GS Privorotsky GS Basics Inference Economy Real Rates Corn Beans

GS Privorotsky - GS Basics Inference Economy. Real Rates. Corn & Beans 22 Sep 2026 Rich Privorotsky · Goldman Sachs · FICC & Equities Tue 22 Sep 2026, 3:15am ET

Markets: Massive Nasdaq rip yesterday. Muse has lit a fire under the CPU chain. ARM +17%, Intel +12%, AMD +10%. Muse is now the No. 1 free U.S. iPhone app and the market has very quickly extrapolated the implication: if a useful personal agent requires a few CPU cores, RAM, storage and intelligence running persistently in a VM, then multiply that by millions and eventually hundreds of millions of agents...Nasdaq +2.3%, SOX +4.3%. Curiously, this hasn't spilled cleanly across the entire AI complex. Asia is better this morning, but the Kospi faded from its early highs (Hynix small down)...TSMC down. This still feels much more nuanced than buy everything AI. Inference is not exactly the same trade as frontier training. Maybe the next leg isn't simply a handful of labs spending trillions building ever larger clusters. Maybe it's billions of dispersed agents, that doesn't kill the frontier trade. it broadens what AI infrastructure demand looks like.

Inference Economy: And I think we need to go much broader than CPUs...sure bullish CPU, a bit memory, storage, networking and power. but it cuts deeper. The easiest framework might be rails vs interfaces. An agent can destroy value in the human facing interface while creating more activity underneath it. I may stop visiting a travel website, but the agent still needs inventory, booking rails and payment infrastructure. My deposits still need custody and transactions still need settlement. Systems of record probably become more important even as some systems of engagement become less relevant. The flipside is enormous. Anything earning excess returns from friction, complexity or inertia should get nervous. Wrong telecom plan. agent switches it. Flight delayed. refund filed. Cash earning too little. move it. Insurance renewal too expensive. re-shop it. Subscription unused. cancel it. Why would I stay in a bad product if an agent is continuously scanning the world for a better one? Comparison sites, affiliates, lead generators, basic brokers and businesses whose moat is basically "this is annoying enough that you won't bother" all look structurally challenged. Advertising: value potentially migrates away from capturing human attention toward being selected by the agent. Whoever owns the default personal agent incredibly valuable (Meta?). Whoever sits between the consumer and the internet gets an enormous distribution advantage (Meta?). What does that mean for labor? Think Gerstner got it right...more likely top lines keep growing while businesses discover they simply need fewer incremental people. Less hiring/less firing economy. That's where we've been and where we'll continue is how the productivity ultimately gets realized.

Rates: This remains the weird bit. Oil fell hard and the nominal curve bull flattened, but the front end barely participated. 2Y yields fell just 0.8bp versus 2.9bp in 10s. Inflation expectations came down with energy, yet the market still wants a lot of real rate compensation at the front/belly. Maybe that is simply a reflection of an incredibly strong nominal growth backdrop. story would be cleaner if Russell/cyclicals were confirming harder. Maybe some of it is supply. Treasury issuance plus hyperscaler funding increasingly hits the front and belly rather than duration. If front end and belly rates genuinely start

diverging from falling energy, that becomes a bigger problem. The market would effectively be saying the economy is running too hot regardless of oil. Watch this closely. falling energy should be giving rates more relief than this.

Energy: Talks remain alive and nothing is getting blown up for the moment, but there is still no confirmed breakthrough. Trump has signaled openness to meeting Iranian President Pezeshkian around UNGA and Iran is in New York, while crude is bouncing this morning after four straight down days. I'm still cautiously optimistic that the incentives ultimately point toward some diplomatic solution. Stakes are enormous and economically everyone should want a deal. The…