Key Takeaways
- • Gavin Baker, CIO of Atreides Management, argues investors have prematurely condemned Big Tech's AI capital spending despite a 35-40% selloff in the AI equity complex from late June highs, asserting the bear case "doesn't survive contact with data."
- • Baker distinguishes the current AI cycle from the 1999 dot-com bubble by highlighting that hyperscalers are generating solid returns on investment (ROI) from AI infrastructure spending, contrasting with the profitless dot-com era.
- • A broad 10-20% market pullback could force 30-40% drawdowns in high-beta AI stocks, creating disproportionate pain for recent buyers despite fundamentals supporting longer-term AI infrastructure demand.
- • Baker identifies edge AI—where powerful models run directly on devices rather than in data centers—as the most plausible bear case for data center spending, though he questions the practical momentum of this "bottleneck trade."
- • The AI infrastructure narrative hinges on whether hyperscalers can sustain ROI from trillion-dollar capital commitments; Baker suggests current valuations at decade lows (particularly NVIDIA) may reflect excessive pessimism about demand.
Summary
Gavin Baker, founding partner and chief investment officer of Atreides Management, has pushed back against the recent market skepticism surrounding Big Tech's massive capital expenditure commitments to artificial intelligence infrastructure. Speaking on the podcast "Invest Like the Best" with Patrick O'Shaughnessy, Baker argued that the intensity of the recent selloff—which saw the AI equity complex decline 35-40% from its late June highs—does not reflect the underlying fundamentals supporting continued AI spending by hyperscalers.
Baker's core thesis centers on a critical distinction between the current AI investment cycle and the 1999 dot-com crash. Unlike the dot-com era, where companies were burning capital with minimal revenue visibility, today's hyperscalers are reportedly generating solid returns on their AI infrastructure investments. Baker characterized this as evidence that "there are no dark GPUs"—a reference to the reality that deployed AI infrastructure is actively generating economic value rather than sitting idle. This ROI generation, in his view, economically justifies continued capital deployment at scale, even if near-term equity volatility persists.
The hedge fund manager cautioned that downside scenarios exist but are narrower than current market sentiment suggests. A moderate 10-20% decline in broad equity indices could cascade into 30-40% drawdowns in high-beta AI stocks, creating acute losses for investors who entered positions during the euphoric phases of the rally. However, Baker argues this volatility risk should not be conflated with fundamental impairment of the AI infrastructure thesis itself.
On the question of what could genuinely derail hyperscaler spending, Baker identified edge AI—computing paradigms where sophisticated models execute on end-user devices rather than centralized data centers—as the most intellectually coherent bear case. However, he expressed skepticism about whether this "bottleneck trade" has accumulated sufficient momentum to materially alter capital allocation decisions. The implication is that while edge AI represents a legitimate long-term competitive pressure on data center economics, it does not yet constitute sufficient cause to abandon trillion-dollar infrastructure commitments.
Baker's analysis suggests that current valuations, particularly for infrastructure leaders like NVIDIA trading at decade-low multiples, may reflect excessive pessimism. His argument implicitly positions the recent selloff as a volatility-driven re-rating rather than a fundamental repricing, with patient capital potentially positioned to benefit as market sentiment normalizes around the durable economics of hyperscaler AI spending.
Baker's core thesis centers on a critical distinction between the current AI investment cycle and the 1999 dot-com crash. Unlike the dot-com era, where companies were burning capital with minimal revenue visibility, today's hyperscalers are reportedly generating solid returns on their AI infrastructure investments. Baker characterized this as evidence that "there are no dark GPUs"—a reference to the reality that deployed AI infrastructure is actively generating economic value rather than sitting idle. This ROI generation, in his view, economically justifies continued capital deployment at scale, even if near-term equity volatility persists.
The hedge fund manager cautioned that downside scenarios exist but are narrower than current market sentiment suggests. A moderate 10-20% decline in broad equity indices could cascade into 30-40% drawdowns in high-beta AI stocks, creating acute losses for investors who entered positions during the euphoric phases of the rally. However, Baker argues this volatility risk should not be conflated with fundamental impairment of the AI infrastructure thesis itself.
On the question of what could genuinely derail hyperscaler spending, Baker identified edge AI—computing paradigms where sophisticated models execute on end-user devices rather than centralized data centers—as the most intellectually coherent bear case. However, he expressed skepticism about whether this "bottleneck trade" has accumulated sufficient momentum to materially alter capital allocation decisions. The implication is that while edge AI represents a legitimate long-term competitive pressure on data center economics, it does not yet constitute sufficient cause to abandon trillion-dollar infrastructure commitments.
Baker's analysis suggests that current valuations, particularly for infrastructure leaders like NVIDIA trading at decade-low multiples, may reflect excessive pessimism. His argument implicitly positions the recent selloff as a volatility-driven re-rating rather than a fundamental repricing, with patient capital potentially positioned to benefit as market sentiment normalizes around the durable economics of hyperscaler AI spending.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.