NVIDIA’s Multi-Billion-Dollar Moves to Expand AI | Tech Field Day News Rundown: September 24, 2025

Hugging face open source models for GitHub, copilot and Lummi AI agents that automate infrastructure management. Cisco net DevNet is now Cisco Automation and portable ai. Micro data centers are in oil fields.

The Department of Justice is going to break up Google's ad business and Sumo Logic has AI agents to streamline cybersecurity. Finally, we'll take a closer look at the billions that NVIDIA is spending on AI partnerships. Welcome to the Tech Field Day rundown, where each time we meet, we run down the IT news of the week with variable degrees of snarkiness.

I'm your host Alistair Cook, and joining me on this beautiful day is my co-host Scott. Rob. Scott, welcome to the show.

Thank you, Alistair. Great to be with you. It's great to have you here as well.

And of course, Tom Hollingsworth this week is at Security Field Day. When you're finished listening to the podcast, maybe tune in to some of the broadcasts there, particularly also on Dogs in Politics Day, I'm not quite sure the role that we want for the dogs in politics, although of course some could be better than our current politicians. And today is also International Sign Language Day, and unfortunately, I can speak neither Americans sign language, Northern New Zealand sign language.

So it's probably a good day for us to head straight into our first item of news Hugging faces. 5 directly in GitHub copilot chat. Uh, within VS k, which I've regularly use as my development environment, users can easily switch between models and providers for a single interface, enabling testing of specialized or language specific coding tools, uh, without leaving their editor.

There's a free tier for experimentation and paid options for when you need commercial support. Uh, the integration gives developers flexibility and control over AI assisted coding workflows. We've had a little bit of cynicism on past episodes of the rundown around the performance, the actual usefulness of these AI coding workflows.

Scott, have you had some experience? What are your thoughts on AI coding workflows? So, I'm not a coder per se, but I don't think it's hard to see how optionality helps here, right?

And competition amongst models and being able to try additional models for different tasks very easily. Um, I think there's two really interesting pieces of this interaction. Um, you know, GitHub has a huge presence, let's say, a lock on the market for source code, uh, source code, um, control and management.

Um, of course there's GitLab, of course there are other tools out there, but GitHub in their social approach to attracting users really dominates this field, right? And so for GitHub to be able to let you change models in for coding assistance, that's a huge impact, right? To let people try things, very different models very easily.

Um, I have a son-in-law who works for GitHub, and I know he exercises this, um, very, very frequently. Um, take that and combined with, you know, what Hugging Face is doing for making open source models and model variants very, uh, easily accessible. I think they just hit over 2 million models available a couple weeks ago.

Um, you know, all the oxygen in the room is, seems to be taken up by open ai, by Anthropic, by Grok, um, by Google Gemini. Um, but there are other choices out there and other, other things to play with. Um, this just gives coders more access to more models and more choice.

I'm not saying all of them are are high quality, but, uh, let the, you know, caveat, let the buyer beware and let people, you know, be able to experiment and try, try their own thing. Uh, pluming introduces AI agents to automate infrastructure management. PMI has launched a preview of pmi, Neo AI agents integrated into its infrastructure as code platform that can diagnose issues, suggest optimizations and autonomously manage IT infrastructure when permitted.

Interesting point on when permitted there. Um, NEO acts like a digital member of the DevOps team handling tasks, monitoring outcomes, and maintaining compliance. While keeping a full history of automated actions, teams can choose whether to use NEO for guidance or to let it execute tasks independently based unassigned permissions.

Now, Alistair, this is firmly in the neighborhood of having, um, becoming an agent boss and now having these digital coworkers that can do things. What do you think about this piece and what you're seeing in this space? Well, I, I think we're seeing some challenges around AI agents not necessarily following the rules they're set up to to follow.

And so PA's challenge here is making sure that NIO does follow the instructions, the boundaries that are given to neo. So having controls about whether Neo is able to correct make changes, maybe make firewall changes, as well as configuration changes in its automated form. Uh, the objective of course is, is to allow this, is to have this AI agent that does the ongoing work.

That is the all seeing eye that we really do want in our infrastructure management, uh, particularly in our fast changing infrastructure management, which is really, we're pulling new targets as DevOps teams. Um, their expectation is that Neo will behave as a new neo, uh, DevOps team member. Yeah, the challenge will be as we kind of move that transition from, you can make recommendations, please, Neo, and I'll choose whether I implement them or not towards the situation of, yes, Neo implements these changes automatically.

Lummi is placing this as being a way of making sure that we have DevOps best practices, and particularly DevSecOps best practices being applied in our pipelines. And that absolutely is a great thing thing that the easier it is to comply with best practices, uh, for both DevOps performance and security. Uh, but the easier we make this for developers, the better, uh, developers care about functional features within their applications rather than wanting to know about all of these lower level compliance and infrastructure elements.

So this could be absolutely awesome. I think, uh, this is still at a preview for plumes Neo, and the proof as always will be in the PUD and the eating of the pudding. It's do we actually get good agents?

Can we trust these agents? And then I liked that there was a whole lot of, uh, audit logging with Plum neo, uh, all of the actions that are being taken are being logged through, so you at least have a, a retrospective view of what's changed. And some of that around change management.

Conflict management is gonna be really vital as we use more of these agents. Cisco has renamed its DevNet certifications to CCNA, CCNP and CCIE automation to reflect the growing importance of network automation skills. Uh, new names make it easier for employers to recognize expertise and improve career opportunities and boost potential earnings for certified individuals.

Existing certifications remain valid, and these programs, uh, prepare professionals for roles from junior automation engineers to engineers to senior network architects. Uh, I would've hoped that network automation had been central to the workflows for most large organizations for, well, I would've hoped more like 10 or 15 years. Scott, are you seeing that in the real world?

So we have a whole organization dedicated to understanding why network automation has not been adopted more fully than it already has, right? Network automation forum. And, you know, there have been some interesting dynamics that have held back, um, companies going all in on automation.

Some of them are cultural issues, some of them are lack of leadership issues. Um, a lot of it actually centers around the focus on the CLI as the primary way to get network devices to do what you want them to do. Um, and on one hand, um, you know, I've been involved in some really interesting conversations on how the CCIE program and the certifications that lead up to it may have perpetuated, you know, this resistance against automation.

Um, that being said, I'll, I'll leave it to the listener to make their own conclusions on that. I think this is a really good move f uh, for the, the DevNet program to be relabeled, you know, CCE XX Automation. Um, not only that, but I do think this is an, an interesting move that takes this particular cert, uh, program away from being strictly vendor specific.

I've had some really good conversations with people who are involved in, you know, developing and administering the program. And, uh, there's so much of what can be done here that's not just Cisco kit. Um, and I won't, I won't, uh, drop all the names that we could for network infrastructure here, but this may be the most applicable vendor cert program across an industry, and not just for being a CLI jockey for one particular router and switch vendor.

North Dakota is testing portable micro data centers for AI in oil fields. Bachan Energy and Startup Armada plan to deploy portable micro data centers across North Dakota's oil fields using container sized units. That's physical containers, not software containers that can be quickly set up, scaled into larger clusters and powered by local energy like flared gas.

These data centers will process data close to the source supporting AI driven oil field analytics, while avoiding the cost and complexity full data centers with the growing demand for edge computing. The project could make North Dakota a test site for containerized AI infrastructure through the first pilot sites, though the first pilot sites are still pending. So, Alistair, what do you think tiny little micro, uh, data centers spread all over North Dakota?

What will that look like? Well, I think it's interesting that it's, it's a micro data center that is a 20 or 40 foot shipping container full of compute and presumably some environmental control. And I think this is a good move because this sort of standardized packaged data center that you can drop into a site that really does need quite a lot of compute, uh, it's gonna be a good mechanism.

Particularly edge computing is, is all about having infrastructure that you don't actually manage on site. You manage remotely and you manage as a fleet and this kind of packaged ship it to site plug in power and, and networking, and you're good to go. I like also that they talk about using starlink for the uplink rather than having to get a, a fiber circuit run out into some oil well in the middle of nowhere.

North Dakota. Uh, to be clear there, none of these have actually been deployed yet. And also data centers in a shipping container are not a new idea.

These micro data centers as an idea have been around for a while, and the modularity building up a larger data center with a collection of these shipping containers was definitely one of the ideas that was being explored probably 15, 20 years ago as new infrastructure for, uh, hyper uh, scale clouds. Uh, I think AI at the edge is definitely driving the need for more compute at these locations. And, uh, oil extraction, this is not an oil exploration, this is oil extraction is increasingly being driven by ai, ai and the bandwidth out to these, these locations doesn't allow that AI to run remotely.

So this is a pretty cool thing. I think we will see more of this. One of the interesting things to see is whether some of the hyperscalers will start going down this road again and building smaller clusters of these, uh, shipping containers because they've had some challenges where they're building these mega data centers are that the, the local communities are up and arms against them.

We covered this I think last week in the rundown, the use of some interesting planning permissioning processes for these larger data center types. Maybe smaller data centers that are gonna have less infrastructure around them will help with the community acceptance. Sumo Logic has launched Dojo ai, a set of AI agents to help cybersecurity teams quickly analyze and summarize security data using AWS's NOVA models and the mobile chat interface agents, reduce the manual work and highlight important alerts and speed up investigations.

Of course, we know the attacks are increasingly being AI driven and AI generated attacks. And these tools, these AI tools from Sumo Logic are gonna help defenders keep up and particularly letting them, uh, respond more rapidly and with a bit more clarity to what's going on in these larger attacks without requiring that massive scale outta humanity. Uh, once again, we see AI agents got looking at more security.

Yeah, there are several reasons. This is very interesting and exciting to me. Um, you know, I think primarily as I've watched, you know, the Knock network operations center and SOC Security Operations Center motions evolve over the past 30 plus years.

I feel odd even saying that number, but it is what it is. Um, we've really seen the security side adopt more automation more quickly and also move to the use of AI for their tooling for helping fight the battle. Um, so there's been a, you know, a, a willingness to not make the humans keep fighting the robots, but actually using some robots to fight the robots here.

Um, also using AI and this reduction of log and alerting information, um, from security systems just makes complete sense, right? That's something that large language models are very, very good at. Um, in my view, the security community through, um, actions like this are leading by example.

I would love to see NOx, um, take more of a similar approach to this, you know, reduc reducing network telemetry, um, and other information. Um, I also see this as hope for the future where I could have a unified view of all my logging and reporting from my security systems and from my network and the IP layer and my optical layer. And if I had this unified field theory that somebody like Sumo Logic could drive forward, boy, I could have this unified view of what's going on with security in the network.

And if I could get Datadog to feed into this, I could have all this information in one place to help me get to problem resolution very, very quickly. I know I'm, I'm pushing a little bit of the boundaries here, but I think there's hope for the future with looking at, um, AI tooling. Just like Sumo Logic has done here.

The US Justice Department has asked a judge to force Google to sell its largest ad exchange in key ad technologies following a ruling. That part that parts of its ad network are an illegal monopoly. The trial's current phase focuses on remedies to restore competition with prosecutors saying Google's practices harm publishers and the market.

Google costs the plan extreme and plans to appeal once any measures are decided. The case targets the core of Google's 305 billion annual advertising revenue. We know how important Google is in Google AdWords serving up advertising.

Alistair, what do you think about this? This is part of an ongoing Justice Department examination of Google. You'll recall we covered earlier that the Justice Department wanted Google to hand off maybe Chrome outta the business and just break up a little.

Now the target is the, the actual business side, because of course, Chrome was never what Google's about. Google has always been about that. Ad revenue, $305 billion is quite a good turnover for advertising.

Uh, and that's of course the money that's going to, uh, Google themselves and Google Ads is pro the biggest marketplace for buying and selling placement of ads. And that's really what the Department of Justices addressing here has. Google used that dominance in the market to exclude others, and how can we reduce that?

How can we restore competition? Now, the initial, uh, justice Department, uh, request comes off a, a judge's ruling that Google is anti-competitive in this market, and something needs to be done about that. The next stage is where the Justice Department asks a different judge to choose how to, or to make a judgment on how Google should be, uh, made to allow further competition.

Of course, Google eyes that $305 billion and says, you know, I kind of like having that income and I don't really want to have to hand it off to somebody else. And so naturally they say, this is a terrible thing. Uh, Google is doing a wonderful job of making access to the advertising revenue available to all of the sites that actually use Google Ads to place.

Uh, now this, this is the danger of success in a capitalist market. Uh, if you are very successful, you take over the market and then the market doesn't like you so much because you have some dominant position that you can use. It's the natural conflict between what's good for Google and Google's, uh, price Google shareholders, uh, versus what's good for the customers.

And it's, it's a never ending story. Uh, we've just seen this continuing. You know, we, we first saw these, I think with Microsoft and, uh, internet Explorer.

Uh, we'll continue to see these stories of, of, um, breaking up large organizations to restore competition and of course, Google Fight this Time. Look story we've seen over weeks, months, years, NVIDIA's dominance as the hardware accelerator for ai. We've seen Nvidia chips being the core of what's driving AI forward, and of course, AI is what's driving Nvidia to be such a massively valuable company.

Uh, just as the, uh, cryptocurrency bubble burst AI came in and saved Nvidia, Hmm, maybe NVIDIA's worried that, uh, AI isn't gonna last them forever. So a couple of moves that we've seen in here, one of them is Nvidia has teamed up with Intel, huh? Intel seems to be teaming up with a whole bunch of new people.

Maybe this is Bhutan's plan is to actually have in, uh, Intel not as the standalone leader, but as partnerships with other organizations. Yeah, this, this is a fascinating investment and, uh, you know, to see a competitor invest $5 billion at the drop of a hat, but a competitor that's in trouble, right? You know, they're working through, um, how they're gonna end up organizationally.

Um, of course there's been talk about investment in their foundries, especially in the US and how those are gonna survive. Some of our friends of in Ohio are probably very tuned into this, um, who may or may not be involved in the production of this, uh, this episode. Um, it's fascinating to watch.

Um, I, I do see the benefit of, um, making sure, you know, for us interests there is silicon production in the US and helping to survive. I think that this investment is probably gonna be key to Intel surviving in some way or form, whether they remain independent or whether they actually become part of an NVIDIA at some point. And the other commitment that NVIDIA's recently made is a little bit bigger.

So 5 billion to Intel, but a hundred billion dollars commitment to OpenAI, uh, and helping OpenAI build 10 gigawatts of, uh, data center power. Uh, this is a huge infrastructure partnership, which comes on the back of the infrastructure partnership we saw between Nvidia and Core Weave that we covered. And that was, um, in the 6 billion.

This a hundred billion with OpenAI is absolutely massive. I think, uh, one of the interesting things on this is that the money that OpenAI, uh, that Nvidia commits to open AI will probably come back to Nvidia because OpenAI is gonna be buying millions of GPUs with those hundred billion dollars. So, hmm, maybe NVIDIA's getting a stake in open AI by giving them GPUs.

And that's an interesting way to run it. I wonder how, if that has some impact on, uh, tariff costs, because of course, the current government's tariffs in the US are part of the driver for Nvidia wanting to team up with Intel and use those, uh, within the us uh, chip fabs rather than being hit with the import, uh, tariffs as they're coming in from overseas. Uh, a hundred billion for a long time with, uh, OpenAI.

Yeah, Well, I, I saw some interesting comparisons about, uh, you know, OpenAI and Nvidia, are they the new Microsoft and Intel, you know, in a sense. So we have a new operating system and a new computing platform. Uh, it's an interesting analogy.

Time will tell if, uh, if that really plays out. Um, it, you know, certainly those geopolitics and tariff, uh, implications are a factor here. Um, I'm not an economist and I don't play one on television, but, uh, you don't have to go too, too deep in your economics textbooks to say, yeah, this is probably a motivator here.

Um, but it's a very significant placing of a bet and investment. Um, and the, the return will come back to Nvidia just like, uh, just like you mentioned. Well, that's our news and our closer look for today.

But let's look a little bit further ahead. Well, to today, even Security Field Day is on right now. Tom Hollingsworth is out in California hosting a wonderful group of delegates, uh, an awesome lineup of sponsoring companies, and you can tune in to those, uh, videos absolutely right now.

Uh, it gets a week off before he is doing a Tech Field Day exclusive with Microsoft Security. That will be on October 9th. com website.

You'll also find that Tech Field Day will be at NetApp Insight. And that will be on, looks at my other screen. That will be on the 15th of December.

Uh, there are abouts, hang on, that's not December. Uh, you'll also find that Tech Field Day will be at NetApp Insight on October 15th and tune in for more video. From that, I get a trip back to the US the following week for Cloud Field Day.

Uh, I'll be running Cloud Field Day, October 22nd and 23rd. Uh, we've got HPE and we've got Fortinet and Oxide computer will be back with us as well. And then the following day, Stephen Foco will be out in California for AI Field Day, October 29th and 30th.

You can find all of the Tech Field Day events on the Tech Field Day website, and of course on the Tech Field Day LinkedIn, you can find the live streams or of course on the Techstrong tv. Thanks for watching the Tick Field Day rundown. You can catch new episodes every Wednesday as a YouTube video or on your favorite podcast application.

The rundown is streamed on Techstrong TV as well. And you can often catch us on other Techstrong and future and group programs, possibly even on the Tech Drunk Gang some days. Uh, we'll be back next Wednesday to talk about all of the other IT News of the week.

That was until then for myself, Alistair Cook and for Scott Reon, and for from all of us here at Tick Field A is wishing you and yours a great day.