White House AI Roadmap Pushes to Block State Rules | Tech Field Day News Rundown: March 25, 2026

SAP is all AI. Sunrise in space. FCC says no new Wi-Fi at home, and Cloudsmith goes real-time.

Zero CVs, Hugh Kubernetes, AI smuggling uncovered, and the White House wants national AI. Let's run this down. Welcome to the Tech Field Day Rundown, everybody, where every week we meet and we run down the IT news of the week with a variable degree of snarkiness.

I'm your host, Alistair Cooke. It is National, in fact, International Tolkien Reading Day. I shall be reading some Tolkien later today.

It may take all week. Joining me is my co-host, who will have some time on a plane to read some Tolkien. Guy, how are you doing today?

I'm doing great. Feeling extra snarky. Being stuck in an airline lounge is definitely a way to end up snarky.

Of course, when you get home, you can have some lobster Newburg to go with all of the beautiful stories that we're bringing you. We will start with a great story here. The way companies pay for their software is changing, and SAP is making a big shift to keep up with the AI era.

Led by CEO Christian Klein, the company is moving away from charging business users per user, and instead plans to charge based on what AI actually does, like completing finance tasks or managing supply chains. SAP is betting its huge amount of business data that it holds, and that'll help it compete with newer AI companies. The goal is to turn its traditional business software into AI systems can run large parts of the company automatically with much less human work.

This seems to be a bit of a trend of AI does things and we need less people, doesn't it, Guy? Well, yes. Now, SAP is very much worth watching in this space.

I think it's, what is it? About a year and four months, or is it two years and four months since Satya Nadella of Microsoft asserted that software and SaaS are dead. Thanks to AI, the whole idea was why write and build software when you can just have agents that you ask what to do, and the agents code everything for you in a trice, run it for you, and then destroy themselves or destroy the code.

So you really don't need software anymore. I doubt he really meant that totally literally, but this is probably the best example we have to date of this sort of thing. Why?

SAP is an extremely careful product development, software development, very customer-focused vendor, and they have a pretty large captive audience, mostly in Europe. So they tend to be very measured and careful and provide a ton of customer and client support, especially for big shifts like this. The last big shift they made was the development and release of HANA, which took them roughly 10 years to transition customers out of Oracle and into SAP databases.

Now, this is changing not just how SAP is delivered, but how it's priced. Critically important. So what they are essentially doing, as our colleague Mike Pizzardi commented to me on Friday, is creating a new licensing model where you're more or less hiring AI agents as virtual employees.

This might sound threatening. I actually think it's encouraging to see a software and pricing model that sort of fits the way AI and agentic AI is infusing itself everywhere. Somehow or other, the software companies have to get paid and be profitable for all of this development and tokenization and use of cloud and their own AI services and all that investment and everything.

And I think the SAP example is a really good one. They are really unlikely to be making a misstep here, so to speak, at least from a product standpoint. And because they always provide a whole lot of support, like I said, and have a very loyal client base, that makes this as an example for everyone to watch.

This could be the future of your enterprise software. The race to dominate AI is moving into space. Jeff Bezos' company, Blue Origin, is planning a huge network of satellites called Project Sunrise that would help run artificial intelligence from orbit instead of on Earth.

This sounds great. The idea is that you can use solar power in space and laser communications between satellites, and that'll reduce the massive electricity and land needed for data centers on the ground, which is so far where all of them are. The plan would be to put Blue Origin in direct competition with SpaceX, founded by Elon Musk, so Jeff Bezos against Elon Musk again in space.

While the project could be very expensive and does raise concerns about space debris, it shows how companies are starting to look to space as the future home for AI infrastructure. Alistair, I always think of you as someone in space. What do you think of this story?

Well, sometimes I am a little spaced out, and this story gets me a little spaced out. This is Jeff Bezos suggesting that he would like to put 51,600 satellites into low Earth orbit to be rather a lot of data centers to run AI. The idea of putting data centers where solar power is plentiful and there is no land to have a land problem seems great.

A couple of challenges I can see in this. One is satellites are not very serviceable. So when something goes wrong, when an upgrade's required, when an update goes wrong, it's a little difficult to send somebody out to service these data centers in space.

So they need to be really reliable, really well-managed. NowBoth Amazon, AWS, through Jeff Bezos's world, as well as SpaceX, have been working in this idea of managing things that are in space and that need to be highly reliable. So maybe that's not so much of a concern, but nobody's put general purpose compute in space.

The ability to operate large-scale general purpose compute in space is still an unknown, unsolved problem. We know there are going to be huge challenges in this. There's already some challenges around the operation of the low Earth orbit constellations for communication.

Making larger low Earth orbit satellites is going to further exacerbate some of these challenges, right? The communication satellites are relatively small, but by the time you're building a data center and you're harvesting enough solar energy to be able to run that data center, you're talking about a pretty huge size for each of these satellites that is going to be running AI. So I am deeply skeptical that this will ever actually turn into a working system.

As we've discussed previously, Tom and I have waxed lyrical a little about some of the challenges involved in putting a very heat-intensive function into space, because it's not like you can tap into a passing river and move the heat out of your data center in space. You have to radiate it, and radiative cooling is not very efficient. That's why we don't use radiative cooling on ground-based data centers.

We typically use evaporative cooling. Shipping a lot of water into space to evaporate for cooling is not going to be workable. So there's a whole bunch of challenges in here that need to be resolved out.

There's a part of me that says, as you said, Guy, this is Elon Musk versus Jeff Bezos again. Both of them have got somewhere in space, and maybe the new race is in a different direction, getting the most satellites or getting the biggest hero numbers. " I'm waiting to see some of these data centers actually get launched, see whether they actually survive, or whether this is just a lot of noise about a concept that really doesn't address any of the real problems we have with the massive build-out of AI data centers.

The Federal Communications Commission has added foreign-made consumer routers to its covered list. This means new models will no longer be approved for sale or import into the United States due to national security concerns. The decision followed a determination by the UNIS National Security Agencies and aligns with policy priorities outlined by Donald Trump regarding supply chain independence and cybersecurity protection.

Existing routers already approved or purchased are not affected, but future foreign-made router models must receive special conditional approval from the national security agencies to be sold in the United States. Steven, this has really caught your attention, hasn't it? Well, you could say that again.

It's caught a lot of people's attention. " And the answer is, yes. Yes, the FCC says no Wi-Fi routers for you.

You can no longer import any Wi-Fi routers as of yesterday. And the answer, as you said, the reason is something, something cybersecurity, something, something national security. It is actually apparently true that this is a reaction to the Salt Typhoon persistent threat that has been sort of a thorn in the side of the United States for the last few years.

By the way, it is apparently an espionage unit of China's Ministry of State Security, and it has apparently managed to infiltrate many networking devices around the world. We've actually covered it here on "The Rundown" plenty of times. In response to this, the idea was that we should increase the security of routers.

Now, the way that we would do that, well, it's interesting. We had a whole group of folks working on it within the United States government that were trying to, within the Cyber Safety Review Board, trying to investigate the Salt Typhoon threat and figure out remediation efforts. They were all fired in April of 2025 by the Trump administration, and instead, what we've got is this bizarre situation where literally on a Monday night, the FCC issues a warning that says no foreign routers can be imported.

The FCC, by the way, only controls wireless spectrum, so this only applies to Wi-Fi routers. It only applies to consumer devices at this point, and it would allow companies to petition to be allowed to continue imports. That process apparently includes a disclosure of corporate ownership interests and a promise to move manufacturing to the United States from wherever it is and quarterly updates on these two issues.

Importantly, what it does not include is anything related to security or any information at all About what makes a domestically produced router. So I have questions, as maybe you do too. Is a compact PC running pfSense an American router?

What if it's running Microtik or OpenWrt? How about the Tesla Starlink router, which is a satellite dish that is actually made, in fact, it probably is the only router made in the United States. It includes Wi-Fi, it includes satellite connectivity, and it really is made in Texas.

Is that legit? Would, for example, maybe SpaceX would come out with a consumer-grade router that you could buy at Walmart? I don't know.

Honestly, to me, the one company that stands best to benefit from this would be HP, because they have the ability to produce products like this. They have in the past produced routers. They could, in the future, produce routers in the United States, and they could get them to market fairly quickly.

Will they? I do not know. But yeah.

It sounds wild to say it, but literally as of now, no more routers can be imported into the United States, and it's going to do nothing to improve security. Developers could unknowingly download malware hidden in regular software updates. At KubeCon Europe 2026, where I just was, Cloudsmith unveiled a security upgrade that adds live threat data to software components.

Using info from OpenSSF and the Exploit Prediction Scoring System, EPSS, system can automatically block suspicious code and risky dependencies before they reach a company's network. With almost half of organizations already hit by third-party software breaches, this tool helps prevent AI-generated code from overwhelming security teams and avoids potential fines under the EU's Cyber Resilience Act. Sounds like a great new development.

Alistair, what do you think? I agree. I think it's really important to get some of that life cycle information to go along with the created artifacts, the containers that are images that are actually being spun up, that contain a whole bunch of software.

So the idea of having the software bill of materials that lists out when you're running this container image, here are all the things that are inside it, having that automatically generated as the container image is being built, and having that bill of materials mapped back to, are these things safe? Are these things allowed in our environment? In addition, having some interactions at the time that that artifact, that container image, is being built, so that the developer maybe who is vibe coding this thing, hopefully not.

Hopefully, they're using more of a functional description way of using AI to build these applications. As these applications are being built, some feedback loop can be in there of you're trying to use a dependency that we know is unsafe in our environment or is not certified in our environment. Having that early feedback loop say, "Solve the problem with a tool that is approved," rather than trying to use one that's not approved.

That early feedback to the developer is really important. Always in DevOps cycles, you want early feedback, and a very tight feedback loop to the developer of success or failure of whatever they're attempting to do, whether that developer's a human or whether it's an AI agent along the way, that tight feedback loop is really important, and this is an important part of what the Cloudsmith innovation announced at KubeCon was all about. I think this is a really good thing.

Hopefully, we'll see more and wider use of software bill of materials and this automation of threat identification and management of threats by policy, rather than by having to deal with every single one independently. So I think that's a really cool thing to see. Speaking of things announced at KubeCon, running AI in the cloud just got safer.

This announcement from RapidFort and Nutanix is announcing a partnership that adds RapidFort's automated container hardening to Nutanix's Kubernetes platform, NKP. By removing unnecessary code, they create near zero vulnerability images, letting companies run AI workloads faster without constantly patching. This approach also provides cryptographic proof that every software component is secure, keeping apps safe wherever they run, in data centers or at the edge, and I think this is again one of those great announcements out of KubeCon of making things safer and easier to operate.

Is this going to be a great innovation for software developers, Guy? Well, first of all, I'll tell you what I like the most about it. Listen, this kind of code inspection is all the rage right now.

It's a rapidly developing area. It's more general purpose, but this is specific to AI and AI workloads, and also what I love about it is that it's integrated. It's integrated with the Kubernetes platform, in this case, NKP, Nutanix Kubernetes Platform.

So what does that mean? If you're running AI container images on NKP or anyone else, they tend to be massive. And so anything that can strip down the size and the operation of, or the existence, I should say, of totally unnecessary code, that's what RapidFort does.

And when it's integrated into the Kubernetes orchestration platform, then it's essentially automated for you. And that way, you're allowed to let developers be developers, which involves a whole lot of junk creation, involves a whole lot of experimentation. In other words, this is automated cleanup.

It's a maid service for your AI containers. And then, for common vulnerabilities, so-called CVE, having to patch containers continuously to address CVE, that's really disruptive to AI pipelines, which depends so much on throughputSo this solution also helps eliminate the vulnerabilities that are potential there at the source, namely the code. So patching isn't as necessary nearly as often.

And the audit trail and documentation and that sort of stuff, it's a really nice match, I think, for what we just talked about from Cloudsmith. Cloudsmith is taking one approach. This is taking another approach to the software bill of materials that reduces it in size, makes it tight, and therefore safer and more reliable.

Shares of Super Micro Computer plunged after US prosecutors charged employees and associates with illegally diverting billions of dollars, billions of dollars worth of servers containing AI chips from Nvidia, diverting them to China. The indictment alleges fake paperwork, dummy servers, and rerouted shipments through Southeast Asia to bypass US export controls. The case highlights the growing geopolitical battle over AI hardware, export restrictions, and the global demand for high-performance GPUs used to train artificial intelligence models.

Alistair, you're in that hemisphere of the world where all of this crazy action seems to be going on. What do you think about this story? Well, I think there's a few things going on in here.

Now, on "The Rundown," we've covered the back and forth about which parts of Nvidia's production estate, which of their GPUs are allowed to be shipped to China, and the changing landscape on that over time. This particular story I find really interesting on two directions in it. Firstly, the mechanics of what happened here.

So let's be very clear on this. Super Micro Computer is not being accused of having done anything illegal as a company. They've not been involved in any illegal activity as far as the allegations that have made.

The allegations that are in here, the indictment that's in here is that a director at one of the founders at Super Micro was involved in this outside of the Super Micro business, but using presumably the Super Micro business in a way that's a little inappropriate, as well as another employee and a contractor. The founder and the employee have been placed on leave from Super Micro while they're under investigation. The relationship with the contractor has been terminated.

So, Super Micro themselves as a company does not appear to have done anything illegal here. The impact on the stock, though, is interesting. Quite a significant drop in the stock price as a result of this court case.

33% reduction in the stock price. I think there's two things that are going on that are causing this. One is when you say, well, you've made billions of dollars from shipping these Nvidia servers, and undoubtedly Super Micro's made money along the way on this.

That turnover is going to disappear, so Super Micro is maybe less valuable for it. But there's reputational damage because a founder has been found to be doing something illegal, even if they weren't doing it inside of Super Micro. So I think that's really unfortunate for Super Micro that one of their founders has been indicted for doing poor things.

The other interesting element in here for me is the mechanics of this. I mean, how the heck do you manage around the constraints that both Nvidia and the US government puts on the shipping of their GPUs and the activation? There's a licensing component for the Nvidia GPU.

You've got to enable that before you can use them, and so there's some real interesting things around it. There must have been a lot of work here done to make this shipping of Nvidia GPUs to China actually work. As some of the coverage talks about dummy servers that have copied serial numbers from Super Micro.

So this is why we see that people who are senior within Super Micro who are involved in this slightly shady side, well, very shady side deal of taking legitimate serial numbers and applying them to dummy servers that contain these GP-- That seems like something you'd read in a book rather than something that's happened in the real world. But again, when you're talking about millions of dollars worth of turnover here, potentially billions of dollars, in fact, it is saying billions of dollars of equipment, yeah, you take extreme measures to make those billions of dollars. Often we find that there are unintended consequences to every regulation we put in place.

All right, what does it mean for GPU vendors? Well, Nvidia is possibly going to be asked to be doing a better job of tracking where their hardware goes. They did at one stage propose having geo-block capability and just not being able to activate your Nvidia chips if they were found to be physically inside China.

No idea how that was going to be implemented. I don't think you're building a GPS chip into the GPU. But Nvidia will get some scrutiny around how could they have better protected against this.

And there'll be some more observation of supply chain because if the United States wants to maintain the dominance in AI, and a closer look will show that the current administration definitely wants to remain dominant in AI, well, some of that is denying the top-tier technologies to potentially enemy states. And China is one of those that the United States has identified as not wanting to supply with top-end GPUs. We'll see how this plays out.

There will undoubtedly be more of the story of controlling where GPUs go and auditability of what's going on in these businesses. While we're looking at the regulatory control of AI, let's take a closer look at a recent announcement from the Trump administration saying it wanted to take AI regulation away from the states and create a single national standard. The White House proposed a framework led by their AI tsar, David Sacks, and aimed at giving tech giants like NVIDIA and Microsoft more freedom to innovate.

The plan would speed up data center approvals and protect AI training under copyright law while keeping some rules for child safety and free speech. Its main goal is to prevent states such as California and Texas from slowing down the AI industry. And one of the big things that's sitting in the way of building out the huge number of AI data centers that the industry is expecting to have is that people don't really want massive data centers right next to them, driving up the cost of power, consuming water from the local environment, creating large amounts of noise, and yet not creating a lot of jobs.

These AI data centers are massive and automated, and there's a bit of opposition in a number of locations. So California and Texas particularly, we've seen. And we've seen this in previous episodes of "The Rundown," that there is opposition to these massive data centers being built.

So this legislative framework that the current administration has put out has some great announcements up on the White House website about protecting children, empowering parents, and safeguarding and strengthening American communities, and respect for intellectual property. And then at the bottom, it has this little bit, which says: "Importantly, this framework can only succeed if applied uniformly across the United States. " Now, that's the announcement rather than the actual framework, but it does become clear that there's a lot of headline, highfalutin', this has got to be good for everybody, without probably a lot of specifics about how that's going to be achieved.

And then the federal government wants to control this, and your local state government should not be getting in the way of the federal government's intention. Guy, do you think AI is that strategic to the United States that it becomes one of those things where a state's choice needs to be overridden by national good? Well, Alistair, AI is of national importance to pretty much every nation.

Is it of national importance sufficiently to block US states from regulation? Well, first of all, let me just say out of the gate that federal US AI policy superseding state policy is a dead letter already. It's a mere announcement.

It's not going anywhere, and I'll tell you why. It's not just Texas and California, it's Kentucky, it's Mississippi, it's Alabama, it's North Carolina. It's all kinds of so-called red states that are dealing with irate local communities, red as they come, that are seeing things like their power bill go up or potentially go up.

So this isn't getting anywhere because it's not going to muster enough votes from the president's own party to turn into law. That said, however, I have not detected in any industry in the United States since its founding, any difficulty in innovation or development based on the state's ability to regulate. In fact, quite the opposite.

California was very much a driving force in the development of electric vehicles, which I think we could all agree is an advance, whether you believe that they are ready for you to drive or not. And in the same way, AI is like much of the tech industry, in bed and connected to the national administration, the Trump administration, and they want to have as much room to run as they can get. And it's really hard to do that 51 times for the 50 states plus the federal government.

They would rather just do it once. So I take this as an announcement and a statement of the importance of AI, which everybody kind of knows, at least within the industry, to try and get political support from these communities, that as you note, are affected not just in power, but in water, in the potential for jobs or lack of potential for jobs. I think it's significant in that sense.

It's very significant from a political standpoint. Where are we as a country in either rallying around or rallying against AI and the development of AI? No matter where you are around the country, we have some great events, in fact.

The RSAC event is just winding up for us, so watch for all of our coverage there. But coming up in April, we have a big month. We have Networking Field Day turning 40.

Tom Hollingsworth will be back out in the Bay Area. He's got a great lineup of some companies that have been regulars with us. We've got Upscale AI, Selector AI, Arista, Reviz, Cisco, Lightyear, NetAI, as well as Netris and Nokia, and a great panel of delegates there as well.

That's the 8th to 10th of April. The next week, it's Stephen's turn. He'll be out in Orlando with Qlik, and he'll be at the Qlik Connect 2026 conference.

It's a nice little panel of delegates there, and we'll be digging into all that's new with Qlik. And then rounding out our April, we have Security Field Day. That's Tom again.

He's going to be heading back out to the Bay, looking at all things new in security. He's just still building out the panel of delegates there, but I'm glad to see Fernando Montenegro, who is the future and research director for cybersecurity, will be at the event. Looking forward to hearing some more coverage from him.

So it's going to be a great time with April being very heavily networking and data-oriented for Tech Field Day. Of course, plenty more Tech Field Day events coming up. " You can catch new episodes every Wednesday as a YouTube video on your favorite podcast application.

"The Rundown" is streamed on Techstrong TV as well, and you can often catch us on other Techstrong and Futurum Group programs. We'll be back next Wednesday to talk about all the IT news in the week that was, and I should have Tom back with me for that one. Until then, for myself, for Guy Currier, and for all of us here at Tech Field Day, just wishing you and yours a great day and a great week.