Jefferies’ Wood Sticks With ‘Picks and Shovels’ Trade

Jefferies Global Head of Equity Strategy Chris Wood continues to favor the AI "picks and shovels" trade, contingent on sustained capital-spending growth. Wood also argues that China’s large domestic market provides a competitive advantage in the AI cycle, supporting a constructive view of AI infrastructure beneficiaries and Chinese technology exposure.

Analysis

The investable implication is not Jefferies itself but whether AI infrastructure spending is broadening beyond hyperscaler commitments into a durable enterprise and sovereign-capex cycle. The highest operating leverage remains with bottleneck suppliers—NVDA, AVGO, ANET, VRT, ETN and TSM—while their valuations now require 2026 demand visibility rather than another upside quarter. Over the next 1-3 months, cloud-provider capex guidance and lead-time commentary matter more than generalized AI optimism.

China's scale can support domestic AI adoption, but it does not automatically translate into investable profit pools for offshore investors. Export controls constrain access to frontier accelerators and may push local workloads toward lower-performance, higher-power domestic alternatives; this favors SMIC and Chinese server-chain exposure only if policy support offsets weaker unit economics. A second-order beneficiary is power infrastructure: higher compute intensity lifts demand for electrical equipment and cooling regardless of which model vendor wins, making VRT/ETN a cleaner way to express the thesis than concentrated semiconductor beta.

Consensus risk is that capex remains elevated while monetization lags, converting an apparent secular boom into a 2026 return-on-investment debate. This would pressure the most expensive accelerator and networking names first, whereas power and data-center infrastructure retain backlog support. The thesis is falsified by a material reduction in MSFT, AMZN, GOOGL or META capex outlooks, falling AI-server lead times, or evidence that data-center power constraints are delaying deployments rather than accelerating infrastructure orders.

Key Decisions for Investors

• Maintain a 3-6 month barbell: long VRT and ETN versus short a basket of higher-multiple AI application software (IGV) to capture infrastructure spend while hedging monetization risk. Reassess if hyperscaler aggregate 2026 capex guidance falls more than 10% or VRT backlog conversion weakens.
• Prefer ANET over NVDA for incremental AI exposure into the next earnings cycle: networking content rises as clusters scale and valuation sensitivity to a single accelerator roadmap is lower. Use a stop if management signals cloud titans are digesting capacity or if AI-related revenue growth decelerates materially for two quarters.
• Do not establish a directional JEF position from this commentary; the direct earnings linkage is immaterial. Treat JEF only as a sentiment read-through unless investment-banking pipeline data show a sustained recovery in technology underwriting and M&A.
• Watch SMIC and KWEB rather than chase Chinese AI beta immediately. Upgrade to a tactical 6-12 month long only on verifiable domestic accelerator order growth, improved utilization, and policy clarity; export-control tightening or weak power availability would invalidate the local-substitution case.