Measuring Returns on AI Infrastructure

Daniel Pilling, Co-PM of the Sands Capital Global Growth Fund, has spent nearly 20 years investing across long-only and long-short strategies, including time at Fidelity, Millennium, and Balyasny. Today at Sands Capital, he takes a very different approach: concentrated, deep-dive investing in high-quality growth companies with the potential to compound for years.

In this episode, Daniel breaks down why he believes the AI investment cycle is still incredibly early. We discuss $NVDIA, $TSMC, $ASML, memory, AI agents, the return on GPU infrastructure, and why compute could remain supply constrained for a long time. Daniel also explains why Anthropic's growth has been unlike anything he's seen before and how Sands thinks about finding the long-term winners as AI diffuses across the economy.

If you're wondering whether the AI trade has gone too far—or whether we're still at the beginning of a much larger cycle—this conversation offers a long-term investor's framework for thinking about what comes next.

"I've never seen anything like this."

"We're going to be supply constrained in terms of compute for a very long time."

“The reason for that is, again, the low penetration and the high ROI of what’s happening.”

"Anthropic and agentic AI is incredible. And it's just going viral and the pace of adoption is unheard of."

Stocks: $NVDA, $TSM, $MU, $ASML, $AMZN, $GOOGL, $META, $AMD, $ZM

Topics: Sands Capital, Artificial Intelligence, NVIDIA, TSMC, ASML, Memory, AI Agents, Anthropic, Compute, Semiconductors, GPU Economics, Long-Term Investing, AI Infrastructure, AI Innovator Fund

*Not Investment Advice

[00:00:00] Introduction to Daniel Pilling, Co-PM of the Sands Capital Global Growth Fund.

[00:01:15] Daniel’s path from banking and multi-manager investing to long-term growth.

[00:03:02] How Daniel became obsessed with investing at 12.

[00:04:13] Why Daniel left Millennium and Balyasny for Sands Capital.

[00:05:14] Sands Capital’s philosophy: concentrated portfolios, deep research, and long-term ownership.

[00:07:32] Why memory and NVIDIA remain high-conviction AI investments.

[00:09:54] NVIDIA’s market share and why open-source AI could support GPU demand.

[00:12:20] Why NVIDIA, Cerebras, Trainium, and TPUs can all win.

[00:14:13] The case for a memory shortage as AI agents scale.

[00:19:00] Why memory may not follow a traditional cyclical pattern.

[00:22:13] AI infrastructure returns and increasingly valuable compute.

[00:23:48] Why rising older-GPU prices challenge depreciation concerns.

[00:24:23] Anthropic’s growth, Zoom during COVID, and rapid agentic AI adoption.

[00:26:46] Why AI compute could remain supply constrained and create an “upside cliff.”

[00:28:34] Why Daniel compares AI adoption to electricity.

[00:30:42] How AI could make investment research faster and more effective.

[00:32:03] Why ASML and TSMC remain key AI infrastructure constraints.

[00:34:28] Generating differentiated returns through multi-year views.

[00:37:52] Why 99% of daily market information doesn’t matter.

[00:39:52] Humility in investing and recognizing when the Zoom thesis changed.

[00:43:42] Why AI remains early, underpenetrated, and rapidly improving.

[00:46:22] Sands Capital’s AI exposure across semis, memory, cloud, and software.

[00:52:56] The case for Meta despite rising CapEx and declining free cash flow.

[00:56:45] The fund’s AI exposure and global diversification.

[00:58:14] The AI Innovator Fund thesis: low penetration, constrained compute, and AI winners.

💡 This episode is powered by:

Oxford Data Plan: Request a Demo 

AlphaSense: Request a Demo

Pitch The PM Links:

📩 Subscribe to our Substack for research updates and new high-conviction episodes from top PMs, and our Job Board: ⁠https://pitchthepm.substack.com/⁠Doug Garber on LinkedIn: https://www.linkedin.com/in/doug-garber-42aa508 

Sands Capital Links:Daniel Pilling on Linkedin: https://www.linkedin.com/in/daniel-pilling-14343116/Sands Capital: https://www.sandscapital.com/