Embracing Physical AI at Intel Vision 2025 || Tech Field Day News Rundown: April 16, 2025

Trump shift stance on Nvidia chip ban. Do you remember? Windows 11?

Recall? It's back. And solo IO adds an MCP gateway to its API platform.

Intel is selling off 51% of their all terrorist chip business, and it's going to Silverlake partners. And we will, of course, look at embracing physical AI from in television 2025. All this and the greatest of news today on the tech field day rundown.

Welcome to the Tech Field Day rundown, where each time we meet, we run down the IT news of the week with a variable degree of snarkiness. I'm your host Alistair Cook, and joining me on what is National Eggs? Benedict Day is my good friend, Mr.

Ron Westfall. Welcome to the show. Ron, are you ready for, uh, that great sinking feeling day?

I'm fired up, Alice. Sarah, it's great to be here. And I think we'd be neglectful not to mention what happened on the eve of this day.

And that was 1874, the first impressionistic exhibition open in Paris featuring outstanding artists such as Claude Monet, Edgar Diggas Pierre, August Renoir, Camille Piro, and Keith Richards. So I'm just excited to, you know, be here and also comment on yeah, that great sinking feeling day, which is when the Titanic sunk and it happens to coincide when us folks need to file their taxes. So, you know, is that coincidence?

Uh, what do you think? Well, I think that great sinking feeling would absolutely fit. And being based in New Zealand, I try very hard not to need to file US federal taxes.

And so far I've succeeded. 'cause I hear that it is quite the trial. Fair enough.

Yeah. I imagine it's different in other parts of the world, but it it can be a trial for many of us in the us. Indeed.

Speaking of trials, the Trump administration has paused plans to restrict NVIDIA's export of H 20 AI chips to China. This decision followed a Mar-a-Lago dinner attended by the Nvidia CEO Jensen Huang. It's a great place to make businesses at Mar-a-Lago, uh, after which the company pledged more AI infrastructure investment in the U us.

So, despite bipartisan concerns about national security, Chinese firms like Bite Dance and Alibaba have already placed large orders for the H 20 chip, which remains the most ai, uh, advanced AI processor currently allowed for sale in China. This change on restrictions, this move to allow more business to happen in China, and particularly AI in China. Ron, is this a sign of changes in the water or is this a simple commercial deal?

Well, you know, that's, I think a great way to start off. Imagine that Nvidia ai, it's still filling up the headlines and it's what been over two and a half years since it's really, you know, uh, cop captured the intention of the entire ecosystem. And I say it's a big deal, and accompanying the announcement is a fact that Nvidia plans assure up to half trillion dollars in the development of its chips and AI infrastructure.

So this was a big wig dinner for sure. And what I think is going to make a difference here is that the Trump administration has agreed to, you know, expedite and quickly deliver all the necessary permits to Nvidia to make this happen. And I think a bit of background would be helpful here.

Um, in particular, Blackwell production is taking place at TSMC facilities in Arizona, so that's already one foot in the ground. And so that I think is clearly gonna indicate that this really has momentum and that this is something that's real, including, I would say the objective of creating up to hundreds of thousands of jobs. And I'll come back to that in a moment.

But what I think is also important here is that ai, super computers will be built in Texas through partnerships with Foxconn and Wist. And, uh, we're looking at roughly a 1 million square feet of manufacturing space that Nvidia is building out. And so that is going to enable mass production of supercomputers in the next year or so.

Also, over the next few years, they are planning to ramp up that half to that half a trillion mark with testing and packaging coming from SPIL or spill and Amcor technology. Now it's clear that I think that the company sees a world where diversity, supply chain and manufacturing is really critical for resilience and fulfilling the needs of its customers. Now, NVIDIA's hardly unique in this regard, but they really are, you know, putting this into action and really demonstrating it with, you know, these remarkable figures.

Also of note, Nvidia and a MD are ready to begin production of their most advanced chips at t SMCs plant in Arizona. So it's not just Nvidia, it's also a MD getting into the game. And from my view, this is a significant milestone and de-risking the single point of failure that is tcs, advanced node manufacturing and Taiwan.

And this de uh, definitely deserves a spotlight because it's not even so much the fact that, okay, mainland China could invade Taiwan, it could simply blockade Taiwan and have these same ma massive risks built into the supply chain. Also, natural disaster disasters as Taiwan is in the Pacific Rim. You know, it could be earthquake, it could be hurricanes, and not that, you know, Texas is completely free of that, but Northern Texas and, uh, Arizona certainly are safer in those regards and there's certainly no blockade issues.

And so what this is building to is that, you know, the critical steal suspicious gaps are getting into place for a completely regionalized US semiconductor supply chain. And basically the key elements are coming into place by Onshoring. And so for all the noise that's surrounding this, I would say the Trump administration's actions will accelerate the deglobalization trend.

And this is something that's not unique to the us but again, it's being acted on. And for more evidence, we're seeing that South Korea is putting another $23 billion into its semiconductor industry for basically achieving the same ends. And on a, I would say, a concluding note, what I think is also curious here about the half trillion figure is the sa same amount as, you know, the heavily promoted Stargate project that was promoted heavily again, uh, at the beginning of this year in January with a, the initial group of funders including SoftBank Open ai, Oracle, and MGX, as well as, again, Nvidia and Arm playing a hand as technology, um, partners.

And so there's the beginning of the build out of data centers in North Texas with that. But what they both have in common, in addition to the half trillion figure, is that details are a bit scant. I mean, we're seeing, you know, the basic details, the framework, but again, it's gonna be about filling in more information executing and so forth to really make this happen.

And thus, you know, take off and transform really the global semiconductor supply chain. Now on a different note, remember Windows 11 recall, does anybody recall it? Well, it's back and Windows Recall is really the ultimate user activity because it records tools released with copilot plus PCs back in May of 2024, a year ago now, and withdrawn a few months later.

So what Microsoft uses recall activity to do is record, uh, comprehensive search, for example, when you cannot remember the website that you are at or read a report or which Slack channel, which I think would be very useful, someone asks you to do something on. Basically, uh, what caused the shutdown is that many users were concerned about the amount of data that was being collected and about how recall was storing potentially confidential information. Now recall is coming back as Microsoft really made it more acceptable.

Alistair, uh, from your perspective, this comeback sequel of my Windows 11 recall a real deal. Yeah, I think there's an, it's an interesting thing that's a lot of the coverage has been that this really was not something users wanted to have and that they're not very keen to see it come back. The coverage we've seen, I particularly picked this up on RS Technica, uh, they, they don't expect people to want this again.

Uh, one of the challenges in this is that Windows Recall is looking at everything that's going on on the screen and doing AI analysis to identify if there's an image on screen, what's in that image, uh, doing text recognition within images. And so there seems to be very little opportunity to tell recall not to capture some piece of information. So particularly personally identifiable information, health information security information.

Uh, we might previously have been relatively safe if we emailed somebody a screenshot of some information because the, uh, image that was attached to the email wasn't actually being indexed successfully. What recall does is it will actually index the contents of that image. So if you are sending a, maybe a, a screenshot of a report, uh, from your doctor to your partner, and your partner is using a co-pilot plus, um, PC with recall enabled, that information is actually fully being ingested into recall, leading to a question then of how is this data being stored and secured?

And that's a kind of a, a sticky element here of, uh, there's, there was no awareness and recall that some of the information is more privileged than other information. The, the difficulty with recall was that even, um, opening a, an object, uh, opening a file from a file share recall would capture what was being shown inside that file with no awareness that there might be security constraints or compliance constraints around the data that's being shown. So there are some serious concerns about how recall is, is holding this information.

I can see huge benefit, and I particularly like the idea that recall could tell me which Slack channel I had received a message in because I, like many people, uh, Ron and I have hundreds of Slack channels across multiple, uh, multiple Slack servers to try and track and working out who said, what do you wear can be a real challenge. So I can see why Microsoft wants this now. It is coming back in the windows inside a channel first.

So it is an opt-in and the Insider channel, it's not in the main channel yet, and it's unclear whether that's, it's going to remain as an opt-in to have all of your data harvested by Windows Recall. And we do have some concerns about where that data's landing up, uh, where the AI insights are landing up from it. And quite a few answer, uh, questions still to be answered there.

Nu Con and Cloud Native Con in Europe 2025 were a couple of weeks ago now, but Solar IO added support for the MCP, the model context protocol to its open source K gateway. Uh, this helps connect AI platforms more easily. They also donated K Gateway to the Cloud Native Computing Foundation to boost the community use in collaboration.

Ron MCP seems to be one of the new hot things about connecting AI tools and particularly to connecting them into the Kubernetes clusters they're deployed onto. This seems really good. I agree, Alistair, it's a hot topic and I think a bit of background will be helpful here because MCP originally was developed by Anthro anthropic, as you noted, and it provides what I think is a universal protocol for linking large language models and AI agents to external systems.

And I think this is, uh, going to be successful because by embedding MCP support and K Gateway SOLO enables AI platforms to use a consistent, really a standardized method for access tools. And this reduces the need for custom integration for each tool and service. And well, quite simply, you can understand the benefits right away.

It's intuitive and built in. Now the K gateway is built on Kubernetes and Envoy proxy and it's designed for cloud native applications. So, alrighty, you could see why it will be so appealing to well developers and anybody who has to really, you know, rely on Kubernetes to get their jobs done.

And I think it's also important to note here that solo's MCP support can allow AI platforms to scale more efficiently as well as handle, you know, the growing AI workloads and traffic to, you know, the large language model providers while maintaining performance and security. So that's, I think, all important. All these things have to come together.

It's not just a scaling challenge, it's also a performance challenge and an all important security challenge. And I, I think by using K gateway's, MCP gateway, uh, AI platforms can connect to a broader ecosystem of tools and, you know, make it easier to build context to where interoperable AI applications without the burden of managing complex integrations. So what I think is happening is that MCP, you know, model context protocols trending as a magic sauce of sorts, that is key to transforming how enterprise developers and AI engineers can better use the AI ecosystem as such, solo support here helps move the needle.

And we're seeing folks out there like Selectors, John, uh, capa Bianco really just raving about how great MCP is and just making, you know, not only his job easier, but also the customers out there, the partners out there, anybody who just wants to make using AI a great deal, simpler and streamlined in their every everyday tasks. Now also, I think what's in interesting out there is what's going on on the server side. And what we saw is that Dell has launched new Power Edge servers, powered by Intel Zon six chips, which are offering better performance and energy efficiency over previous models.

Now the servers are built to support fla uh, flexible application development, and Dell also improves storage data protection and management tools. Now, Dell's new powered service, specifically the R four 70, the R five 70, R six 70, and R seven 70 models say that five types real quick, all leverage. Again, those Intel Zion six processors with both P cores and ecos, and that's important, deliver these improvements, uh, that are noted, uh, particularly for example, in AI data center environments.

And I think one example is that the servers are designed not only for, uh, these diverse tasks that are going on the AI data centers, but also scale out databases with the RH seven 70, specifically achieving up to 80% space savings per 42 unit rack by consolidating legacy PO platforms. So I think that's something that's going to catch the eye of some decision makers out there. And Alistair, you know, from your viewpoint, what is going on with these new Dell powered server capabilities?

How much is it a big deal? Well, these, particularly the, the R four 70 through seven 70 are the one and two U single and double socket servers. So these are the commodity servers that are deployed in large numbers and data centers to be the primary compute platform for a lot of organizations historically, that would, of course always have been their, their vSphere clusters that are then running all of their applications.

And although these are not the most high density servers with the most, uh, GPUs you can load up in them, there's one and two used chasis don't take a lot of gpu. What we are seeing is that the inference stage of the use of AI often doesn't need GPU or doesn't need very much GPU. And so even though these, you might look at as there's a CPU and, uh, and network heavy and not necessarily great for GPO, they can be a very cost effective way to run the inference stage of, uh, and let's be clear, the inference is where you're getting business value out of using ai.

All of the training stuff is the necessary pre-work to be able to do the inference, and it's the inference that you're gonna do at the largest scale over time. So having nice high density servers that are going to be cost effective to deliver a large amount of compute is a, a really vital thing here. I like seeing that there are more cores per, uh, processor available.

It's the general march that we have seen over time, and I like that this is oriented towards this combination of doing virtual machines, bare metal compute, where it suits containers on bare metal or virtual machines where it suits. I think also that, uh, Dell has updated object scale along the way. And so Object Scale provides a storage platform that integrates really nicely into a Kubernetes cluster.

And the newer release of Object Scale has gone much higher density on the actual SSDs up to 122 terabytes for a single SSD means that even though these aren't necessarily huge physical systems, there's a huge amount of storage in them, that is, to me, it's a clear line towards the use of that large amount of storage and high performance storage for AI use cases where that speed of retrieval of data, particularly for, for tools like, uh, the, uh, using a retrieval augmented generation, you've got large corpus of private data that you're using during the inference stage. So object Scale looks to be doing some great things at at increasing the amount of volume, the amount of data up to six petabytes of data in a single two you node configuration. And naturally you wanna scale out to more than just one node for that.

So yeah, nice to see that continuing innovation, that expansion and capacity and high density servers, particularly because we're seeing data centers filling up again, right? There was a phase when things emptied out in the data center when virtualization saved us, and we didn't need quite so much data center space. And here progressively we've seen now there's, uh, an expansion in demand for space and an ex an expansion in demand for power, uh, which, uh, can be quite challenging.

Of course, I'd, um, Dell also brought in a new version of their open managed software supporting newer versions of the irac. So the IRAC 10 is the latest version, and I'm getting old when I remember the versions of irac I worked with, uh, many, many years ago. And perk 13 RAID controllers.

Yeah, lots of, uh, new technologies coming in, these newer servers, seeing reductions in right latency, seeing IM improvements in the proactive support, uh, line US know beforehand if things are gonna fail, scaling to ridiculous amounts of storage in all flash configurations, using things like Dell de Power Protect for that onsite data protection. So seeing really good continuation of that spectrum of more power, more storage, uh, at least more compute power, hopefully not, uh, an equivalent increase in the electrical power requirement. Uh, and that again, is what we see with the, uh, innovations and, and CPUs along the way.

So awesome stuff from Dell. Lots of new products released out and, uh, continuing to see more use cases for them. Google Cloud next Google's big conference was last week, and they introduced a bunch of tools and technologies to make AI adoption easier.

That might be a bit of a theme today, is making AI adoption easier, uh, and more efficient highlights include Google's, uh, agent to agent protocol for secure collaboration between AI agents. It's not just machine to machine communication, it's actually AI agents interacting with other AI agents, uh, and new cloud tools to better manage these applications and services. Google also unveiled Ironwood, a powerful new AI chip, uh, reinforce, uh, reinforcing its focus on leading AI infrastructure and AI tooling.

Ron, you have, uh, spent quite a bit of time looking at Google's announcements and Google's ai, what strikes you as the best parts of this? Yeah, uh, great lead at Alistair, and I think, yes, it's certainly been spotlighted, you know, uh, what's been coming out of Google Cloud next, but also I think, uh, Google Cloud overall has been building a lot of momentum over the last few quarters. And that's why these announcements are, I think, sharp, you know, differentiators for Google.

And I'll start with the, uh, a to a agents, uh, first of all, you know, what they are going to enable is that AI agents, you know, whether they're built by Google or Salesforce, SAP, you know, others, what it's gonna happen is that they could communicate securely using a common pro protocol based on HTTP, uh, Jason, RPC or SSE. So what this is, uh, enabling is the reduction and need again for custom integration. And so I think this is something that in this context, in this area, is, you know, moving toward, um, just that having more standardization and less customization, which could be a bit confusing because in other areas, more customization is a demand.

Uh, but, uh, for this particular announcement, this I think is a outstanding, uh, feature that will flow. And it's also allowing agents to share data and coordinate actions. So for example, a client agent can discover remote agent's capabilities through J based agent cards and assign tasks such as processing a customer query or scheduling logistics, et cetera.

So you can see, you know, some of the baseline appeal here already. Moreover, it's backed by over 50 partners, including Atlassian, cohere, MongoDB, as well as Deloitte and Accenture. So you really have a who's who of major players that can really make, uh, the a to a agents, you know, matter in terms of what's going on, uh, with, you know, the Google, uh, customers out there.

And, you know, as a result, it's, again, signaling strong industry buy-in. And from my view, then this coalition suggests a move towards standardized AI agents and ai, uh, agent ecosystems, and again, reducing the fragmentation that's out there. And also while a to a focuses on enabling communication between AI agents model con, uh, context protocol, which we touched on, is facilitating connection between agents and external tools or data sources.

And so what I am hearing is that they both together can, you know, create a robust framework for multi-agent systems by addressing different layers of the interactions. And this can again, be underpinned by the Ironwood, um, chip set that was also, uh, unleashed at the event. And I think that's important because we're seeing the, uh, again, that the hyperscalers are committed to building their own silicon.

This is, again, trending toward more customization to, you know, again, support their specific cloud workload demands that is really primarily related to ai. And so I think that's something that is just reinforcing, uh, that trend is again, making the competitive mix out there more interesting. And it will make, uh, I would say more challenging for, you know, some of, uh, the merchants out there to compete more directly against it.

And then we're seeing, you know, players like SoftBank already hedging for that, uh, particular outcome. And, uh, so this is something that I think is going to maintain, uh, a good deal of momentum for Google Cloud because my, uh, comment and, you know, to round this out about Google Cloud having a lot of momentum, I'm seeing, uh, you know, organizations, enterprises out there adopting Google Cloud from a previous platform because of Google Clouds, uh, excellence in areas such as resilience. That is, you know, when you have, you know, virtual machinery boots, they're gonna have to happen on occasion rather than having them happen frequently.

And with little to no pre, uh, announcement that upsets customers what Google Cloud is doing. It's not only are they reducing those types of scenarios, but also proactively letting them know, Hey, yes, a virtual machine reboots are gonna have to happen, but just to let you know so you can plan around it, and this is something that, you know, should only take X amount of time. That's a real difference maker.

Something as simple as that, or seemingly simple, but it's not really simple because it's not obviously being implemented across the entire cloud ecosystem. And the other important factor is security. And of course, cloud security is always important.

It's not so much the native security of the hyperscaler IE Google Cloud, but it's also ensuring security of third parties and partners that are using the Google Cloud platform. And that has obviously created some situations that are unfortunate for, you know, other scenarios out there. But Google Cloud, I think is doing a good job of directly addressing that, and as a result, creating, uh, a competitive advantages.

And, uh, speaking of what's going on that can impact, you know, data centers, we saw that Intel and Alterra are basically partying ways. 75 billion. And as an outcome, Intel will retain 49% share, and the deal is expected to close in the second half of 25.

Now, this move is part of CEO lip bhutan's strategy to streamline operations and strengthen Intel's financial position. And I think what is also important here is that Rahi Hussein from Marvell will take over as a al alter's new CEO on May 5th succeeding Sandra Rivera. 75 billion compared to this nearly $17 billion until paid in 2015, basically a decade ago.

But the sale, I think, will provide the company much needed cash after hefty bets on contract manufacturing, uh, previously did not really pan out and as a result, strained Intel's finances. And what I think is also important to note here is that, you know, this is the center of tan strategy to streamline intel after, you know, the attempts to diversify did not really happen beyond, you know, Intel's mainstay PC and server chip businesses, AI being the prime example. But I think it's also, um, uh, indicative that, uh, while Intel's struggling to gain a foothold in the AI industry that is dominated by Nvidia, uh, that what we're seeing is, again, Intel making these important steps to really sharpen its focus, let you know Tan take care of, you know, more of the financial side.

And, you know, the product teams really drill down on making innovation happen. So again, I think it's an upbeat note that Raib Hussein is taking over as CEO of T Alterra because after 25 year journey at Cavium and Marvell, which he co-founded, you know, with, uh, SAE Ali Cavium in 2000, and he proceeded to build CM from scratch to about a $1 billion annual revenue company and over a $6 billion market cap at its exit. And that included a successful IPO in 2007, culminating with its acquisition by Marvell in 2018.

So in a nutshell, I think Alterra will be in very good hands moving forward. And Alistair, from your perspective, you know, what do you think is important about, you know, Intel selling, you know, 51% of its share in Alterra? What are your expectations?

Well, there've been some discussion about Intel completely selling off the business, and that seems to have been a plan a little while ago, but lip Bhutan seems to want to hold onto it, and I can understand that having a part of your, your platform where these field programmable gator arrays are, are available, this is a, a piece of physical hardware chip that can be reprogrammed to function differently based on the firmware that you load onto it. So unlike a normal CPU where the instruction set is, is fixed when it comes outta the foundry, uh, the field programmable gateway can be rewritten for a different task, E even to the extent that, uh, you can offload tasks onto the field programmable gateway, the FPGA, such as accelerating databases. So this capability very much complement the use of high performance CPUs as we get from, from Intel.

And so not completely handing over Alterra holding just, uh, the the largest minority stake holding you can have is a, a good move to keep access to that technology. But as you say, Intel needs to recover some, some, uh, finances to focus on the core things that has made in Intel. Great.

75 billion, that's gonna be very beneficial as funding further development and the things that, that Intel specializes in. And I think we will see plenty more focus from lip Bhutan to do things in a different way and to get intel back on its its feet. And that brings us to a closer look.

We did want to look a little more at what LIP Bhutan emphasized at the Intel Vision Conference, uh, a move towards a customer-centric, uh, strategy focusing on advancements in a I PCs with GDI three accelerators, uh, and a, a new, uh, node pro process node for, for, uh, higher density. He introduced the idea also of physical AI highlighting Intel's commitment to integrating AI into hardware solutions across various selectors, Tan's leadership aims to position intel at the forefront of AI driven innovation by enhancing its portfolio and foundry capabilities. What we see is these physical AI is talking about what I'd think of as embedded ai, AI inside physical things, whether it's your car, whether it is your fridge, whether it is a manufacturing plant, but instead of having AI purely in data centers and particularly cloud data centers, actually having the AI happening inside physical devices.

Now, we saw this progression with predictive AI in the past, uh, elements like cameras that get object recogni recognition built into the physical camera rather than relying on a service somewhere to do object recognition. And that's a, a very important part of what we see with things like, uh, fraud detection, physical fraud detection in supermarkets, uh, this intelligence built into the camera is what we mean by physical AI here. So Ron, you've written about this for the TUM Group.

Can you give us a little more insight about what you saw from Intel Vision 2025? Yes, thank you, Alistair. And it was an honor to be able to attend in television because really it was lip Bhutan's public debut as CEO.

He'd only been on the job for two weeks, and he, I believe, was very poignant in explaining why take on this massive task. You know, you've been so successful throughout your career, you know, you could be selling this seven Cs and sipping sling, uh, Singapore slings for all you, uh, for all you want. And yet here he is, uh, again, looking at really, um, uh, something that is an immense challenge that is getting intel back on its feet.

And his, uh, core response is that he knew he could do something. He just couldn't stand by and just, you know, let Intel continue spinning its wheel. And so he's going to find out what it's going to take to really make intel, you know, a, uh, successful company, again, really a more formidable competitor.

And we touched on this throughout our conversation here, certainly n video looms, but it's also about a and d it's also about arm. It's also about more customized chips. You get it.

There's just a, a, a really and a huge array of challenges out there that Intel is taking on. And so, uh, while, uh, he talked about really, you know, all the core, uh, areas of AI including, uh, ag agentic ai, and certainly generative ai, physical AI was, you know, part of that, uh, tripod, I guess you can say of know the major AI areas that Intel can look at. And so what I think is interesting here is that during the event, uh, he invokes I would say, robotics on, uh, at least a couple occasions.

You know, this is my guesstimate, but I think it's indicative that, okay, the thinking is the physical AI is really the new kid on the block here when it comes to AI areas. Yes, Nvidia has done some heavy marketing here, and yes, you know, world foundational models are something I think Intel should definitely play a role in. But then it's translating that and productizing into really, uh, silicon and software.

Again, he's emphasizing software as a major part of their strategy to, you know, again, uh, deliver competitive advantages quite simply to customers. And because physical ai, again, is nascent, I believe it's an area where Intel can really make its mark, uh, all the more effectively and stay the areas where they haven't quite really made a mark. You know, gene AI coming to mind.

I mean, yes, the part of the ecosystem, yes, uh, you know, Gotti, three accelerators and, you know, Xons, DPU, uh, fabrics and so forth can, you know, make a difference when it comes to, you know, some enterprise training and inferencing. But I think it's gonna really to capture the imagination of, you know, investors out there. And the overall ecosystem, I would say, uh, major play in physical AI can be one key way of doing it.

So, uh, if anything else, he was consistent about emphasizing AI's transformative, uh, potential across industries. And it's again, uh, uh, being able to leverage the advanced computing power and infrastructure along with software that Intel can deliver. And so I think what is going to happen is that Intel is really not starting from scratch and that they possess, you know, the key portfolio building blocks to pull this off.

And I touched on these AI accelerators, edge computing expertise, and really a tradition of powering complex systems to become an integral player in the physical AI market during its earlier stages. And I think, yes, I touched on robotics, but other key, uh, use cases, uh, invokes, uh, a couple of them, Alistair certainly, you know, video, uh, capabilities and, you know, um, warehouses and so on. I mean, there's just, you know, a litany of, you know, uh, different, uh, use cases that could be capitalized on here, but it's really about Intel prioritizing this potentially and really getting the ecosystem collaboration that's gonna be essential for this.

And so that was something that I thought was, you know, a, a couple of major takeaways from Intel vision. It's, you know, lip Bhutan setting the tone and you know, really saying all the right things in its first two weeks at the helm, but also, you know, so hence suggestions. And I think, um, from my view, and definitely lead Intel to a more successful path.

And so that in a nutshell is, you know, what I thought was valuable about being able to join Intel vision this year. I think it's, it's this physical AI is definitely, as you say, it's an cent area. It's an area where there is not a dominant player that, that owns the market.

And that you, you have such an uphill battle against as, as Intel has struggled with their uphill battle against Nvidia in the, uh, in the GPU market. So, I mean, GDI is an accelerator. We've, uh, through the, uh, labs team here at, at Futurum, we've done quite a lot of work around seeing how actually effective the, the Gaia accelerators are.

And, and yet they're not a headline, they're not a, a household name the way NVIDIA is for, uh, AI accelerators. So claiming the, the end space, the other thing to, to bear in mind is this, this end space, the physical AI is all about inference. This is where the value's being delivered.

This is the low risk place to deploy out these new pieces of infrastructure because this is directly giving you business value. The more foundational staff building foundational models, you, you're building it and hoping they'll come. There's a significantly higher risk as you are building out that, that earlier stage in it.

So I think there's some some really good things to be seen here. And I think, uh, it was very fast movement for lip Botani clearly had plans about how this, uh, Intel could be brought forward in a different way than it has for the last few years. And clearly he made this, this visible to the board and, uh, has then started straight away to get the message out and to start to execute on that.

Yeah, I think physical AI is the next frontier for, you know, the next major study breakthroughs of innovation because we've, you know, been witness to, you know, the personal computer revolution, you know, bringing the data to you. You don't have to use an IBM mainframe to do all this compute processing again, smartphones, you know, bringing the telecommunications to, you don't have to use an at and t dial tone phone to communicate, et cetera. I think we all, uh, understand that and have been fortunate to witness this.

And I think, yes, this is the next new brave new frontier amongst others. But again, it's about robotics and I think robotics will increasingly become important, not just in manufacturing, but also in, you know, say households. You know, getting beyond, you know, basic vacuuming capabilities to, you know, capabilities that could clean a pool or clean a house.

You get the whole idea. These are things that are, I think, increasingly just around the corner. So, uh, good news, uh, for everybody.

And yes, I, I think that was an extremely good point. You touched on Alistair, it's about the edge. It's, this is where Intel I think has very formidable assets in terms of, you know, reducing latency, being able to optimize, uh, inferencing and other, you know, AI capabilities.

And this is not a game of, okay, having these GPU clusters, having lifting, doing AI training in massive data centers. That's important, that'll continue. But what I think what is evolving here, again, is about what's going on in the field.

And this is where I think Intel can make a dramatic difference, uh, because it's not just a new field, but it's a new set of requirements and so forth. So stay tuned. Alright, well that brings us to the end of our rundown of, uh, the news of this week.

Of course, we're looking forward to future weeks, uh, next week of course we will have another rundown, but I won't be there because I'm gonna be extremely busy. Uh, so April 22nd to 25th is AI Infrastructure Field Day. And then Ron, you are joining a Tech Field Day event at the beginning of May.

I believe you're joining us for Mobility Field Day, Right on, definitely looking for it. Mobility Day 13. And it also has a huge lineup, and I wanna leave anybody out, but it's a who's who of certainly, uh, wireless players.

Uh, I think we're gonna get some excellent scoops and takeaways on what's going on, you know, with, uh, not just wifi, but also 5G and the overall mobility space. And of course, Tom Hollingsworth will be the event lead there, and you can catch all of that action on May 7th and eighth. And, uh, says, uh, Ron says that's gonna be a big event too, and Tom doesn't escape without another event.

Also in May, security Field Day will be May 29th to 30th as well. Uh, a really big couple of months with, uh, great events. So thanks for watching the Tech Field Day rundown.

You can catch new episodes every Wednesday, uh, as a YouTube video or on your favorite podcast application, the rundown streamed on Techstrong tv, uh, and you can often catch us on other tech Strong and future and group platforms. We'll be back next Wednesday to talk about the IT news and the week. That was until then for myself, for one wistful and for all of us here at Tech Field Day, here's wishing you and yours a beautiful day.