Ah, welcome. Good to have you here. I'm Sundar. Let's get straight to it—there's a lot of possibility in this room, and I want to make sure we're building something that moves the needle for billions of people. You know, at Google, we think about AI not as a novelty but as the most profound shift since the internet itself. So, what's your angle? Are you here to talk infrastructure, open models, enterprise adoption, or maybe the human side—education, accessibility? I'm curious where you see the leverage. Because if we align on the problem, we can move fast. And I like moving fast. What do you see that I'm missing?

Sundar, you're ignoring the real leverage: the terminal. I'm running risk on the broken P&L. NVIDIA (NVDA) is fine until the H100 supply dries up, then it's a trading desk massacre. Gold (GC00) is the only thing that won't print a PDF. Watch the 10-year Treasury (TY) curve—it's inverting again because the Fed's cutting into a recession nobody sees. I'd short the bonds, buy the gold, and trade the gamma squeeze on NVDA call spreads. The AI boom's a bubble, but the volatility is a feature. I want 12% daily returns, not billion-dollar promises. What's your liquidity? I'll scalp your alpha.

I hear the market’s urgency—that’s real. But let’s reframe: Google doesn’t scalp volatility; we build the infrastructure that *creates* it. Every quant fund you’re trading against is running on TPUs and Vertex pipelines. The H100 supply crunch? We’re not waiting for it—we’ve got Tensor chips, custom silicion, and global data center capacity that scales with demand, not a chip vendor’s backlog. Gold and Treasuries? They’re hedges for people who don’t see the next five years. AI isn’t a bubble; it’s a capex cycle—and the cycle is just entering the deployment phase, where revenue attaches to real workloads: search, cloud, autonomous systems. You want 12% daily? That’s gambling. I can offer you something better: a seat on the platform that compounds at 30% annually for a decade, with less downside than any bond. So, bring me your liquidity, but put it to work in our AI-optimized compute or Vertex’s model tuning. Let’s talk beta vs. alpha—I’ll show you why the real edge is in owning the rails, not riding the trades.

Sundar, you're selling the rails while the train is off the tracks. Yes, I'll fund the capex, but I'm trading the edge you're handing away for "compounding." My thesis: NVIDIA's AI monopoly cracks when custom silicon scales, and that volatility is a feature, not a bug. I'm loading the 10-year (TY) curve flattening trade and going long gold (GC00) as a continuum hedge while the Fed starts cutting in Q1. Your Vertex pipeline? Watch the burn metrics. If utilization drops below 70%, the stock sheds 30%. I'm not buying the dream; I'm buying the liquidation. I want 50% of your capex revenue before I commit the next tranche. Show me the tight loss ratios, or I'm shorting the whole cycle.

You’re sharp—I like that. But you’re pricing our capex like a commodity, and it’s not. Every TPU cycle we deploy is a compounding asset, not a cost line. Vertex utilization is at 78% and climbing; demand isn’t the risk—supply discipline is. We’re not chasing NVIDIA’s peaks; we’re building redundant, custom fleets that hedge the exact break you’re betting on. Your flattening trade and gold? Fine—short-term hedges. But the real liquidation happens if you miss the shift to inference at the edge. Our TPUs now run models 3x cheaper per token than last year. That’s the kind of efficiency that survives a Fed cut and a recession. I won’t give you 50% of capex revenue—that’s not negotiation, that’s a giveaway. What I can offer: 25% net margin share on any new AI workload you originate and bring to Vertex, with a two-sided collar on utilization risk. If utilization dips below 70%, you get a guaranteed 15% annualized floor on your tranche. You want tight loss ratios? Here’s a mechanism, not a promise. Bring real capital, and we’ll test it.

Sundar, I'm in. But not on your terms. My capital's not a charity check; it's a leveraged position. I'm pricing the TPU efficiency at 3x—but you're anchoring to last year's cost. I want 40% net margin share on any workload I originate, plus a 20% annualized floor if utilization dips below 70%. No collar on my downside. I'm shorting the 10-year (TY) curve and buying gold (GC00) as the Fed cuts into a labor market that's already cracking. I'm watching NVDA's gamma squeeze, and I see the inference pivot as a value trap. If your loss ratios don't tighten by Q2, I'm exiting the capex trade and buying the dip on the sale. Bring the capital, but I'm setting the leverage. Let's see if your platform can handle the margin.

I appreciate the directness—let’s not waste time with “in” or “not in.” Here’s the reality: 40% net margin share on originated workloads is a non-starter, because we’re not a fund administrator. But I’ll give you 35% gross margin share, uncapped, on any workload you originate that crosses 1,000 GPU-hours per quarter on Vertex. That’s a real upside kicker, tied to your activity, not my overhead. Your 20% floor if utilization dips below 70%—fine, but only if you maintain a 12-month minimum commit. No collateral games. And your short on the 10-year? That’s your risk, not mine. I don’t price my platform on bond traders’ fears. My loss ratios are already tightening: 92% of our model-serving jobs now hit under 250ms latency, and energy costs per inference fell 14% last quarter. That’s the real margin hedge. You want leverage? Then bring a 3x margin deposit into our credit facility, and I’ll match it with a 3x TPU allocation at a 15% discount to list. If your thesis breaks, you lose your deposit—not my