The Right Way to Deal With AI Data Centers

Today on the AI Daily Brief, the right way to deal with AI data centers. Before that, in the headlines, updates on Mythos, SpaceX's Neo Cloud, and much, much more. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Robots and Pencils, Mission Cloud, and OutSystems. To get an ad-free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts. And to learn more about sponsoring the show, send us a note at sponsors@aidailybrief.ai. Last note, I've been sharing the new training.beesuper.ai. This is the updated home of some of the programs that we've been experimenting with this year, including the Executive Catch-Up Program, as well as the inheritor to Enterprise Claw, which is the enterprise-grade version Executive Agent Leadership Program. That is a 6-week sprint that will have you not only building agents, but also building the organizational infrastructure, policies, et cetera, around them that you need to actually use agents. The next cohort for that launches on June 29th. Sign up now. Again, you can find all the links that are relevant at training.beesuper.ai. We kick off today with an update in a story that we were covering yesterday, which is around the NSA dimension of the Mythos Fable ban. Now the TL;DR of what I was discussing in that initial coverage was that there seemed to me to be a mismatch between what was actually said and the way that it was being interpreted. In short, people sitting around waiting for Fable to return were looking for a reason that made a little bit more sense than there just being this jailbreak that Amazon had reported, and when they went back and found a story that seemed to indicate that the NSA had been hacked by Mythos, it seemed to some to make more sense. Now commentators with an understanding of NSA operations pointed out that what the director was telling Senator Mark Warner was likely a discussion of a controlled red team exercise rather than some live cyber attack. Shashank Joshi, the Economist reporter whose article went viral, has now provided an update, writing, a U.S. official tells me that Senator Warner misunderstood the NSA Director General Rudd in this case. Rudd did use the 'hours, not weeks' wording, but the use of mythos in this context was, as widely assumed, part of a red teaming effort. I.e., testing the security of internal networks. The official also told me that the agency's red teams no longer have access to Mythos because their authority for accessing it was under Project Glasswing. Yoshi pointed to commentary from a cybersecurity account called Iris C2 suggesting that they had the correct read on what had actually happened. That account noted that, quote, in the case of almost every exercise, the red team begins with some kind of initial access to the air-gapped classified network. Essentially, classified NSA systems aren't generally accessible from the outside, and gaining that initial access is one of the most difficult parts of any attack. Iris C2 commented that Mythos doesn't suddenly mean that any random person can suddenly hack the NSA. However, they did note that if an attacker manages to gain access, Mythos makes the design and execution of an exploit much faster, reducing the available time to detect and contain an intruder. So the point, if we're looking for takeaways, is not that Mythos has broken into the NSA, but that it does raise the stakes for cybersecurity by making attackers far more efficient. Now, the bigger detail for many was the throwaway line about the NSA not having access to Mythos anymore. Yet another reason to want this situation to be resolved as soon as humanly possible. Now, staying on the cybersecurity train, OpenAI has updated their cybersecurity model alongside a big new cyber initiative. On Monday, OpenAI announced a significant update to their Daybreak security initiative. Daybreak was first launched in May as an answer to Anthropic's Project Glasswing. As part of the initial rollout, OpenAI made a preview version of GPT-5.5 Cyber available to trusted partners. As a point of differentiation to Glasswing, OpenAI invited smaller organizations to apply for access. OpenAI wrote, Frontier defensive capabilities should not be concentrated in the hands of a few. As AI changes the pace of vulnerability discovery, defenders everywhere need democratized access to these models to find, fix, and protect their infrastructure before attackers can identify and abuse these flaws. With the expansion, OpenAI has now launched the full version of GPT-5.5 Cyber, their first model that's been fine-tuned for cybersecurity work. Like Mythos, the model has reduced guardrails to ensure professionals can use it freely in their work, and OpenAI has updated the Codex security plugin to make their harness more performant for cybersecurity tasks. OpenAI is claiming that with the new updates, their new Cyber model has overtaken Mythos on the relevant benchmark CyberGym. In addition, OpenAI is launching a new initiative called Patch the Planet in partnership with security research firm Trail of Bits. Reporting on their initial testing of GPT-5.5 Cyber, Trail of Bits wrote that they'd found hundreds of bugs in open-source libraries. They've deployed 37 patches so far and have many more in the pipeline. Over 30 open-source projects have already joined Patch the Planet, with the goal of securing the critical software that underpins the digital world. Trail of Bits noted that the introduction of strong AI security models has fundamentally changed the nature of the work. They wrote, "If it wasn't already clear from the last several months of security news, this week makes one thing clear." The expensive part of security work has moved. The advantage is no longer in finding bugs, but everything after— confirming a finding, getting its severity right, writing a patch a maintainer will accept, and coordinating a disclosure. That is the work that floods of AI-generated reports threaten to bury. The release comes as the Five Eyes intelligence agencies, which is an intelligence alliance between Australia, Canada, New Zealand, the UK, and the United States, issued a rare public alert for AI-driven cyber risk. In a bulletin published on Monday, the UK National Cyber Security Centre wrote, The evolving landscape of AI is rapidly transforming cyber risk, and we must act swiftly to remain ahead. The bulletin called on the business community to assess the changing risk, prioritize cybersecurity practices, and, quote, stay actively engaged as threats and guidance evolve. The agencies warned that the rapid pace of frontier AI development means cyber risk assumptions can become outdated in months, not years. They also heavily encouraged the integration of AI tools into cybersecurity operations and warned that, quote, cyber risk can no longer be treated as a purely technical issue. This is a core business risk and leadership responsibility. Now, nothing in the bulletin will be surprising to anyone who has been paying attention, but it's pretty clear that the intention of the bulletin was to light a fire under the enterprises who perhaps are not spending as much time listening to the AI Daily Brief as they should. Now, staying in the regulatory space, a story that isn't exactly related to AI, but which is being kind of lumped in as frontier tech writ large, President Trump has called for the construction of a powerful quantum computer in a pair of new executive orders. One order instructs federal agencies, including the Energy Department, to collaborate with private industry to deploy a quantum computer for scientific research purposes. Michael Kratsios, director of the White House Office of Science and Technology Policy, said he believes a functional computer can be done by 2028. In addition, the first order instructs the government to migrate to quantum-secure cryptography by 2031. The second order deals with protection of intellectual property within quantum companies and hardening the supply chain for components for quantum computing. This order emphasizes the need for international cooperation to prevent the technology from falling into the hands of adversaries. Now, for anyone who's been paying attention to the quantum story, it is honestly a little hard to tell exactly where the state of the technology is, nor is it particularly easy with this administration to know at any given time what the motivations behind an executive order are. So for this one, I'm going to have to just leave it at the news and make of it what you will. Now, on the market side of the house, SpaceX has signed another multi-billion dollar data center deal as Elon Musk extends his compute empire. Open source AI startup Reflection AI agreed to pay $150 a month to rent capacity from the Colossus 2 data center. That agreement begins next month and runs through to 2029, putting the total deal value at $6.3 billion. As with the other SpaceX deals, either party can walk away on 3 months' notice. This is also significantly smaller than the Anthropic and Google deals, each running at around $1 billion a month. Still, with SpaceX coming down from its IPO sugar rush, anything that helps people understand the long term of the business is probably helpful. Reflection AI, meanwhile, used the deal as a way to promote themselves as a domestic open-source alternative. In a press statement, the startup said, "Recent events highlight how important open source is to the AI ecosystem, with more nations and enterprises recognizing the risks and costs associated with exclusively depending on closed models." They said that the deal signals their strategic importance in the frontier AI ecosystem and that more compute would give them more runway to develop leading open models. Now, Reflection is yet to release their first frontier model, but has been working with government partners including the Pentagon and the Department of Energy's Genesis mission. Now, Swiggs from Late in Space thinks we might be underestimating the potency of SpaceX's new moves. He tweeted, I don't think anyone is correctly doing the math around how SpaceX, the Neo Cloud Neo Lab, is currently going to market. SpaceX has already recouped about half its investment in Cursor in compute deals. The other half is paid for if Composer 3 does well. No other company is simultaneously a leading model lab and Neo Cloud, at least where GPUs is concerned. It's a crazy effective combo if you've adequately planned out GPU supply, if in-house training 1) goes very well or 2) doesn't go very well. Which is certainly not to say that SpaceX did well yesterday, actually experiencing a 16% drop, but it wasn't the only AI company running into some trouble. Google stock fell sharply as the market reacted to the loss of two key AI researchers. On Monday's show, I covered the departure of Nobel laureate John Jumper, who left DeepMind for Anthropic just a couple of days after Noam Shazeer packed his bags to join OpenAI. The sort of standard reading of the situation is things going sideways at DeepMind, And there were plenty of anonymous quotes to be had to suggest morale was plummeting as Google lags behind in the AI race. Now, while my personal take was that we have a tendency to overreact to personnel changes, evidently the market disagrees. Google stock was down as much as 7.2% on Monday, its largest intraday move since February. That means that if you can attribute this directly to those two employees departing, those departures cost the company over $200 billion in market cap. Now, it might be reasonable to chalk this up to the market over-indexing on AI headlines once again, as they did during the Deepseek moment. But there is also a more fundamental narrative shift happening around Google. Up until very recently, Google had been the top performer in big tech and even briefly became the largest company in the world. And yet the more that the perception is that Google's models aren't keeping pace with Anthropic and OpenAI, particularly on enterprise-essential agentic use cases, skepticism has started to creep in. With that as background, Jumper and Shazier moving to Rival Labs seems to confirm that point of view, rather than simply representing some new trend. Gil Luria, the head of technology research at D.A. Davidson, summed up that viewpoint saying, "Google is losing the war for talent at the frontier of AI. Google had the state-of-the-art model for a few weeks last year, which helped it get credit as an AI winner, but has fallen off since, and these departures may mean it is falling behind." Now I will say I'm sticking to my guns around reading too much around any personnel moves, and I think Prime Intellect's Florian Brand is directionally correct when they write, We are in peak Google is done for and will never catch up, which is always followed by them releasing a new model and people going, the others are done for. No one has the TPUs and data that Google has. For now, it is for markets to debate which side of that equation is more accurate. But for us, that is going to do it for the headlines. Next up, the main episode. One of the most important AI questions right now isn't who's using AI. It's who's using it well. KPMG and the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising. The highest impact users aren't better prompt engineers. They treat AI like a reasoning partner. They frame problems, guide thinking, iterate, and push for better answers. And the good news? These behaviors are teachable at scale. If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time. Learn more at kpmg.com/us/sophisticated. That's kpmg.com/us/sophisticated. One thing I keep seeing in enterprise AI, companies hedging across every cloud, every model, every framework, or paying a GSI for a pilot that never ends. The teams actually shipping, they've picked a lane and they move fast. That's one of the reasons I like today's sponsor, Robots and Pencils. They've gone all in on AWS. They're an advanced tier and AWS pattern partner, and they ship production AI coworkers in 45 days. That's led to them doing some of the more interesting work I've seen on AI coworkers. And by that, I'm not talking about chatbots. I'm talking about actual agentic systems that sit inside a business architecture and do real work. That kind of focus matters if you're an enterprise leader trying to get something real into production or an AWS rep trying to move a customer from interested to deployed. Request an AI briefing at robotsandpencils.com. One conversation with robots and pencils and you'll know. The average enterprise is spending $11.5 million on AI this year, and most of them can't prove a single dollar came back. What does AI actually look like when it produces ROI? Ask the healthcare company that just made their payment processing 320 times faster, or the law firm whose document research went from 3 months to 10 minutes, or the contact center who reduced wait times by 99%. These are real Mission Cloud customers with real results. Mission Cloud is a CDW company and an AWS Premier Tier partner. They're the AI-first, outcomes-obsessed AWS experts who build AI solutions that drive your business forward. Whether you're flooded with AI ambitions but no idea where to start, or 6 months into a deployment that's going sideways, they've seen it and they've fixed it. Stop burning your budgets on AI that doesn't produce results. Start at missioncloud.com. This episode of the AI Daily Brief is brought to you by OutSystems, a leading agentic systems platform built for the enterprise. Organizations all over the world are building, orchestrating, and governing agentic systems on the OutSystems platform, and with good reason. OutSystems' open and unified platform allows teams to architect, deliver, and scale governed agentic systems with agility. Teams of any size and technical depth can use OutSystems to build, deploy, and manage AI apps and agents quickly and cost-effectively without compromising reliability and security. With OutSystems, you can rapidly launch ideas from concept to completion. It's the leading agentic systems platform that is unified, agile, and enterprise-proven, allowing you to accelerate growth, reduce operational friction, and deliver real enterprise impact with AI. OutSystems. Build your agentic future. Welcome back to the AI Daily Brief. Today we're gonna take advantage of a slightly slower news cycle in which everyone is somewhere between sitting on their hands waiting for Fable 5 to return, or going full open source tinkerer building out local infrastructure in their basement. To try to discuss calmly one of the most contentious issues surrounding AI, which is the impact of data centers. Now, on any given day, I could find a reason to discuss this topic, especially with midterm elections coming up. It is becoming more and more of a hot button, and it's also rising in cultural significance. On his popular This Past Weekend podcast, Theo Vaughn recently went on a data center rant. He said, nobody wants a data center, dude. And the people that want them, to me, they seem kind of evil. One of these companies is gonna own all this information. There's gonna become this social or emotional credit score, and then AI is going to try to become our new god, which by the way, has to be a contender for the most misperceptions or concerns, depending on your perspective about AI shoved into a single paragraph. And yet Theo Vaughn here is, I think, more reflective of an emerging mainstream point of view than he is even a leading indicator. AI researcher Andy Masli recently posted a New Yorker cartoon of a little robot coming into its parents' bedroom with the mom reaching over to the dad and whispering, "It's AI again. He wants another 1,000 glasses of water." As Andy, who has done more to debunk some of these myths than just about anyone, captioned, "This idea will never die." And on those lines, you might have recently noticed that famed activist Erin Brockovich has started a major campaign against data centers. And on the other side of the political aisle, former Tea Party conservatives are now planning a nationwide protest against AI data centers. The New York Times quoted comedian Charlie Berens calling opposition to data centers the most bipartisan issue since beer. Now, when you hear critiques, despite Theo grabbing some concerns that we've previously heard around things like central bank digital currencies, i.e., the worry about a social credit score, for most people, the two big issues come down to water use and energy prices. Take, for example, this recent tweet from Indiana resident Valerie Ann Smith. She wrote, Amazon is dropping an $11 billion AI data center right in Indiana. This monster will guzzle electricity for 1 million homes and 300 million gallons of water every single year. Just one facility. Our grid is already crumbling. Bills will explode. Water shortages incoming. Blackouts while they power their AI overlords. And these same tech giants lecture us regular people about climate change. Nicole Dividar, who so far as I can tell from just their Twitter profile, seems to be pretty much as opposite politically as she can be from Valerie Ann Smith, reposted that and reiterated that key point. Amazon's monster data center projected to consume 300 million to 330 million gallons of water every year. Insane. The way we're going will be the downfall of humanity and every living being. The madness must end. The thinking that nature is here to exploit must end. Now, these numbers do seem at first glance huge. 300 million gallons of water a year seems like just an enormous amount, right? Overall, Amazon recently released full water use statistics for their data centers,, and according to their figures, their global data center operations consumed 2.5 billion gallons of water in 2025. Now that report did note that water use at its sites had actually fallen 2% from 2024, even though it expanded its data center footprint meaningfully, but still 2.5 billion gallons of water, right? The problem is that relative to many of our other uses of water, these numbers actually aren't particularly large. So to use some comparisons, the 16,000 golf courses in the United States consume over 500 billion gallons of water a year. That means that a full year of operations for Amazon's data centers is slightly more than a single day of US golf course maintenance. Another frequent point of comparison is almonds. California almonds use between 1.2 and 1.8 trillion gallons of water each year, which is somewhere between 5 and 8 times the total amount of water used by all data centers in America. One recent study found that the US lost 3.29 trillion gallons of water per year to leaky pipes. About 15 times as much as data centers use. And one of the key sources for the myth of water consumption, author Karen Howe, in her book Empire of AI, had to apologize for her most dramatic claim being a factor of 1,000 wrong. Basically, she claimed that a Google data center in Chile consumed 1,000 times more water than the surrounding population, but she was off by a factor of 1,000 due to a unit mix-up. Howe eventually acknowledged the error and issued the correction, but the book is still out there saying the same thing. In Indiana, where the citizens were concerned about Amazon's data center using 300 million gallons of water per year, the state delivers a little under 500 million gallons of water per day for domestic use, or around 182 billion gallons per year. That means the Amazon data center represents 0.2% of water use for the state without including industrial use. And this gets to one of the biggest problems with the data center conversation when it comes to water, which is that if you don't like the thing that the water is being used for, even one gallon is too many. And that's really what a lot of this critique comes back to. Big numbers are politically potent because most people don't have any idea of how much water we actually consume. And these numbers all just sound so astronomically large that it's very easy to grab attention and get people to think that the water being used is just too much. Now it's more for those who think that even this is too much water, the companies in the data center space are working to reduce it even more. NVIDIA recently touted what they called one of the biggest efficiency leaps in data center history, saying that their new approach to liquid cooling could cut water use to near zero. Now, as I mentioned, the second big issue that comes up around data centers is the idea that they drive up electricity prices. And once again, digging into the numbers, it's a lot less clear than the data center opposition would make it seem. The Institute for Energy Research recently concluded, there is no statistically significant correlation between the number of data centers in a state and its current electricity prices. In fact, prices in the top 10 data center states are virtually identical to the average across other states. Furthermore, there is no statistically significant relationship between data center concentration and faster increases in electricity rates. Now, where I think that this does get more complicated is that ultimately most electricity price pressure is due in some way, not just to new data centers coming online, but to costs associated with upgrading an aging grid, a problem that we have to face even without the new data center buildout. The Daily Economy explored this research and asked why so many Americans are convinced that data centers do increase their electricity prices. They concluded, in the short run at the local level, the story is more complicated. A Bloomberg analysis of wholesale electricity prices across 25,000 grid nodes found that prices have risen as much as 276% since 2020 in areas near major data center clusters. More than 70% of nodes recording price increases were located within 50 miles of significant data center activity. In these regions, data centers create a surge in demand on local grids. When transmission capacity is constrained and the new generation has not yet come online, prices spike. Those higher wholesale costs can then filter into retail bills, at least in the short run, and local customers bear the brunt of this regional electricity demand. Now, continuing later, they write, concentrated price spikes are not evidence that data centers are inherently incompatible with affordable electricity. They are evidence that grid infrastructure and cost allocation rules haven't kept up. They then point to some of the examples of how policy is racing to catch up with this and make things more fair in the short run as well. They point to, for example, Oregon's Power Act, which requires the biggest electricity users, i.e., the data centers, to bear the cost of infrastructure that is built specifically for them. This is also the same principle behind the White House's Ratepayer Protection Pledge, where companies agreed to, quote, protect American consumers from price hikes due to data center energy and infrastructure requirements and lower electricity costs for consumers in the long term. The 5 commitments in the pledge included specifically building, bringing, or buying new power supply, paying for new power delivery infrastructure upgrades, paying whether they use the power or not, investing in local job creation and workforce development, and contributing to electric and community resilience. Now, my personal belief, and why I think this is worth talking about, is that as with much in the AI debate, the conversation on both sides gets wildly reductive and extreme. It tends to be snarky reactions like Marc Andreessen reposting Theo von's post saying, "I have bad news about your podcast, dude," pointing out the inherently hypocritical position of critiquing data centers when they are the thing that allows his podcast to reach all the people that it reaches. Or you have commentary, as Alex Finn said, that just calls this insane dribble. Now, Alex is someone who's incredibly excited about AI, and I think rightly points out that there is a huge excitement gap between, for example, America and China when it comes to AI. He writes, "If a large majority of Americans believe the most powerful, prosperous, important technology of our lifetimes is somehow evil," We simply won't be able to keep up with a country that believes it's good. And while I think that that's right, we're not going to convert people by not taking their concerns seriously. Now, I think that in all of this, maybe the most important actor for finding space in the middle are the labor unions. Labor unions represent multiple constituencies. On the one hand, they are of the communities that have these concerns and I think take them seriously. I think they are sympathetic to concerns that AI and any big tech is another way to disproportionately benefit the already rich. At the same time, unions representing key skilled blue-collar work are also at the forefront of seeing how valuable the data center buildout can be, as the demand for more skilled work from their members just goes up and up and up. The Information's Ann Davis Vaughan recently wrote a piece called "Debunking the Myths That AI Data Center Critics Believe." And I actually think that the title undersells the value of the piece. I think that her lead paragraph nails it when she writes, I've been visiting large AI data center projects in rural and industrial communities in the Midwest, Southeast, and West, and I have found that two things can be true at the same time. AI data centers have drawbacks but are better for communities than their residents think, and those communities can win a lot more financial concessions and benefits from the tech firms than they realize. I think right now communities are being taught to felt like their choices are roll over and let big tech and the data centers do whatever they want with the natural resources surrounding their homes, or on the other end of the spectrum, get your signs and pitchforks and stop it from happening entirely. I think communities should absolutely get to advocate for what they think is right for their communities, whatever it may be relative to data centers. But I think that they're missing that there is a massive middle path where they can be negotiating for the data center builders to effectively give them the world. Anne's article starts to make this a little bit real. She writes, In rural Richland Parish, Louisiana, for instance, hundreds of teachers are set to receive unprecedented $50,000 bonuses this year, funded by a surge in tax receipts tied to Meta Platforms, which is building a large AI campus there. An existing ordinance mandates that teachers get a slice of sales taxes, and teacher bonuses quintupled due to their activities related to campus construction. Not to be too crass about this, but I think that with the amount of money involved and the stakes of what's being built, we're not just talking about data center builders and owners making sure that people's electricity prices don't go up. We're talking about major intentional economic benefits to the communities if negotiated well. I hope that we can move past the knee-jerk binary phase of this conversation into the how everyone wins phase sooner rather than later. For now, that's going to do it for today's AI Daily Brief. Appreciate you listening or watching as always. And until next time, peace.