Macro Matters: Rockefeller’s Chang on AI, Rates and Bond Markets

Rockefeller Global Family Office CIO Jimmy Chang discusses a market in which accelerating AI capital expenditure, rising energy costs and Treasury yields at multi-decade highs have left few assets appearing obviously cheap. The discussion frames AI investment as a potential support for the economy, while elevated rates and valuations remain key constraints for investors.

Analysis

The relevant transmission mechanism is not broad “AI bullishness,” but the widening spread between companies that can monetize incremental compute and those financing compute-heavy expansion with long-duration cash flows. MSFT, GOOGL, AMZN and META can absorb elevated capex through existing high-margin cash generation; smaller AI infrastructure buyers and leveraged data-center developers face a materially higher hurdle rate. This favors hyperscalers over second-tier cloud, software names whose valuations still assume rapid AI revenue conversion, and highly levered digital-infrastructure vehicles such as certain data-center REITs.

Power availability is becoming a binding constraint rather than simply an input-cost issue. Regulated utilities with transmission exposure and credible rate-base growth—including CEG, VST, PWR and ETN—should retain earnings visibility, while merchant generators without contracted capacity may see more volatile upside. The second-order risk is that grid interconnection delays defer data-center revenue recognition, causing AI-capex beneficiaries to incur depreciation and interest expense before utilization ramps.

Near term, crowded AI leadership remains vulnerable to a real-yield shock because a larger share of expected value sits beyond the next two years. Over 1-3 months, the key catalyst is whether hyperscaler results demonstrate AI-related revenue and utilization rising alongside capex, rather than another capex-guidance step-up. Over 6-18 months, the structural winners should be power equipment and grid-service providers, but the thesis is falsified if utility load forecasts or large data-center project pipelines are delayed, canceled, or fail to secure power contracts.

Key Decisions for Investors

• Maintain a quality-AI barbell: long MSFT and GOOGL versus a basket short of unprofitable/low-visibility software via IGV, sized as a 1-3 month relative-value trade. The thesis is that self-funded platforms can sustain investment while duration-sensitive software multiples compress if real yields rise; exit if IGV materially outgrows hyperscaler AI revenue guidance in the next earnings cycle.
• Accumulate PWR and ETN on market weakness for a 6-18 month grid-buildout exposure; use a 10-15% downside risk limit from entry. Their opportunity is tied to transmission, electrification and data-center interconnection spend rather than a single model vendor’s capex cycle; reduce if backlog growth or utility capital plans decelerate for two consecutive quarters.
• Avoid initiating broad data-center REIT longs until financing, contracted-power, and preleasing data are verified. For existing exposure to EQIX or DLR, hedge duration risk with a partial IYR short over the next 1-3 months; the hedge is no longer needed if long-term yields decline materially while leasing spreads and development yields improve.
• Do not add outright index exposure solely on the AI narrative at current uncertainty. Treat the next hyperscaler earnings cycle as the decision point: raise AI exposure only if incremental capex is accompanied by accelerating cloud/AI revenue, stable operating margins, and no deterioration in free-cash-flow guidance.