The Capex Boom Is Real, but the Return on Equity Question Remains Open

What's New

Jim Zelter, President of Apollo Global Management, argues at the Milken Institute Global Conference 2026 that the unprecedented investment grade capex boom may not translate into commensurate returns for equity holders, even as it powers the broader economy. Speaking on the Global Capital Markets panel alongside Jon Gray, Jenny Johnson, Robin Vince, Waleed Al Mokarrab Al Muhairi, and David Faber, Zelter frames the AI buildout as historically distinct from prior leveraged finance cycles. The implication: consumers and society may capture the gains while the economic equation for the financing equity remains unresolved.

Why It Matters

The consensus view on this panel and across allocator conversations is that AI capex plus productivity gains equals a virtuous cycle. Zelter accepts the capex math but separates it from the return math. If trillions flow into investment grade balance sheets to fund AI infrastructure, but the science benefits accrue mainly to end users, the asset owners financing this buildout could end up holding capital-intensive assets with compressed returns. That reframes the question for allocators sizing exposure to hyperscalers, infrastructure equity, and the open architecture financing structures Zelter himself is building.

Big Picture Drivers

• Investment grade has eclipsed Treasuries: Net issuance in the IG market this year will exceed net issuance in the US Treasury market at north of $1.1 trillion, a structural shift in where capital is being formed.

• Capex is full-stack and multi-year: A handful of companies announced $750 billion in spending in a single week, and this scale will continue for years across power, real estate, cooling, chips, and semiconductors.

• Open architecture replaces brutal competition: Unlike the GFC era when firms were pure competitors, lenders, borrowers, and financers now collaborate across IG, infrastructure, and private credit because no single structure can absorb the capital need.

• Asset-light is becoming asset-heavy: The Magnificent Seven and similar companies are transitioning to capital-intensive business models, which changes the equity return profile relative to the prior decade.

• Delivery mechanisms are outdated: Drawdown vehicles with 8x leverage and three-year exits are misaligned with assets that should be owned with more equity, less debt, and evergreen structures to match aging populations and longer liability horizons.

By The Numbers

• $1.1 trillion: Net IG issuance this year, exceeding net Treasury issuance.

• $750 billion: Capex announced by a handful of companies in a single week.

• $230 trillion: US household net worth, up from $60 trillion at the GFC.

• 37 years: Zelter's tenure in markets, which did not prevent him from missing the SVB credit call.

• 15th: The Chinese five-year plan, which doubles down on manufacturing dominance rather than consumer rotation.

Key Trends to Watch

• The 2028 to 2030 ROE reckoning: By the end of the decade, the question of whether AI capex generated real return on economic capital for equity holders will become answerable, and Zelter signals the answer may disappoint.

• Private credit's structural growth despite headline noise: BDC concerns of the past eight weeks miss the broader ocean of private capital, which will continue to grow because the capex stack cannot be funded from any single source.

• Evergreen structures over drawdown funds: Infrastructure and real estate will increasingly be held in perpetual vehicles with lower leverage, replacing the lever-and-flip model that defined the prior two decades.

• Value investing versus building from scratch: With AI enabling rapid company creation, buying incumbent value assets becomes harder to justify against the option to build from zero, a tension every capital allocator will face.

Memorable Quotes

• "The big plot is the ocean of of private capital in aggregate." Zelter dismisses recent private credit anxieties as missing the structural scale of the asset class.

• "Are we going to find that the economic equation didn't really follow through for the economic model for that massive investment?" His central question about whether equity returns will justify the AI capex boom.

• "We're all practitioners. I know we're put up on the panel to be predictors but we really are practitioners. I think there's humility in us. I was at the front of the table and when SVB went down saying it was going to be a massive credit crisis, I missed it. I had only been in the market 37 years." Zelter framing why he is cautious about the consensus AI optimism.

• "China being the fitness center of the globe." His takeaway from the China Development Forum on industrial competition, reinforcing his concern about Europe and incumbents.

The Wrap

Zelter is not predicting an AI bust or a private credit crisis. He is predicting that the locus of value creation may diverge from the locus of capital deployment, and that allocators who assume capex spending automatically translates to equity returns are repeating a familiar mistake. His thesis succeeds if, by 2028 to 2030, the realized ROE on the trillion-dollar AI infrastructure stack disappoints relative to the consumer and productivity gains it enables. It fails if the productivity step function is large enough to lift both end users and asset owners simultaneously. The next 24 to 36 months of earnings from hyperscalers, power generators, and data center owners will start to answer the question.