データセンター減速はBig Techに追い風になるのか

The conventional wisdom heading into the elections were that if Democrats took power as they appeared that they were going to, it would be very bad for data centers. Certainly one interpretation of the market sell-off in chips and neoclouds over the preceding few weeks was anticipating this potential reality. And yet now that the thing has actually happened, people are starting to point out that there are some very specific reasons why not only might the blue wave not be as bad as people thought for data centers, but that in fact, a data center slowdown could actually be good for big tech. 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, C-Data, and Blipsey. To get an ad-free version of the show, go to patreon.com slash AI Daily Brief, or you can subscribe on Apple Podcasts. If you want to learn more about sponsoring the show, send us a note at sponsors at aidailybrief.ai. And finally, if you haven't yet, please go check out my new podcast, How We Got Here.

It is the deep dive backstory on important issues, many of which are in tech, but not all of them, that goes way more in depth than I have a chance to here on the AI Daily Brief. This week's episode is about how we made the wrong things cheap and is particularly relevant, I think, in this cost of living election. You can find How We Got Here with NLW on any of your podcast apps or just go to howwegothere.com. Welcome back to the AI Daily Brief Headlines Edition. all the daily AI news you need in around five minutes. Is Ask Chat the new Google it? OpenAI says ChatGPT is starting to compete with Google as a search engine. During a press event earlier this week, OpenAI's consumer product lead Adam Fry said ChatGPT is now processing 20% as many search requests as Google. And while 20% may not sound like much, Google has held a 90% plus market share in search for over a decade, meaning no other competitor has come anywhere close to these numbers. Now, Fry did explain that OpenAI is classifying any query that results in a web search as a search request. So the numbers might be a little juiced by ChatGPT's propensity to

verify answers on the web. But again, no one has come anywhere close to competing with Google in this manner. And Google themselves have noticed users turning to AI search more and more. During last week's earnings call, Sundar Pichai said that strong growth in AI mode queries played a role in slowing search revenue growth because of lower monetization. The market sold off on those results, But with OpenAI finding a foothold in search, it's clear that Google has no choice but to disrupt themselves. For OpenAI, gaining market share in search is of course extremely promising for the growth of their advertising business. Ads were hugely controversial when OpenAI introduced them at the beginning of the year, but shocker, backlash from actual users has been much less intense. OpenAI's Fry again said, Ads have not negatively impacted the product or the usage of the product, and most people do not seem to mind ads in the product. He added that the relatively well-received rollout has, quote, Maybe shown us that we could be doing more of this, actually. So if you are a person who does not like ads in your chat GPT, it might be a good time to jump to the paid plan. By the way, it also looks like after a couple of quiet weeks, we might be in model release season again soon.

GPT 6.2 Astra and GPT 6.2 Sol have shown up on the Google Cloud backend, which usually indicates an imminent release. For many, there is one and only one question that matters. Can GPT 6.2 beat Opus 5.5? Next up, a bunch of acquisitions in the news. Meta has acquired AI coding startup Conductor. Conductor was founded in 2024 and focuses on the orchestration layer, offering a platform that lets developers run multiple coding agents in parallel. Meta launched their own coding agent UseCode in August, but it's received relatively little buzz. Sources said the acquisition is expected to help Meta build their next generation of AI coding products. And it sounds like Meta is going to try to stand out from the crowd with a new approach to coding harnesses. Conductor founder Charlie Holtz wrote in a blog post, We have some big, crazy plans for the next interface. If you liked Conductor, you're going to love what's coming. In other acquisition news, OpenAI has agreed to buy Pierre Computer Company to enhance their Codex product. The three-year-old startup produces open-source developer tools, which help teams store their codebase and track changes.

Codex already leans on Pierre's tool, so it makes sense to bring the team in-house, particularly if OpenAI is still aiming to add integrated services and eventually start to compete with GitHub. One thing to watch over the next year will be whether competition between frontier models flattens out and we start to see more focus on the agentic product layer. These acquisitions seem to indicate that the leading labs know it won't be enough to just have the best model, and that in fact they'll need to have the best overall product experience as well. Over in enterprise adoption land, Apple has cut their internal AI budgets as part of a cost-cutting measure. Bloomberg's Mark Gurman reports that Apple quietly lowered their employee spending caps last month from $300 to $200 a day. Another source said another cut is potentially on the table and could reduce limits to just $100 a day. Now, these cuts were framed as impacting engineering teams specifically, so this at least isn't about Apple trying to trim AI costs in marketing or finance departments. Non-engineers already have $100 spending limits and could see those cut to as little as $25 a day. And broadly speaking, there's nothing all that surprising about Apple looking for cost savings in their AI use. This has been a theme across corporate America over the past six months,

and is as much a byproduct of the fact that AIU shifted from a flat monthly subscription plan number to an effectively unlimited API number at the beginning of the year. Apple's limit of $300 a day represented over $6,000 a month limit. In contrast, Uber imposed a $1,500 a month limit in June. It's also not like Apple's policy here is particularly coherent. On the one hand, they are pushing to modernize their engineering process and other internal operations using AI, but then at the same time asking employees to rein in their experiments to keep it under budget, and the mixed messaging extends externally as well. Last week, newly installed CEO John Ternus told Wired, We don't want to ship AI for AI's sake. When we're doing it right, people won't have to think about whether it was AI that made this happen or more traditional software. What matters is, did we build something that they really love? However, during this week's earnings call, Ternus was hammering the AI theme. AI and Apple intelligence were mentioned 45 times, a new record for Apple. CFO Kayvon Perrick identified AI features as a key revenue driver, commenting, We feel AI offers a compelling reason for customers to upgrade and switch in order to

get the latest capabilities. Bloomberg's Apple insider Mark Gurman's sources inside Apple seemed just as confused about where the company is heading. Some said the $300 a day limit was excessive, with few engineers ever getting close. Others thought that the AI cuts were a sign of deeper cost-cutting measures to come, including potential layoffs. Reportedly some Apple divisions have brought in consultants to evaluate their efficiency and operations and Ternus is believed to favor a flattened management structure to help the company run faster and leaner I think one of the more interesting conversations coming out of this comes from OpenClaw creator Peter Steinberger, who wrote, Do individual limits still make sense? As capabilities increase, you want to work with and support an agent that represents your project, that runs loops and evolves your code base. The humans merely curate the automations. Put differently, when we talk about the shift from individual to team AI, the whole presumption of individual budgets gets thrown into question as well. Way beyond the scope of the headlines, but definitely something interesting for us to keep an eye on. Now, staying on the adoption theme, while U.S. megabanks are banning agents, some smaller banks are finding a way to make agentic finance work. Last month, CEOs from both J.P. Morgan and

Bank of America said that they'd be blocking agents from accessing customer accounts, the risk of agents making mistakes being seen as too large for the big banks. The information, however, reports that a crop of smaller fintech-friendly banks are experimenting with agentic payments. One example was LeadBank in Missouri, which is allowing their fintech partners to issue bank account numbers for agents. These function in a similar way to a credit card issued to agents, which allow customers to set spending limits and restrictions on what an agent can do. The accounts also have a panic button that locks an agent out of it if it goes haywire. Internet-only Increase Bank is also offering segregated accounts for agentic use, as well as providing an agent-friendly API. Bank owner Derek Buckley said, We don't need the entire industry to coordinate banking for agents. It's sufficient to have just a small cohort of technology-forward banks. Finally, one update on the theme that we are watching of AI and biology. Google's AI drug discovery spinoff, Isomorphic Labs, is in talk to raise $5 billion in outside funding as they scale up operations. Isomorphic was founded in 2021 as a vehicle to build on top of DeepMind's protein folding model, AlphaFold.

Earlier this year, then-DeepMind CEO Demas The Sabas moved on to the chairman role, and he reported at the time that reduced duties at DeepMind were in part to allow him to focus on scaling isomorphic labs. Sources said the round would value the company at $50 billion, and the new capital raise seems to signal that Google will treat isomorphic and, by extension, AI-assisted drug discovery as a business capable of standing on its own. For those who are excited about AIs for health science and drug discovery, this sounds like it would be capital well invested. And that is going to do it for the headlines next up, the main episode. A new study from KPMG in the University of Texas at Austin found that when people work with AI, similar skills don't guarantee similar outcomes. Researchers studied more than 500 early career professionals and found that the best performers consistently amplified the value of AI by guiding, evaluating, and refining its outputs. These top performers, called AI amplifiers, weren't defined by what they knew alone, but by how they worked with AI. Learn more about what

separates AI amplifiers from everyone else at kpmg.com slash us slash AI amplifiers. The best teams don't have a single star carrying everyone else. They know their own strengths and each other's weaknesses and play to both. That's the team Robots and Pencils has built on purpose. Nobody there is grinding through busy work to pad a headcount number. People come for the hard problems and they stay because everyone around them is leveling up at the same time. In a market full of companies that are just trying to hire fast, that's worth a look. Check out robotsandpencils.com slash careers. This episode is brought to you by Cdata. Token spend is climbing, security is holding agents back, and companies rolling out AI initiatives might not know how it's truly being used. The root cause, AI is cut off from the systems and context your business runs on. That's why Cdata built Connect AI Gateway, one control point between your AI and your systems. Connect AI Gateway is not just another static model router. Prompts get intelligently routed to the right model with exactly the tools and data it needs governed down to the record.

The result? Accurate answers at a fraction of the token cost with a context engine that learns your business with every interaction. Built on the connectivity layer powering Palantir, Google, Salesforce, and hundreds more. Give it a try and start your free trial at cdata.ai. That's C as in connect, data.ai. Blitzy's deep code-based understanding unlocks the thing every roadmap owner cares about, shipping new features. Here's the truth about building inside a massive enterprise codebase. Writing code was never the bottleneck. Context is. Which system does this touch? Which contracts can't break? Which standards apply? Blitzy already knows because it reverse-engineered your entire codebase into a dynamic knowledge graph before feature work began. With that complete picture, Blitzy builds features end-to-end. Architecture, APIs, UI, and tests all validated against your existing systems. One Blitzy customer built an AI-native application from scratch with 100% autonomous completion, saving over 2,700 engineering hours. Features that respect your code base instead of fighting it. Stop letting your backlog grow faster than your team. Accelerate your roadmap at blitzy.com.

That's B-L-I-T-Z-Y dot com. Welcome back to the AI Daily Brief. The day after the big Democrat sweep, everyone's attention now turns to implications. Whatever a person's particular issue or industry, Wednesday, November 4th was spent exploring what the new crop of leadership in Washington and in state houses around the country means for those issues. And certainly the magnitude of the sweep is shaping those perceptions. According to the AP, Democrats have at least an 8-seat majority in the Senate and a 45-seat majority in the House, and they also picked up five governorships, including the pivotal win in Texas. In short, outside of the Oval Office, the rest of Washington is going to be blue for the next two years. How much and what specifically will change, though, is another question. Now, certainly for our purposes on the AI Daily Brief, the obvious first question is what's going to happen with data centers? I think a fairly understandable first-layer analysis says that this is bad for them.

43 out of the 44 data center moratorium-type votes around the country passed, clearly giving an indication of where the public sentiment is. And Elon Musk joked in a way that wasn't joking at all, data centers in space looking pretty appealing right now. Now, the state where the changes are perhaps most significant when it comes to data centers is, of course, Texas. In that state, we already saw soon-to-be former Governor Greg Abbott running away from data centers, trying an 11th-hour switch that ended up being unsuccessful, and the assumption is that incoming Governor Gina Hinojosa is going to be much less amenable to data centers in the state, which could be hugely problematic given how much Texas has been the epicenter of the data center build-out. There are certainly some indications on Wall Street that investors believe that the data center construction boom is basically over Chip stocks and neoclouds plunged on Wednesday including a 3 drop for NVIDIA and a 10 drop from CoreWeave Citrini Research put out a piece called The End of the Trade as We Know It. They declared that we're past the peak for AI CapEx, that risk has shifted to the downside,

and that the data centers that do get built are going to get more expensive moving forward. Now, what's interesting, though, and what we'll come back to, is that Citrini's analysis is not just about policy. In fact, Cetrini's analysis is more about how policy interplays with other existing challenges. They wrote, investors start looking at the extended completion dates up against the existing debt maturity dates, and you start hearing phrases like takeout risk and completion defaults. Now, in that environment, they say policy is no longer ambiguous. While they argue that none of the incoming governors in Texas, Ohio, or Georgia are likely to put an outright stop to the build-out, the cost, they say, is almost certainly going to get a lot higher.

Larger municipal concessions, clean energy requirements, ratepayer protections, and slower permitting. However, in maybe the most important line of the entire piece, they write, Perhaps the bottlenecks become the excuse some people need to quietly trim their own projections without causing a rush for the exits. Now, put a pin in that because we're going to come back to that particular market analysis in just a minute. Because it turns out that when you actually look past the headlines and the obvious default assumptions, a lot of folks are taking a much more nuanced view of what the likely implications of this Democrat sweep are. Yesterday on X, I posted a summary of my analysis, which you heard on yesterday's show, which is that I don't think data centers stop getting built. Instead, I think that this newly emboldened set of leaders demand no NDAs and full process transparency. I think they demand the builders pay the full cost of infrastructure buildouts, including subsidizing everyone else's electricity. I think we start to get actual consideration of quality of life concerns and more environmental rigor. And I think that we see big money involvement in local communities in the form of direct

subsidies or payouts to citizens, incentives, build out of unrelated community infrastructure, whatever it takes to get the deals done. I believe that even though Satrini is worried about this increasing the cost for the buildout, that the data center builders will absolutely jump to agree to all of this. Those new leaders, meanwhile, get to deliver a win for everyone while still being tough on tech. And the whole system keeps turning in a way that's honestly a lot more sustainable. Now, the most reasonable pushback I got on that on X was a sort of assertion that my argument assumes that people are rational actors, with the person basically implying that the reason that people were against data centers was that they had been poisoned by a digital psyops campaign, probably architected by China. However, if you've listened to my first episode of the new podcast, How We Got Here, about how tech lost America, I think that even if there was a massive China psyops campaign, people were fairly likely to be negatively predisposed to data centers regardless, and I think we also have to account for the possibility that data centers as they have been constructed so far just haven't been very good neighbors.

But what evidence do I have? There are certainly some places that you can look where the new crop of leaders are just advocating for bans outright. And we'll actually come back to the new support for Pausing AI in just a moment. But if you look at some of the biggest stars of the new show, they're laying out a path for compliance and creating a narrative architecture that the data centers can abide by. In his media tour following his election in Texas, incoming Senator James Tallarico said, We have to crack down on these big data center companies taking advantage of Texas communities. We've got to make them pay for their own energy so they're not driving up utility bills across the state, make sure they recycle their own water, and ensure they get the consent of people who live in these communities before they start building. And it's not just me that's viewing that not just as political rhetoric, but as a bid that can be built upon. JP Powers of RWA Wealth Partners wrote, There's definitely some fears about what this might mean for the AI build-out, especially on data centers. And those names have already been under pressure because of the expected democratic victory. But I think it's probably a little bit overblown. It's going to put more onus on the states, but I don't think that will impact existing

projects all that much. It could result in higher cost of production if these companies have to provide maybe more local assistance or navigate more regulation across the board to get some of these projects through, but we still need these data centers. The Wall Street Engine account on X wrote, Gina Hinojosa may not be as bad for Texas data centers as it looks. Her plan asks for no subsidies, union labor, no NDAs, and proof household power and water bills won't rise. Much of that is already underway. And even more than individual points of analysis, outside of basically the neoclouds, it was far from doom and gloom yesterday on Wall Street. The S&P 500 saw its strongest single-day move since August, with the S&P 500 up 1.6% and the NASDAQ rising by 0.9%. Obviously, we need to see a little more green to call it a full-on trend reversal, but it does kind of look like a near-term bottom for the market. The conventional view is that markets hate uncertainty, and gridlock in Washington takes a lot of uncertainty off the table. However, even in the Reuters article that presented this viewpoint, they struggled to find many sources that actually believe we're in for a calm two years in Washington.

The next phase of the Iran War, a potential government shutdown in December, and interest rate hikes from the Federal Reserve all present the potential for major shocks to come. Seema Shah from Principal Asset Management said, We expect market volatility to continue. It will be a white-knuckled ride for the next year. And yet the biggest signal that many took from Wednesday's markets was a rise in big tech. The MAG-7 added a trillion dollars in market value, with Microsoft gaining 6.7%, Amazon up 6.1%, and Google gaining 3.3%. Now, part of that may be that markets are just taking a sigh of relief that the election is over, but there is increasingly an argument about how the hyperscalers could actually benefit from an AI slowdown. Remember that line from Citrini, perhaps the bottlenecks become the excuse some people need to quietly trim their own projections without causing a rush for the exits. And remember that heading into the election, the conversation has been nothing but when the AI bubble will burst. And in this case, the AI bubble that they're talking about is not just in the valuations

of the companies, but in the amount that they plan to spend on the AI data center build out. The question that many are now starting to explore is whether this presumption of a policy induced slowdown allows air out of that trend rather than having to wait for a full pop Again as Satrini pointed out there were already a lot of challenges to the build-out. Each of Google, Meta, Amazon, and Microsoft came in under their CapEx targets in Q3, flagging construction delays across their projects. Microsoft was specific, stating that a $5 billion miss on projections was down to a rented facility that wasn't ready on time. In September, Oracle gave notice that they might declare force majeure in relation to a New Mexico project that is facing power connection issues. And while in the aggregate, data center construction is still surging ahead, it's clear that the pace of construction is straining the supply chain, tapping out the labor pool, and putting stress on balance sheets. It's not crazy then to think that the hyperscalers might welcome an external reason to give them some cover to lower CapEx forecasts for next year. And it could even give them a PR win. The Hill reports

that hyperscalers are already in talks with environmental groups to draft a joint framework for data center development. And while some will certainly assume that this is cynical greenwashing, if a framework does come together, it could be an easy win for Democrats that ran on bringing the data center build-out to heel. Employee America Skanda Amarnath suggested the second-order benefit, commenting, I can attest to the fact that the group chat was discussing earlier today whether the hyperscalers would welcome political blocking of data center construction as a capital discipline coordination mechanism. And market analyst Matt Farley wrote, Slower AI CapEx is bullish, not bearish. Harder to build data centers mean GPU prices flatten, compute spot rents climb, CapEx falls, margins expand, FCF flips faster. Scarcity doesn't kill compute owners. It makes them rich and feeds investors what they wanted all along. Higher returns on capital while maintaining positive free cash flow. Can the constraint now be the moat? And was it always a race to getting the most infrastructure built the fastest before this inevitably happened? So instead of the doomsday scenario of data centers halting and the AI trade crumbling,

taking the rest of the market with it, instead, you've got the potential of hyperscalers taking advantage of the political and narrative moment to lower at least a bit their CapEx projections for the next couple of years, which makes markets happy for a variety of reasons, including an increase in the free cash flow of those stocks, and a potential bubble lets air out instead of having to pop. Jessica Lesson from the information made exactly this point. She wrote, The market reaction today points to the fact that investors have a very nuanced lens on the AI trade, which could be good news for those worried about an eventual huge crash. Rather than a big wipeout of everything today, we're seeing the market trade up a little, hyperscalers up a little, chips down, and neoclouds down. That suggests that the AI trade doesn't play just one way, which is probably good news for those who are concerned that weakness in some companies could pull the whole market down. Obviously, major changes in the macro, revenue growth halts, etc., could still collapse the whole pie, but the way things are trading now suggests a more nuanced picture. Ultimately, one historical truism on Wall Street is that it's very rare that a bubble actually features years of everyone correctly asking when a bubble will pop.

Historically speaking, bubbles tend to happen when most people don't see any risk and your Uber drivers are starting to hand out stock picks. But if things might not be so bad for the data center build-out and actually could have some benefits to a slowdown, what about broader AI policy? I talked yesterday about a lot of the very reasonable, low-hanging fruit, common-sense-type proposals that I think a big cross-section of politicians can get behind. Things like additional liability for the labs, disclosure of safety incidents, third-party auditing. But what about more severe measures like the Bernie Sanders proposal, which includes jailing anyone that develops AGI? Heading into the election, Abdul Saeed came out in favor of an AI pause. And at least some other progressives seemed to be getting on board. Arizona Congresswoman Yasemin Ansari wrote, top of the Democratic agenda on day one needs to be a pause on advanced AI development, regulation of this out-of-control industry, and a ban on artificial superintelligence. AI news aggregator Andrew Curran commented, This is probably going to be a huge part of political coverage for the next two years. The Democrats are going to push hard for an AI pause, new industry regulations, and probably a Bernie-style ban on superintelligence itself that exceeds the human level.

It will be interesting to see what the opposition ends up looking like. Many people on the right are also extremely anti-AI, so this will probably make for some strange bedfellows on both sides. Now, given how many other issues there are, I am somewhat less convinced that an AI pause inevitably becomes central to the progressive agenda. But let's say that it does. What would that mean for the industry in AI politics? Frankly, a legislated AI pause seems like basically a non-starter during this Congress. Trump is the most stalwart supporter of AI, or excuse me, superintelligence, or excuse me, SI, in the entire Republican Party at this point. And Cannon seems to be eager to veto anything Democrats passed, and I have a very hard time seeing enough Republicans switching sides to get enough votes to override a veto. And it's not like there's clear consensus. This week's election vote was not a unified endorsement of a single type of Democrat politics. It was a rejection of the GOP. But the folks that came into power represented the full range from moderates to the DSA. And even though they are now in power,

Democrats are going to have to do a lot of work to get to party alignment on some of these key issues. Semaphore reported that there's already a schism within the Democrats on how to best deal with this particular issue. Josh Gottheimer, one of the more centrist members of the caucus, commented, We agree that this technology is changing how Americans work and live. We also agree that we need common sense guardrails so that it works for families, workers, and kids, not against them. We'll continue to have constructive discussions on how best to get there. I look forward to working with every member, new and returning, to get it done. Despite being extremely not in favor of an AI pause as such, The fact that it's going to get broader political discussion doesn't stress me out at all. I think somewhere along the way, we forgot that the whole point of a democracy is to be able to have fierce debates in service of coming to the right answers. AI is getting more powerful with every single model release. This is indeed the one thing the accelerationists and the safetyists agree upon. Over the past year, the rest of America has been brought into that conversation. Now we actually get to have the debate. The debates which I will continue to argue have much more room for agreement than the extremes of either position would have you think. This is obviously something that we'll be spending

a bit more time on in the future. For now, however, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always, and until next time, peace!