off the back of what Patrick said around fine tuning is just absolutely integral to make sure that your Gen AI environment is working the correct way. I mean it's the important and the focus of fine-tuning those Gen AI to create that competitive advantage, you know, you continually need to refine the data sets and ensure that you've got the right data in the right place.

From an AI Report with MIT, it was deferment that you know 30% of the respondents rank their company actually, readiness to create it is not in place.

So we ascertained that many organizations underestimated the requirement for an effective implementation leveraged by their propriety data, so you know, drawing to your attention to the bottom row of boxes, 19% of the respondents recognize they need to uplift the volume of data  to their envisioned AI user cases, but most notably only 13% and 7% respectively indicated they had the necessary data storage infrastructure and compute platforms for their AI addition requirements. So in mind those folks are really you know, we are only starting their journey and they're not ready for it so  you know if we consider the early adopters and and leverage from what they've seen, we've actually seen strikingly companies who have progressed most  you know in their AI transformation have even less confidence you know, eye-opened in their current IT infrastructure so they can see they need to even spend more  infrastructure and investments to create the best Gen AI environment for that so you know put it in another way you need to start thinking about your infrastructure strategy now, regardless where you are in your journey whether you're out the front or out the back you still need to have a look because you know people are not getting it right first go unfortunately. 

And then finally back to you know Patrick when he was talking about coming back into the private environments,  you know data security privacy we know what happened with GDPR in Europe, they are changing the rules as we speak right now, and people are now what you see is moving away from hyperscalers and putting it back into a private location for two reasons. So they can access their data, they can protect it, and they can leverage at that as well. 

So you know therefore both the IT business reasons you know are go hand in hand, the cost considerations, and we see this actually the pendulum swinging  back to a private onsite environment. Okay Patrick, before you continue you know just to add on what Truman just said is that the obviously all audience today you must actually consider I think Truman can help later as well during the Q&A is the type of challenges you will face when you think of how to set up an infrastructure in the context of Generative AI.

Now Patrick continuing on what Truman said, can you leverage I mean elaborate a little bit more on the infrastructure side of things? Sure, so the way I look at it really in terms of infrastructure really three pillars and which are linked to the AI workflow. The first thing you need to do is to ingest and consolidate data so you have a storage area, a data lake so that in itself is a choice to make, whether you know you will have something proprietary or in a public cloud or in a cloud should we say. Typically if you are considering the associated compute if you were to run it in your own data center you look at you know densities of around 10 KW per rack potentially for for that type of storage. Then obviously once you have ingested your data you need to train it this is where all the magic happens you train your models, you use GPUs and accelerated compute at scale, and there this is where we see the biggest challenge certainly in terms of cooling and you're looking at if you run it inhouse or proprietary you looking at racks of up to 80 KW 100 KW per rack should we say with liquid cool service. 

And then the last stage is really the inference, which is once your model has been produced it needs to run into production and again, you still need to use accelerators GPUs, TPUs but the density is probably a little bit lower, it's very dependent on latency, latency sensitive and they potentially you're looking at 20 KWs per rack. 

For these three pillars you potentially can deploy in the cloud and obviously you then need to take into account you know, are you going to have elasticity, can you sort of use the cloud to its benefit, or are you at a stage where you already mature and you lose the elasticity you constantly run these things, in which case it's probably more appropriate to consider private environment. So I would say again, it depends on your maturity and then in terms of capabilities I think this is where the challenge starts to be if you want to do it inhouse you need to start investing really in your data center because we are are at an inflection point when it comes to cooling technology and cooling techniques. We're moving from a standard of 5 to 10 KW per rack to something which is going to go, as I said there's no limit for the training 100, 200 kilowatts, and on an average we're going to hit the 15 20 KW so you know certain companies can do it they they make the choice to invest directly in their proprietary data center but this is really the exception. And this is why we also work very closely with all of the OEM providers on the service side with Nvidia to really understand you know the future road map and the future technology challenges. 

Okay I also understand that most of our attendees today have different workloads environment as well. 

Now again, Truman and Patrick, you mentioned a lot on infrastructure and also related to integrated data and data management. 

Patrick, just now you mentioned effective training model for Gen AI to ensure its integrated well. Can you elaborate more on that, if you can talk about what sort of data challenges they will face, as well as what are the ideal solutions for this particular setup.

Then I'll get Truman to contribute on that later, would that be okay Patrick?

Sure, so you know I've just described a little bit what was happening within the data center in terms of densities and compute and now in this picture really is to illustrate a little bit our reason of being, which is essentially the move to hybrid IT. So you know I think everyone is familiar with hybrid cloud, whereby you will go to and from cloud and edge, now with the advent of digitalization the data economy and AI what we getting to is a point where you will have severe bottleneck between the edge and the cloud. And the way to solve this what what is needed really is to go from a two-tier architecture to a three-tier architecture, where you start to introduce a zone of data exchange to literally alleviate you know some of the bottlenecks. And what we think is so this was this has always been our reason of being but now that AI is really taking off, you're going to see it even more acute you know the the amount of data which is generated at the edge you know I think of the overall data is 90% plus, so in order to get your data assets in order you need to start thinking about this type of architecture and not just rely on your edge or just rely on your cloud or just rely on your on-prem data center, you really need to think about zones of data exchange and that helps you address problems that we call data gravity, which is literally means that you know all applications and compute will tend to gravitate towards you know critical masses of data.

Truman, you know Patrick mentioned a lot about data and now this is time for you to show your real expertise, especially on the data gravity side of things.

Go ahead. 

Yeah look, as Patrick said and let me reiterate a key point that we've raised is and I'm sure that everyone agrees is that data generation and usage continue to grows at an exponential rate the amalgamation of physical and digital data across the world  is created an accelerated amount of data because of that machine to machine and the machine to human you'll see digital reactions that now create more data. So AI is actually generating more data, and then that follows into my second sort of point is data gravity that also Patrick turned about whether the data is sitting out on the edge, it's sitting out within the data centers, it's sitting out with the the humans, as well on their devices, it's how to pull it together and what we see with data gravity which is not a a new term is how we can pull it all together and compose it because it's a critical consideration on a successful Gen AI deployment. And then thirdly is the hyperconnected digital infrastructure and what do I mean about that, is that we work with our customers to understand the critical importance of right fit, right size, and right locate approach, to their data and their AI deployments. So right fit focus on determining the most suitable technology to run your AI applications and underlying data estate. Many organizations will need a radically simplified on their data infrastructure to maximize their business outcomes. The right sizing comes in where we match resources to workloads making sure that you are running a digital infrastructure in the most efficient manner possible. You know for example hyperscalers consumption model can be suitable for many applications. But IT leaders now need to consider the scale of the data as we previously discussed as well as the volume of computations that their AI workloads will demand. And then finally we have the right locale, focuses upon ensuring your application AI models, and datas reside in the IT environment that is the most suitable for the workload and business requirements for modern enablement workloads weighing up considerations of latent sensitivity, data sovereignty, as well as compute, and storage economies of scale, can be highly complex. 

So just to close in summary, the hyperconnected AI era organizations need to be cost effectively balanced, exponential growth, and also demand for data as well as competitive strategy going forward. So as you bring all these things, we have to meet the core business objectives of performance, security, and then regulatory compliance. 

All right, I understand that some of the concern I think we have question coming in just now as well is about compliance, regulations, privacy,  we'll walk through that later.

Patrick I understand yourself, and also Truman work together a lot on the project.

Can you Patrick, take us through some of the AI deployment examples you have seen from Digital Realty's perspective? 

Sure, so I think you can see three logos here on this slide and I think they were quite representative of the type of early adopters that we see in terms of you know scaling AI, starting on the enterprise side with Kakao bank so typically this is a new Bank running in Korea which is all digitized so if you remind yourself a little bit about the stages we were at digital economy and data economy and this is a prime example of an organization which is you know completely embracing the data economy and literally using AI to produce mortgages you know really leveraging all their data assets and the power of AI, to generate new products. So typically Kakao bank has chosen to deploy in colocation a compute cluster with some Nvidia 800 GPUs, so very good example of a one of the early adopter should we say. Now the second pillar we see CoreWeave, so this is as well a new type of company they are involved in providing a GPU Cloud so this is a cloud which is purely aimed at enabling AI workloads, both on the training and the inference, and they really have developed  a core competence in deploying very large scale GPU clusters and running them extremely efficiently. 

They really master the challenge of scaling infinite band and connectivity between all of the servers so they really deploy at scale with us in the US and globally, and thirdly a good one to mention as well I think is G-Core Labs which originates a little bit more within the gaming industry CDN but which has sort really distribute compute all over the globe and is now positioning themselves to be a provider of infrastructure for inference. So if you think about training versus inference for AI, training tends to be quite concentrated centralized maybe you'll have one or two training zones per region, whereas the inference need to be distributed next to the users and this is where these guys are positioning.

Great great great. Well thank you so much Patrick and Truman.

We have covered quite a bit, now before we move to the partnership segment I would require yourself and Truman to just sort of think about what muscle deflex but before do that ladies and gentlemen I know you're dialing in from different country. If you have any question now think about it and put in a chat box, I already have a couple of them I'll read it through later. And you know throughout the our webinar today if you have any questions that you want to reach out to Patrick and Truman, we will also share their contact later.

Okay now let's move on to the partnership questions. Now what you just mentioned it sounds very complex to deploy AI  deployment. How can services provider like Telstra and Digital Realty assist companies that are from all over Asia today and New Zealand deploy large scale AI more? Now I understand the friends aside, Patrick you love Truman's sunshine living barbecue and beer, and Truman you love the where Patrick have lot of French wine and great european food, but when it comes to working together helping client how do you both do it? 

Shall we start with, let's see, Patrick? 

Well Jeffrey I think it's a little bit the same as you know the analogy that you just mentioned it's all about complimentarity isn't it so the beauty of us, I think is is you know we really specialized in running efficiently a global platform of data centers, and you know Telstra really is if you think about connectivity, professional services, so there's great complimentarity, you know if we align our capabilities on both sets of organization then all of a sudden we have a global set of capabilities that cover pretty much all of the major, I mean I would say the part of this stack linked to infrastructure. Okay so that's the first thing. And if you're an organization and you start to think well

I can't do everything in the cloud, therefore I'm going to have to start going hybrid, therefore I have potential infrastructure headaches, I think this has got tremendous value. And you know the the positive thing is once we get together and solve these infrastructure issues it means the CTO and CIOs can then focus on high up the stack you know in terms of all the intelligent things they need to do at software and at business level. So I think this is really at a high level how we we help enterprises and and it's a very complex ecosystem at the moment. It used to be, it used to sort of start to rationalize with hybrid IT and cloud and all of a sudden you have AI and you have all these different breeds of providers of infrastructure providers, and it's starts to be a little bit complex for enterprises to navigate the range of solutions and I think with Telstra in particular with their professional services I think we're very well positioned to provide solutions and advice on these complex challenges.

Okay.  Yeah, no I'll follow up on Patrick because 100% you know, with the digital networking components within Digital

Realty and their landing space and their power and calling creates a very complimentary solution. I mean just in one hand you know we have worked very closely with a large European Bank with whom you know Telstra developed a strategic relationship with Digital Realty and they were looking to future proof  their operations by enhancing their network core infrastructure data management systems. So you know the solution for them that we provided both as a partnership was to address, we collaborated with Digital Realty to transform their technology environment and this transformation involved migrating their outdated systems their data center from their legacy data center provider we put them into a state of an art facility, we upgraded all of their core infrastructure, we realigned all their networks across, and we modernized their software and then enhanced their connectivity to ensure they have a seamless transition. 

So what this built for them was you know a successful landing zone where they could put their data lakes in place to leverage them the best way that they could get for their Gen AI. So they were looking to put in a customer service environment, so back to the point of fine-tuning the data sets that they had in place. In essence, Telstra and Digital Realty put that all in place for them. So it was a resounding success, the conclusion we revitalized their infrastructure, so we modernized it, we future proofed it, we had new AI initiative put in place. So you know the case study is that you know not only does it show you know how well we can work together but how we can help our customers you know to the point of reducing that complexity, let our customers focus on their AI transformation, we're absolutely here to help them, but you know we can put you and then give you that leap frog into the new environment. And you know it was a resounding success and you know this is the reason why we're sitting in front of you right now. 

Okay great. Now for most people in New Zealand they probably know what is Telstra and if they do some research, pretty sure they did, they understand Digital Realty have a lot of footprint in the global. I did some research myself and I noticed that if I'm not mistaken, Telstra has a lot of lines that help clients. Am I correct, Truman? 

So down into the the Australian area we're including New Zealand, yeah I mean Telstra is a large corporation that you know covers the mass. So we do massive underwater subsea cables, you know we have thousand undersea cables put in place around the world.

So in essence you know, if you want connectivity, connection to the internet, highspeed low latency networks, you know Telstra is your provider. But the other part is that we integrate into the Digital Realty data centers, so you know in essence they're diverse in their network PoPs. However, Telstra is predominantly in every single data center for Digital Realty around the world as well.

Right Patrick, you guys have a lot of setup around the world. I mean, most of our audience are dialing in from

Asia Pacific, this including New Zealand. So if you can walk us through who are you really because we didn't basically get you to that. We now understand the partnership, but we also want to understand the company itself. 

Sure, so we well so we operate data centers across the world as you understand. We a platform company which means effectively, we run a platform global platform of data centers, so we have 350 data centers across the world.  20 of which are in APAC, we are in about 50 plus metro so really global coverage. So when we talk about a platform think of it a little bit as a meeting place, so in all of these data centers, they all are occupied by essentially enterprises or service providers and the power they benefit from actually interconnecting with each other within the data center or in between you know the data center. So we run a global platform which allows enterprises to scale globally, and really have a similar experience whether they are served in Dallas, or Sydney, or Singapore. But we really truly believe in the power of partnerships, the power of three, hence the partnership with Telstra which also helps as we said you know really sold for the enterprise at a higher level in the stack.

Okay, we have one question coming in actually a few, but they geared towards similar target, I mean similar objective which is ESG.

Now, when it comes to data center, massive deploying of data management infrastructure that obviously ESG is a key question for a lot  of our audience today. So let's discuss ESG goal and accuracy reporting. 

Okay now, for companies that have set their ESG targets and building their infrastructure for AI model, how can Telstra and Digital Realty help meet the ESG goals?

Patrick, would you like to start first and then we go to Truman? 

Sure, so I think the first thing is I'm really glad that we bring this up you know it can sometimes be a little bit the elephant in the room when we see all of this growth and compute which you know AI necessitates.

Obviously AI can help solve some of the climate change are challenges that we have, and and I'm sure it will do.

Nevertheless you know, it is very important for us to be as efficient as we can when it comes to energy and I think I'd like to take a few examples, proof points that we have in APAC. For instance, what we strive to do really is to actually use AI to optimize the energy efficiency within our data centers. So we do this with a platform which is called Apollo and via this platform, we can detect all of the inefficiencies that might occur in a data center; leakages, faulty valves, and things like this. We've deployed this across the 20 sites in APAC and to date we can say that we are saving around 1.8 gigawatt per hour, that's what we have saved so far by the virtue of actually optimizing and detecting inefficiencies. Another big challenge in APAC is obviously water saving and if you look at what we have done in Singapore with new initiatives around some of the cooling towers that we use was also being able to to save about you know 1.8 million litres monthly of water alone in Singapore 10 and with this initiative actually we won some awards from the the Singapore Environment Council 24, Singapore environmental achievement awards. 

And the last example I want to give is around the what we call the sort of virtual power purchase agreement that we do around, that we have done in Australia in particular so imagine what we try to do is whenever we build a data center we try and match it with a new source of renewable energy and if we do this, it means that by virtue of building a data center you actually reduce the carbon intensity locally. And I think this is really important because it's all quite easy to sort of deploy your workloads in places where you already have renewable energy but you know what we need to do is solve the problem locally and examples in Australia where essentially 100% of our Sydney power consumption our data centers is matched by renewable energy, and in Melbourne it's actually 60%. So you know we contributing to the saving of about 70,000 ton annually of CO2.

Oh wow, so not only that you actually have global best practices you also deploy in know local. 

Absolutely.

Okay Truman, your thoughts on the ESG question? 

Yah look, for Telstra doing business responsibility means doing the right thing for our customers, our people and our community at large.

We have the responsibility to operate sustainability to actively consider the impact that it creates, for our customers, our communities and the environment. This is why doing business responsibility, is one of the key pillars that Telstra has within its manifesto. 

You know, we take bold climate actions to protect our environment, and so as as such, Telstra has been carbon neutral in our operations since July 2020. We also aim to reduce our emissions by 50% by 2030 and so to achieve that, we achieved a 19% reduction from FY 23, and then we're trying to get this down to a cumulative 30% now from FY 19, so as you can see we've gone from 19 to 30, and then you know we're on our way to the 50%. And then finally we're trying to decarbonize our grid through greater investment in renewable energy, and thereby reducing our non-renewable energy via fossil fuel. So our target is to enable renewable energy generation equivalent to 100% of our consumption by 2025. So you know we're committed to the ESG standards and our statements going towards our shareholders, our community, and also the environment at large.

Wow, that's really good to hear, I mean among all our audience today I think a lot of government as well as corporations, they move towards that goal as well. 

Now ladies and gentlemen, thank you so much for your participation. That brings us to the end of our sessions.

Again, if you have any questions, write them all down. Team Telstra, team Digital Realty will reach out to you, what you've given for your one to one's consultations on how they can actually build Gen AI and your return investments for you, right now. 

But before we end, one last question. 

What actionable step would you recommend for the audience to take towards your journey in adopting Gen AI? 

What are the key considerations, most important thing that they should they should prioritize really?

So I can lead away. So number one for me is I encourage everyone to download the MIT report that Telstra and MIT created  because I've only shared a tiny tiny little part in today's session, and the report has many compelling insights from IT and business leaders  as well as industry experts, so that link will be in this webinar chat. I think number two, summarizing a couple of areas is,  the organization about the progress of the AI journey and increasing their trajectory, focusing on the fine-tuned Gen AI capabilities. General purpose is just the entry point, but fine-tune is where you need to be. 

As our report with MIT highlights, organizations will only achieve competitive advantage and necessary IT infrastructure using that fine-tune environment. The challenge many IT business leaders are now facing is dealing with the immediate scale of data.

So you know in essence, we need to focus that on that as well. So that gets us to the point of the right fit, right size, and right locate as my final component because once you have those components in place, you're well on your way. So I think you know as everyone understands to succeed in this landscape, it's critical you proactively and holistically incorporate AI into all facets of your business and try to get that in place. You know Telstra is supporting our organizations with Digital Realty with assessments that we have going forward so I'm happy to then sit down with the customers and then sort of listen to them, what their requirements are, and then sort of promote our Purple services or professional services to create an AI and data envisaging customer workshop going forward, because I think once you have that, then we're well in a way to creating a successful Gen AI environment for all of our customers going forward. 

Okay, Patrick?

Well Jeff, I think the main thing is to think you know don't leave your infrastructure challenges at the end, and the way we look at it you know there are certain design imperatives that you need to take into account when you architect for AI. The first one is you need to solve for latency, I mean obviously there is a lot of compute going on you know you need to train new models, but latency is also very important. Imagine when you deploy your AI in terms of inference you know it's going to be latency sensitives, and your ROI for AI is going to be depending on how well you tackle latency. There is obviously risk that you need to also take into account you know failure to manage the risk can result you know in a damage business reputation, imagine compromising all your data assets that are powering your Gen AI strategy you know that can be a huge huge issue, I mean complexity is a fragmented environment you really need to take into account this complexity and have a proper design in order to pave the way for your future architecture.

Capacity, I mean capacity is such a challenge if you look at the scale at which AI and Gen AI is taking us, so we are there to help you with planning this with that Telstra as well at the network level and obviously sustainability in particular I think the regulation is actually coming into place which will force enterprises to be a lot more precise into their carbon footprint. And again with Telstra we have all the tools to help you so from a Digital Realty perspective we have an ebook, an AI ebook which we invite you to download. We'll make sure to include the link and all of these five themes are well described in there and we'd be happy to exchange a bit further also with our solution architects on these topics.

Well, well said Patrick. Thank you for joining us today. We hope this webinar was informative for you, like I said, we have both Patrick and Truman who already know everything. I think you should just save some time, just contact them build everything for you and let's us make the most of this opportunity to learn to grow as well to learn how Gen AI works. Thank you so much for today, bye-bye.

Cheers!

[Music]