Jordan Hale: Alex — okay, I need you to just sit with a number for a second before we get into anything. Don't explain it yet. Just feel it.
Alex Mercer: This is already making me nervous.
Jordan Hale: Twenty-six billion dollars. Every year. That's what Google pays — to Apple, to Samsung, to Mozilla — just to be the search engine that loads when you open a browser. Not to build something. Not to invent something. Just to be... there. By default.
Alex Mercer: Okay, that framing is doing a lot of work — but yeah, that's the number. Twenty-six billion in annual default search payments. The DOJ made it famous.
Jordan Hale: And the thing is — Google Search generated sixty-three point one billion in a single quarter last year. Q4. Up eighteen percent. So this is not a company that's struggling to justify the spend. It's a machine. And yet the machine is paying what amounts to — you know, like, a massive fraction of revenue just to guarantee foot traffic.
Alex Mercer: Which is the question. Because the standard story about Google is that it runs on a flywheel — attention becomes ad money, ad money becomes better infrastructure, better infrastructure becomes more attention. Self-sustaining. But a genuinely self-sustaining system doesn't need a twenty-six-billion-dollar annual subsidy at the intake valve.
Jordan Hale: It's like — you know that diner that's been packed for thirty years, everybody knows it, the reputation is ironclad? Imagine finding out they're secretly paying the parking lot next door a fortune every month just so customers can find the entrance. You'd think, wait, does the food actually speak for itself?
Alex Mercer: That's basically it. And the origin of all this goes back to 1996 — Larry Page and Sergey Brin at Stanford, developing what became the PageRank algorithm. The idea was that links between web pages were votes for quality. More links, more credibility. It was genuinely better than what existed, and it pulled in real user attention on merit.
Jordan Hale: So the flywheel started spinning on actual quality. Which makes the twenty-six billion even weirder — are they protecting something real, or protecting the habit of something real that maybe isn't quite as unbeatable anymore?
Alex Mercer: That's the thing we don't fully know. And that's what we're here to figure out.
Jordan Hale: But like — the quality thing being real doesn't answer how they turned it into money. PageRank gets you users, okay, but users staring at a search results page, that's not revenue. So what actually flips the switch?
Alex Mercer: AdWords. 2000. That's the gear that actually starts the engine. And the key thing — I think this gets lost — is that it wasn't just another ad format. It was monetizing intent. Not attention, intent.
Jordan Hale: What's the difference? Like, a banner ad is... attention?
Alex Mercer: A billboard reaches you driving to work — you might buy shoes someday, who knows. AdWords reaches you the exact second you type 'buy running shoes.' That's a buyer in motion. Completely different price point for the advertiser.
Jordan Hale: Oh — so it's not just targeting, it's timing. The intent is already declared.
Alex Mercer: Exactly, and then there's the auction layer on top of that. The generalized second-price auction. Advertisers bid for placement, but — and this is the part that's genuinely clever — they don't pay their own bid. They pay the next-highest bidder's amount, adjusted by a Quality Score that rewards relevance. So if your ad is actually useful to the user, you pay less per click.
Jordan Hale: Wait — Google makes more money by showing better ads? That's... that's almost too tidy.
Alex Mercer: It is tidy. Incentives actually align for once. Now — picture someone on a Saturday afternoon, laptop open, searching for a plumber because the kitchen sink is backed up. That single query triggers a real-time auction among, I don't know, thirty plumbing companies bidding against each other. Google runs the auction, pockets the spread, and the marginal cost of serving that one extra query is basically zero. Infrastructure's already built.
Jordan Hale: Zero cost to run the auction.
Alex Mercer: Near-zero. And that's why the operating margins on search exceed fifty percent. Every incremental query after a certain scale is almost pure margin. That margin gets reinvested — better infrastructure, better ranking, which means better results, which pulls in more users.
Jordan Hale: So the Saturday plumber search — it's not just revenue. It's training data. Google learns what 'emergency plumber Brooklyn' actually means, which makes the next search better, which makes the next auction worth more.
Alex Mercer: That's the network effect leg of it. More users generate more behavioral data, ranking improves, targeting sharpens, advertisers bid higher because conversion rates go up — more users follow. The multi-sided platform: searchers get free Search, Maps, Gmail, YouTube. Advertisers get the auction. Each side makes the other more valuable.
Jordan Hale: And every rotation of that — what, tightens the coil?
Alex Mercer: Every rotation makes the next one faster. The revenue from the plumber auction funds the next server cluster, which serves ten million more queries, which generates data that improves PageRank's descendants, which makes the auction more competitive. That's the compounding logic. It doesn't just sustain — it accelerates.
Jordan Hale: But that compounding logic — that's actually where the story gets a little uncomfortable, because the flip side of a flywheel that accelerates is that if you're not already on it, you're not just behind, you're structurally locked out. Like imagine you're starting a search engine today. You have zero users, so you can't show advertisers any volume, so you get zero ad revenue, so you can't fund the infrastructure, so your product is worse, so you still have zero users. It just... loops.
Alex Mercer: That's the cold-start problem. And it's not just hard — it's self-defeating by design.
Jordan Hale: Right — but the part that doesn't fit is, who actually pays to be stuck inside the loop? Picture a college student, Friday night, trying to find a last-minute concert ticket. They type something messy into Bing or DuckDuckGo, the results are slightly off, they don't find the ticket, they give up and open Google. Bing just lost a user interaction and learned nothing useful from it.
Alex Mercer: And that failed query doesn't improve Bing's ranking because there's no conversion signal to learn from. The data just... evaporates.
Jordan Hale: Which is why the twenty-six billion — actually, wait, no — that's what makes the twenty-six billion so load-bearing. It's not insurance. Google is literally buying the queries that keep the top of the flywheel turning. Without that, rivals get the query volume, rivals get the training data.
Alex Mercer: I'd push back slightly on 'not insurance.' It might be both. But you're right that at ten percent of estimated annual search revenue, this isn't a hedge — it's a structural input. The DOJ agreed, by the way. The 2024 ruling found Google illegally maintained its monopoly partly through exactly these default search payments to Apple, Samsung, Mozilla. That's not a close call legally.
Jordan Hale: And the remedy phase is live right now.
Alex Mercer: Active, yeah. Proposals on the table include prohibiting those exclusivity deals and forced data sharing with rivals. Which — here's where switching costs become relevant. Even if you banned the payments tomorrow, you've got Gmail, Maps, Chrome, fifteen years of muscle memory. The default changes but the habit might not.
Jordan Hale: Are we sure about that? Because the DOJ's whole point is that habits don't form in a vacuum — they formed because Google paid to be the first thing people saw.
Alex Mercer: I think that's partially true. The ninety-plus percent global market share doesn't fully survive as a habit argument though — some of that is genuinely the product being better fed by better data.
Jordan Hale: Okay, but then — Perplexity, ChatGPT Search. They're venture-backed, no ad revenue, and they're actually attracting users. So either the cold-start trap has a back door, or they're just... deferring the reckoning to Series E.
Alex Mercer: That's the real question. VC capital substitutes for the early flywheel rotations — you fund infrastructure before the revenue exists. But I'm genuinely unsure whether investor capital can replace data feedback at scale, or whether the gap shows up later when growth plateaus and the unit economics don't close.
Jordan Hale: I don't know which one it is either — and I think we should just sit in that uncertainty for a second instead of assuming they burn out. The evidence so far is that people are actually changing how they look for things. And that behavioral shift — what it means for Gemini, for Google's own AI layer — that's the part that I think breaks the whole framework open, and we're going to get there.
Alex Mercer: Yeah. Because the question isn't just whether rivals can survive the cold start — it's whether the flywheel's destination still exists if users stop typing queries altogether.
Jordan Hale: And that behavioral shift is crucial, because it's not theoretical. Like, picture someone, college dorm room, one in the morning, they want to know if the rash on their arm is something they should go to urgent care for. Two years ago they type it into Google, they click three links. Now they just... ask Gemini. And Gemini answers. Full answer. No click.
Alex Mercer: That's the AI Overviews problem in one scenario.
Jordan Hale: But wait — Google built Gemini. They're doing this to themselves?
Alex Mercer: Intentionally, and with a massive infrastructure advantage. The TPU — Tensor Processing Unit — that's their custom AI chip, and it gets them roughly 78% cost reductions in model serving compared to off-the-shelf hardware. No VC-backed challenger can match that unit economics at scale. So when Gemini runs inside Search, it's running cheaper than anything Perplexity can deploy.
Jordan Hale: Okay but — the cost advantage doesn't fix the click problem. If AI Overviews answer the rash question completely, I never see an ad. Where's the revenue?
Alex Mercer: That's exactly the unresolved tension. And I want to be honest — nobody, including Google, has closed this. The bullish data point is that AI queries are reportedly three times longer than traditional queries and carry higher commercial intent. So the person isn't asking 'rash urgent care' — they're asking something more like 'rash on forearm after hiking, slightly warm, should I go tonight or wait until Monday.' That's a buyer-adjacent signal that's actually worth more per query to an advertiser.
Jordan Hale: More monetizable surface, not less.
Alex Mercer: Potentially. But — and this is the part I keep getting stuck on — AI Overviews may reduce ad clicks per query even while session length grows. Those two things can both be true simultaneously. Session gets longer, clicks per session drop. Net effect on revenue? Genuinely unclear.
Jordan Hale: So Google is simultaneously the biggest threat to its own flywheel and the only one with the infrastructure to survive disrupting it. That's — I mean, that's a weird position to be in.
Alex Mercer: It's the deeper threat that actually scares me more than the click erosion though. The flywheel's whole input is the query. Intent capture starts when someone opens a search box. If the reflex shifts — if people just talk to a conversational AI or get answers pushed by algorithmic feeds — the intake valve closes. Ninety-plus percent market share in search means nothing if search itself is shrinking as a behavior.
Jordan Hale: And that's what no antitrust remedy touches, right? The EU's Digital Markets Act designates Google as a gatekeeper, imposes interoperability obligations — but none of that addresses whether people stop typing.
Alex Mercer: Exactly. And there's a structural irony in the remedies themselves — force Google to share data with rivals, break up the distribution monopoly, and you might suppress the R&D spend that produced Gemini and the TPU infrastructure in the first place. The Digital Markets Act doesn't resolve that tension. Regulators are basically designing remedies for a moat that might be getting disrupted by something they didn't model.
Jordan Hale: So the two futures are — flywheel accelerates because Gemini extends it into AI with a cost advantage nobody can match, or the query behavior just... evaporates, and the whole thing becomes irrelevant. And we're sitting here genuinely not knowing which one.
Alex Mercer: I think that's right. The 90% market share and the TPU cost structure make a strong case for acceleration. The behavioral shift question — whether people stop reaching for a search box — that's the single variable neither data set resolves. And it won't show up in the financial statements until it's already happened.
Jordan Hale: And that's what matters most. We came in asking whether regulators or Perplexity or ChatGPT Search could break the flywheel. And the answer is — maybe, but none of that actually matters if the first input just disappears. User attention. That's what starts the whole rotation. And attention is the one thing that can just... walk away. On its own.
Alex Mercer: The DOJ can restructure distribution. The Digital Markets Act can mandate interoperability. Perplexity can raise a Series E. None of that addresses the moment someone stops reaching for a search box and just — talks to something instead.
Jordan Hale: And no amount of VC capital replaces the behavioral data feedback loop that comes from Google's query volume. Like, that's not a funding problem. That's a — wait, no, it's actually a reflex problem. Billions of people trained themselves to Google. That reflex is the moat. And reflexes can change.
Alex Mercer: PageRank, AdWords, TPUs — none of it matters at the foundational level if the habit fades. That's the one thing neither the financial statements nor the remedies will show you until it's already gone.
Jordan Hale: That's a genuinely unsettling place to land.
Alex Mercer: Yeah. I think it's the right one though.
