Stone Sentinel Capital · Q2 2026 · Published 3 July 2026 · Quarterly letter
Marcel Gozali argues that the hyperscalers' AI arms race needs $500 billion to $1 trillion of new revenue to earn 10% on $1 trillion of 2027 capex, and that low expectations make unpopular stocks such as Marex, a non-bank clearing firm, the safer buy.
Marcel Gozali reports that Stone Sentinel Capital gained 7.6% in the first half of 2026, against 10.2% for the S&P 500 Total Return Index. He argues that unpopular stocks carry low expectations and less downside, while AI-related stocks carry high expectations. He calculates that the largest hyperscalers would need $500 billion to $1 trillion of new revenue to earn a 10% return on their 2027 AI spending. He judges investors by their process rather than their returns, and he presents Marex, a new position.
Goldman Sachs projected roughly $7.6 trillion of cumulative AI capex between 2026 and 2031.Show at the source ↗
The largest hyperscalers (Microsoft, Alphabet, Amazon, Meta, and Oracle) are projected to spend $1 trillion in 2027 alone.Show at the source ↗
Goldman Sachs estimates that the hyperscalers would consume roughly 94% of operating cash flow on AI infrastructure before any debt financing.Show at the source ↗