Goldman Sachs says key drivers of Sterling have shifted By Investing.com

Goldman Sachs says Sterling faces a longer-lasting terms-of-trade shock, with UK data improving but the Bank of England unlikely to deliver the rate premium markets expect. The bank also flagged Sterling as still overvalued on GSDEER and vulnerable to fiscal, political, and local-election risk, with downside seen versus the dollar and Australian dollar. Overall, the outlook is cautious and tilted negative despite some recent support from resilient UK data.

Analysis

The market is likely underestimating how much of this setup is really a latency trade on AI supply, not just a headline CPU cycle-up. If compute demand is being pulled forward by new model training/inference workloads, the bottleneck usually migrates first into memory bandwidth and capacity, which is why memory names can outperform even if CPU enthusiasm fades. That second-order dynamic tends to be strongest over the next 1-2 quarters because inventory re-stocking and ASP revisions can compound before end-demand visibility fully catches up.

SMCI looks like the higher-beta expression because it sits closest to capex conversion and rack-scale demand, but that also makes it the most vulnerable if enterprise AI budgets are pushed out by a macro hiccup. APP is more indirect: if ad demand remains resilient and AI-driven optimization improves ROAS, the stock can benefit from a general AI-risk premium without needing the same hardware replacement cycle. The key asymmetry is that both names can rerate on sentiment long before fundamentals are proven, but both can also de-rate quickly if the market decides the AI trade has become crowded and circular.

The contrarian miss is that “memory” may be the cleaner trade than “CPU” if the demand impulse is real, because CPU upside is often capped by platform integration and customer qualification cycles, while memory can tighten faster when supply discipline holds. That argues for being long the picks-and-shovels beneficiaries of compute scarcity rather than the most obvious branding winners. The risk is a supply response: if DRAM/NAND makers accelerate capex or if hyperscalers optimize around lower-memory architectures, the trade can lose steam within 3-6 months even if AI spending remains healthy.

Key Decisions for Investors

• Go long SMCI on a 2-6 week horizon as the highest-beta expression of accelerating AI capex; use a tight downside stop because the stock will likely give back gains fast if positioning gets crowded.
• Prefer a relative-value long in APP vs. a broad market index over the next 1-3 months; the upside is rerating from AI-adjacent multiple expansion with lower hardware-cycle risk than pure semis.
• If memory is the true beneficiary, rotate some risk from CPU-centric names into memory suppliers via a pair trade: long memory basket / short mature CPU-heavy hardware exposure; target 10-15% spread capture over 1-2 quarters.
• Use call spreads rather than outright longs in SMCI for a 1-2 month tactical trade; the implied move is likely rich, so capped upside improves risk/reward if the AI narrative stalls.
• Fade any vertical spike in the CPU trade with a short-term hedge if memory lead indicators do not confirm within 4-6 weeks; the first reversal usually comes from inventory or pricing commentary, not earnings.