GS Privorotsky GS Basics De Risking Solve Energy to Solve Rates Oil Inventories

GS Privorotsky - GS Basics De Risking Solve Energy to Solve Rates Oil Inventories 15 Sep 2026 Rich Privorotsky · Goldman Sachs · Managing Director, Global Banking & Markets 15 Sep 2026, 2:57am ET

Markets: Felt like a de-grossing event yesterday. Hardware fell about 500–800bps beyond what I expected, depending on the vertical. Electrical, data c

Markets: Felt like a de-grossing event yesterday. Hardware fell about 500–800bps beyond what I expected, depending on the vertical. Electrical, data center, optical… anything tied to the right tail of AI capex got obliterated. Meanwhile Meta +2.5% and Alphabet +2.7% as the market effectively said, “… maybe they don’t have to spend quite so much." Think market saying the highest incremental infrastructure growth assumptions are maybe too aggressive… and hardest hit where earnings/multiples relied most heavily on continued acceleration. On the flipside: Software went vertical. Software vs Semis +12.5%, Ai winners vs AI at risk -9.96%, 3m momentum +6%...High Beta TMT momentum -11.5%.

Asia: What makes it stranger is Asia this morning. SoftBank +7.5% (bouncing back), TSMC green and barely touched through the whole episode. Hard to square that with the idea the AI trade itself has fundamentally broken. Feels much more like repricing the right tail of infrastructure growth than AI demand itself.

AI: I think people are confusing an intention to slow frontier development for alignment with an intention to slow spending. I still see very little evidence this changes capex through 2027/28. We don’t have enough compute for the inference, RL and training workloads already coming, while hyperscalers are still raising, not lowering, spending intentions. Trump basically reinforced that yesterday… politically there is no desire to lose the AI race to China/open weights. The real issue is politics/permitting. Grassroots opposition is broadening and moratorium proposals are spreading. Even if hyperscalers don’t cut capex, you probably have to haircut the right tail of how quickly and where capacity can actually get built. That deserves a lower multiple for electrical, power, cooling and DC nuts/bolts exposures. But the derating has already been savage. On Blomberg blended 24m forward numbers SOX has gone from ~23x to ~13.85x… back toward April / peak oil shock levels. Hard to call the bottom, but there’s now a lot in the price. Beyond 2028 I want margin of safety. In some pockets the selloff gets circular… behind the meter power arguably becomes more valuable if grid access gets harder. Optical isn’t simply a frontier training trade either… more inference still means huge bandwidth/interconnect requirements. Treating all physical AI infrastructure as one trade feels too simplistic.

Rates + Energy: This is still the big one. Prices keep going higher (more Saudi/Houthi escalation overnight) and that is pushing yields, a toxic combination for risk as stagflation concerns rise. There are finally a couple of small potential release valves on energy. Trump asked Zelensky to stop hitting Russian refining infrastructure, while Ukraine has signaled openness to a reciprocal energy ceasefire if Russia is serious. Trump also said Iran “wants to make a deal quickly and desperately.” Iran pushed back,

unsurprisingly, but after weeks with basically no diplomatic signal, at least the tone has shifted slightly. But let me be categorical… I don’t think risk can properly work until energy stabilizes. Oil is still breaking out, alternative routes are insufficient and inventories are too tight. The marginal barrel sets the price. With diesel/product this elevated, the inflation problem broadens through freight, food and PPI. Yields then have very little reason to sustainably fall. Unless the military strategy suddenly works much better, the credible off ramp is diplomacy. We got small hints in that direction yday but so far no convincing signs. Think Xi/Trump meeting next week could be important.

Risk: I think the baby went out with the bathwater across large parts of hardware yesterday. At these levels risk/reward is materially better almost everywhere. Hard to reconcile SoftBank +7.5% and TSMC holding in with second/third…