Honeymoon’s Over | From Spend to Return

Four major hyperscalers reported earnings April 29, with investor reaction split as updated revenue and CapEx outlooks sharpened focus on AI returns. GOOGL surged on strong Google Cloud results, even as chip constraints capped sales, a sign its CapEx is translating into AI-driven demand and revenue growth. META, by contrast, saw investor pullback after raising its 2026 CapEx guidance, underscoring concern about heavier AI spending without a clearer near-term payoff. Based on disclosed plans, 2026 capital spending by GOOGL, AMZN, META and MSFT totals roughly $695 billion to $725 billion, up from prior high-end expectations of about $670 billion.

Hyperscaler CapEx remains a key driver of data center growth. Enverus Intelligence^® Research (EIR) forecasts load growth of 11% to 19% by 2035 across scenarios ranging from 2% to 7% CAGR (Figure 1), with data centers accounting for a substantial share. The buildout era is maturing, and capital alone is no longer enough to move markets.

Research Highlights:

• Planning Under Uncertainty – Market Fundamentals and Investment Decisions – This Enverus EVOLVE 2026 presentation frames U.S. power-market planning as a scenario problem, not a single forecast. We build bottom-up load cases driven by data centers and EVs, then link demand paths to investment outcomes by jointly modeling energy, capacity and REC markets across policy and cost levers — reducing mis-sized builds, mispriced PPAs and missed capacity value.

• Valuing the Stack – Drivers of Asset Economics and IPP Valuations – This presentation from Enverus EVOLVE 2026 examines the shifting economics of power generation and its impact on asset and IPP valuations. We explore how merchant price curves, ancillary revenues and contract structures are reshaping returns across renewables and gas-fired generation — and how these dynamics are driving valuation premiums and discounts as market participants recalibrate their strategies.

• The Binding Constraint – From EUV Machines to Megawatts – We model the full supply stack, from EUV tool delivery, advanced packaging and high-bandwidth memory to show AI compute capacity can only expand as fast as new EUV tools are deployed into high-performance computing-serving fab lines.

Top 3 Takeaways

1. Why did investors reward some hyperscalers and punish others?

Because markets are no longer impressed by AI spending alone. Google’s results showed AI investment translating into real cloud demand and revenue growth, while Meta’s higher CapEx guidance raised concerns about spending faster than returns can materialize. Investors are now separating spend that drives near‑term value from spend that still feels speculative.

2. What does rising hyperscaler CapEx mean for data centers and power markets?

Hyperscaler CapEx continues to fuel data center growth, making it a major driver of electricity demand. Enverus Intelligence® Research expects power load growth of 11 to 19 percent by 2035, with data centers accounting for a meaningful share. However, as the buildout matures, simply adding capital is no longer enough to influence market outcomes.

3. What is the emerging constraint on AI growth going forward?

The bottleneck is shifting from money to infrastructure. AI compute can only scale as fast as critical inputs like EUV tools, advanced packaging, memory, and megawatt availability allow. In other words, returns will increasingly be limited by physical and power constraints, not a lack of capital or ambition.