Ankur Crawford on Choosing a Job You Love

Bloomberg interview with Ankur Crawford, portfolio manager at Alger Capital, on how she builds an investment strategy. The conversation focuses on interpreting AI and compute investment cycles, but does not cite specific financial results, policy changes, or trades that would likely move markets.

Analysis

The signal here is less about one manager’s view and more about the fact that AI is still being framed as a capacity cycle, not just a product cycle. That matters because the market is implicitly paying up for names whose earnings are most levered to hyperscaler capex; if spend merely normalizes rather than accelerates, the multiple compression can arrive well before the revenue inflection slows.

The biggest second-order winners are likely the less obvious infrastructure bottlenecks: networking, power, cooling, and foundry capacity. Those businesses can keep compounding even if GPU unit growth moderates, because the system-level bottleneck migrates from chips to racks, data-center power, and interconnect. By contrast, pure-play application software is at risk of being squeezed between cheaper model access and higher customer scrutiny on ROI.

The contrarian view is that consensus is still underestimating how fast AI economics commoditize at the margin. Better model performance can expand usage, but it also lowers switching costs and weakens pricing power for vendors without proprietary distribution or data. The key falsifier is any indication that hyperscaler capex growth is rolling over on the next earnings round; that would hit the highest-duration AI names first and force a rotation from beta-rich semis into the more defensive picks-and-shovels subset.

Key Decisions for Investors

• Prefer a relative long SMH / short IGV pair for the next 1-3 months: if AI spend stays strong, semis and infrastructure should out-earn software; if spend slows, both legs likely de-rate, but IGV should underperform first.
• Use ANET and AVGO as higher-quality ways to stay exposed to the AI buildout versus chasing the most crowded GPU names; enter on any post-earnings weakness, with the thesis invalidated by a clear slowdown in hyperscaler order commentary over the next two quarters.
• Stay alert to TSM as the cleaner proxy for sustained AI capex, but treat any rally as vulnerable if the next major cloud capex guide comes in below the market’s current trajectory; this is a 3-6 month catalyst trade, not a long-term structural call.
• Avoid initiating fresh longs in pure-play AI software until there is evidence of monetization, not just usage; if enterprise buyers keep demanding ROI proof, the risk/reward remains better for infrastructure than application-layer names.
• If you want a lower-beta expression, consider simply staying long SOXX/SMH and waiting for capex guidance from MSFT, AMZN, GOOGL, and META before adding risk; a two-quarter deceleration in capex growth would be the clearest signal to reduce exposure.