Key Takeaways
- • Gavin Baker, managing partner of Atreides Management, argues that investors are prematurely judging Big Tech's massive AI infrastructure spending and that improved cash flow metrics suggest returns are materializing.
- • Meta's operating cash flow rose 25% year-over-year in Q2 2026, despite a 91% year-over-year decline in free cash flow to $784 million, which Baker characterizes as "material acceleration at this scale."
- • GPU prices remain elevated in 2026, indicating sustained demand for AI computing capacity, contrary to investor expectations that prices would normalize after 2024-2025.
- • Cloud providers have locked in older, lower-priced compute contracts at significant discounts to current spot market rates, creating future monetization upside as existing infrastructure contracts renew at higher prices.
- • Baker contends that the recent July sell-off in memory and chip stocks was driven by unfounded concerns about Big Tech's inability to generate returns on AI infrastructure investments.
Summary
Gavin Baker, the prominent hedge fund manager who gained recognition for early SpaceX investments, has pushed back against investor skepticism surrounding Big Tech's substantial capital expenditures on AI infrastructure. Speaking on the "Invest Like The Best" podcast, Baker argued that the market is rushing to judgment on whether these companies can achieve acceptable returns on their massive technology investments.
Baker highlighted improving financial metrics as evidence that Big Tech's AI spending is beginning to generate results. He pointed to Meta's Q2 2026 operating cash flow, which increased 25% year-over-year despite the company's free cash flow declining 91% to $784 million. This divergence, which may appear counterintuitive on its surface, Baker characterized as "material acceleration at this scale," suggesting the underlying operational dynamics are strengthening even as certain cash flow metrics compressed due to capital deployment.
The hedge fund manager also addressed the persistent strength in GPU pricing throughout 2026, citing this as validation of robust demand for AI computing resources. "I don't think anyone in '24 or '25 thought that the prices of old GPUs would still be going vertical in 2026," Baker stated, underscoring how the market has underestimated the durability of GPU demand.
Baker identified a structural opportunity within cloud providers' existing contract portfolios. Many providers have locked in compute capacity under long-term agreements signed when GPU prices were substantially lower. With current spot market prices for compute significantly higher, these legacy contracts represent assets trading at steep discounts to replacement value. As these contracts renew or expire, providers will have the opportunity to renegotiate at materially higher rates, creating future revenue acceleration.
In July 2026, investors had broadly sold memory and chip stocks, driven largely by concerns that Meta and other technology companies were deploying capital on AI infrastructure without clear paths to profitability. Baker's thesis challenges this narrative, contending that improving cash flow dynamics and pricing power in GPU markets demonstrate that return questions are beginning to find answers.
Baker highlighted improving financial metrics as evidence that Big Tech's AI spending is beginning to generate results. He pointed to Meta's Q2 2026 operating cash flow, which increased 25% year-over-year despite the company's free cash flow declining 91% to $784 million. This divergence, which may appear counterintuitive on its surface, Baker characterized as "material acceleration at this scale," suggesting the underlying operational dynamics are strengthening even as certain cash flow metrics compressed due to capital deployment.
The hedge fund manager also addressed the persistent strength in GPU pricing throughout 2026, citing this as validation of robust demand for AI computing resources. "I don't think anyone in '24 or '25 thought that the prices of old GPUs would still be going vertical in 2026," Baker stated, underscoring how the market has underestimated the durability of GPU demand.
Baker identified a structural opportunity within cloud providers' existing contract portfolios. Many providers have locked in compute capacity under long-term agreements signed when GPU prices were substantially lower. With current spot market prices for compute significantly higher, these legacy contracts represent assets trading at steep discounts to replacement value. As these contracts renew or expire, providers will have the opportunity to renegotiate at materially higher rates, creating future revenue acceleration.
In July 2026, investors had broadly sold memory and chip stocks, driven largely by concerns that Meta and other technology companies were deploying capital on AI infrastructure without clear paths to profitability. Baker's thesis challenges this narrative, contending that improving cash flow dynamics and pricing power in GPU markets demonstrate that return questions are beginning to find answers.
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.