UK wage growth slows to 3.8% in 3 months to January

Regular (ex-bonus) annual wage growth slowed to 3.8% in the three months to January versus a 4.0% Reuters consensus. A jump in energy prices after the start of the Middle East war has introduced new inflationary pressure, reducing the likelihood of a 25bp Bank of England rate cut and leaving the BoE expected to keep borrowing costs on hold at the March MPC meeting. Implication: a more hawkish BoE outlook supports gilt yields and may bolster sterling versus the view of imminent easing.

Analysis

Large OEM demand for high-end accelerators is shifting the microeconomic equilibrium in the AI stack: supplier-side scarcity is increasingly the gating factor for feature rollout timelines, not just software engineering. That gives leading GPU vendors durable pricing power for the next 6-18 months, and forces hyperscalers to either accelerate capex or accept longer project timelines — both outcomes expand near-term revenue visibility for fabs and OSATs even if end-user demand is lumpy.

For OEMs integrating external accelerators, the second-order margin effect is non-trivial: buying top-tier GPUs to shortcut internal ASIC development reduces near-term R&D-to-revenue conversion time but raises COGS and unit-level power costs, pressuring gross margins unless offset by subscription monetization. This tradeoff compresses the optionality of in-house silicon roadmaps and increases strategic value of software lock-in (frameworks, toolchains), which creates higher switching costs for OEM customers over a 2–4 year horizon.

Key risks live at the intersection of geopolitics, capacity and macro: export controls or a supply shock at fabs could flip pricing power into order cancellations within 60–120 days, while a coordinated hyperscaler capex pause would mute fabs’ forward bookings over 6–12 months. Conversely, a visible guidance beat from a dominant GPU supplier would likely re-rate the whole compute supply chain quickly; monitor OSAT lead times, fab utilization and hyperscaler inventory adjustments as high-frequency signals.

Key Decisions for Investors

• Long NVDA via a 3–9 month call-spread: buy a near-ATM 6–9 month call and sell an upper strike 20–30% out to finance premium. Rationale: captures upside from continued enterprise OEM procurement and potential guidance beats while capping time decay; target 30–80% return if supplier pricing/allocations tighten, max loss = premium paid.
• Pair trade — long NVDA / short AMD (or long NVDA / underweight other GPU-exposed names) for 3–12 months: capture asymmetric benefit from NVDA’s software ecosystem and OEM pull while hedging cyclical GPU demand risk. Risk/reward: expect 1.5–2x relative outperformance if NVDA sustains premium ASPs; tail risk is shared supply shock or broad AI capex drawdown.
• Convex TSLA exposure via long-dated LEAP calls financed by selling 3–6 month calls (calendar sell): play optionality that faster external-accelerator adoption shortens time-to-revenue for autonomy software. Timeframe 12–24 months; reward if monetization cadence accelerates (2–4x on option notional), risk = regulatory/autonomy setbacks and higher near-term dilution from capex increases.
• Risk management: set alerts on three high-impact triggers — (1) fab utilization moves ±5 percentage points versus current baseline, (2) announced export-control changes or license denials, (3) hyperscaler inventory revisions. If any trigger fires, cut option exposure by 50% within 7 trading days to preserve asymmetry.