Bulls vs Bears EP117 Datacentres Baby!
The verdict? If you're not building data centres, you're falling behind — and the market is making that very clear. This week the team breaks down Google's blockbuster earnings (spoiler: it's all about the infrastructure), digs into Meta's AI talent war and its ongoing $80B…
Google: The Slow and Steady Winner The week's dominant story was Google's earnings, and the team wasted no time making their position clear — Google is "the greatest monopoly in existence," and this result proved it. What makes the result so significant isn't just the numbers, but what they signal about the AI race more broadly. While competitors like OpenAI are scrambling to secure compute from Amazon, and Elon's xAI is building out its Colossus cluster, Google already owns the entire stack. Their custom TPU hardware, built in-house the same way Meta builds its own silicon, means compute cost is structurally lower for them than for almost anyone else in the game. Kai pointed out something often overlooked: Google didn't just benefit from AI — they invented the underlying technology. The transformer architecture that every major language model is built on came out of Google Research. And with Demis Hassabis — Nobel laureate and the mind behind AlphaGo — running DeepMind, the team argued that if a genuine breakthrough in artificial general intelligence is coming, it's most likely to come from Google or Elon. The market thesis is now simple: if you're spending on data centres, you deserve a higher multiple. The re-rating is underway. Data Centres Are Holding Up the US Economy One of the more striking data points of the episode: strip out data centre capital expenditure, and the US economy would actually be contracting right now. The infrastructure build-out isn't just a market theme — it's a macroeconomic life raft. JT flagged the US's structural energy advantage here. Texas's natural gas fields, courtesy of the fracking boom, give American hyperscalers a cheap, abundant power source that most of the world can't match. The playbook, as one energy analyst put it: skip the nuclear plant, put a gas turbine next to the oil field, and drop the data centre right beside it. Moving data between states is trivial. Moving power isn't. The conversation also touched on what comes next. Jensen Huang is apparently already thinking about space-based data centres — less cooling overhead, and Elon's reusable rockets make the logistics more conceivable than they once were. Separately, the team discussed how zero-gravity manufacturing in space could produce superior fibre-optic cable, free of the oxygen bubbles that degrade signal quality on Earth. We are, it seems, officially in the era of long-range infrastructure speculation. One underappreciated constraint on the whole AI build-out: water. Every standard Google search consumes roughly a bucket of water in cooling. An AI query consumes around ten times that. At global search volumes, the numbers become staggering — and it's a resource constraint that almost nobody is talking about seriously yet. Meta: Big Spender, Uncertain Returns While Google was celebrated, Meta had a rougher reception. The team acknowledged the scale of Zuckerberg's ambition — $145 billion committed to AI capex this year — but questioned whether the company was actually getting results commensurate with the spend. Part of the problem traces back to Yann LeCun, one of the three "godfathers of AI" and Meta's long-standing Chief AI Scientist. LeCun was publicly sceptical of large language models for years, arguing they weren't a path to true artificial intelligence. That philosophical resistance put Meta on the back foot as the LLM wave accelerated, and the company has been playing catch-up ever since. Enter Alex Wang from Scale AI — now heading Meta's AI division at eye-watering compensation. The rationale for the hire, Kai explained, goes deeper than talent. All the internet's human-generated content has essentially been consumed by the major models. The next frontier is synthetic data: AI-generated training material that can extend learning beyond what humans have recorded. That's what Scale AI specialised in, and it's why the acquisition made strategic sense. The Metaverse was also revisited — briefly, mercilessly. "It could have been vibe-coded in a day," one host offered, referencing the legless floating avatars that became the emblem of the $80 billion pivot. Meta hasn't changed its name back to Facebook, which the team took as evidence that admitting mistakes publicly remains off the table. Instead, Zuckerberg has simply switched the narrative — same wallet, new destination. Oil: What the Futures Curve Is Telling You JT brought a more nuanced read on crude oil than the headlines suggested. Yes, oil had a strong push early in the week before pulling back about $5. But the more important signal was in the back months of the futures curve. At the start of the Middle East conflict, only the near-term contracts (March, April, May) spiked — the back months (October, December) barely moved. The market was pricing in a short disruption. This week, when front months ran, September and October moved with them. December didn't break to new highs, but it pushed to the top of its range. The interpretation: the…
Read on MPC Markets Mosaic
