Goldman Sachs Exchanges
2026-06-22 · Hosted by Allison Nathan · Goldman Sachs
Executive Summary
Maverick Capital co-CIOs Ben Silver and David Tychonsky discuss how they hunt for AI winners and manage a public-equity fund with more than three decades of alpha-driven performance. They argue the AI trade has evolved beyond GPUs into the entire hardware, infrastructure, and energy ecosystem, and that the value bottleneck — having migrated upstream toward fabrication, tools, and obscure materials — is now poised to swing back downstream toward infrastructure and applications that deliver real productivity. They flag the key market risk as an “air pocket” during the handoff from the AI training buildout to transformational end applications, and highlight life-science tools as an underappreciated AI-and-reshoring beneficiary. Funding sustainability (capex now under 100% of operating cash flow vs. ~200% in the dot-com era) and Chinese industrial competition frame their bull/bear debate.
Key Stories & Changes
1. The AI Trade Has Outgrown GPUs
The AI trade is “no longer just GPUs” — it now spans the broader hardware, infrastructure, and energy ecosystem, plus the software and services on the other side of disruption.
Central question for the market: how sustainable are current capex levels that “seem really high.”
Key difference vs. the dot-com bubble: funding source. At the dot-com peak, cumulative capex ran ~200% of operating cash flow (externally funded); over the last couple of years that number was well under 100%, funded by the largest, best-capitalized companies in the world.
The offset: a multi-trillion-dollar market-cap company recently tapped equity markets for the next tranche of the buildout, making the funding cushion “feel a little bit tighter than it was before.”
2. The Value Bottleneck Is Migrating — and About to Swing Back
The trade inverted the 2000s/2010s paradigm: value shifted from the software application layer to the hardware and infrastructure layer.
To monetize, Maverick followed the bottleneck upstream — from GPUs to fabrication, to the tools that make chips, even to “obscure materials listed on a Japanese stock exchange” — chasing the sharpest revision torque, often in low-margin businesses.
New view: the migration is “going to begin to swing back the other direction” toward infrastructure and the application layer — servicing AI and transforming businesses with productivity gains.
Implication: CPUs and databases become critical choke points; AI agents integrate LLMs into pre-existing enterprise workflows rather than living on an “island,” bringing value closer to the end user.
3. Healthcare / Life-Science Tools as a Coming AI Winner
Healthcare has seen “a large sucking sound” of capital leaving for AI, but Maverick sees pockets on the right side of major trends.
Life-science tools (products to discover and manufacture complex drugs) should benefit from reshoring of drug manufacturing to the US, driving a capex boom in equipment — expected to s
how in n
TRENumbersDS_IDENTIFIED
1. From Training Buildout to Inference and Applications three
Maverick frames the next market risk as the “handoff” from the AI training infrastructure buildout toward the underlying applications that are genuinely transformational for productivity. They warn of a potential “air pocket” in the interim that could create volatility even for believers in the AI trend. To date they see a commensurate pickup in agentic inference (e.g., coding), but argue the use case must break out into broader knowledge work to sustain demand and support out-year projections. to six months out.
AI is also accelerating drug discovery, with early signs of pickup in consumable usage; the space has consolidated for 20 years and several $5–10 billion companies are seen as ripe for M&A.
2. The Bottleneck Migrates Downstream Again
After years of value accruing upstream (fabrication, tools, materials), Maverick believes the locus of opportunity is rotating back toward infrastructure and applications. Because LLMs are being integrated into existing enterprise stacks rather than operating standalone, choke points like CPUs and databases gain importance and value shifts “closer to the edge and closer to the end user” — a meaningful reweighting of where AI dollars will land.
3. Modern Mercantilism / Reshoring as an Investment Theme
Reshoring of manufacturing to the US — explicitly drug manufacturing in the life-science-tools example — is treated as a structural tailwind layered on top of AI. Maverick expects a capex cycle in domestic manufacturing equipment, positioning certain healthcare names as both “AI winners” and “modern mercantilist winners.”
4. Funding Quality as the Tell on Sustainability
Rather than debating valuations directly, Maverick anchors on how the buildout is financed. The shift from externally funded (dot-com) to internally funded (today’s hyperscalers) is the strongest structural difference — but a recent equity raise by a mega-cap shows the cushion narrowing, making the ROI on AI spend the ultimate arbiter. —-
Sentiment Analysis
Overall Market Sentiment: Constructive but Watchful
The co-CIOs are long-term AI believers (“full-throated”) yet explicitly preoccupied with the training-to-application handoff and structural competitive risks, producing a measured, opportunity-seeking tone.
Risk Factors Highlighted
Training-to-application air pocket: A gap between the training buildout slowing and applications scaling could open volatility even within the AI trend.
Stretched AI capex / funding cushion narrowing: A multi-trillion-dollar company tapping equity markets signals the internally-funded advantage is tightening.
Uncertain ROI on AI spend: Returns must translate into broad productivity gains beyond coding to justify out-year demand projections.
Chinese industrial competition: Hardware, materials, lasers/optics, and analog semis are historically prone to Chinese commodification, threatening current beneficiaries.
Capital flight from healthcare: Near-term “large sucking sound” as money rotates out of healthcare into AI.
US political dysfunction: Difficulty making rational long-term decisions given short-term political incentives.
Geopolitical “cold war” with China: Cited as a broad, ongoing risk beyond the industrial-competition angle.
This episode was covered in today’s The Market Signal — 2026-06-22, a cross-source synthesis of multiple podcast reports.