AI Fund Manager Dumps Nvidia Over Memory Squeeze, Capex Risk
Chawkat Nammour of Bainbridge Partners sold his entire Nvidia stake, citing tight memory supply and rising cloud hyperscaler capital intensity that could pressure hardware margins.
Chawkat Nammour, portfolio manager at London-based Bainbridge Partners, has sold his entire Nvidia holding, citing a tightening in memory supply and rising capital intensity at cloud hyperscalers. Nammour manages the Strand fund, which invests in AI, aerospace and engineering software, and reduced the fund’s gross exposure from about 170 to roughly 100.
Nammour pointed to higher hardware requirements as a key factor. He referenced Nvidia’s Rubin Ultra platform, due in 2027, which will need greater memory capacity and faster memory performance per chip. He also said memory suppliers typically expand capacity by about 20% a year and that new, meaningful supply usually takes two to three years to come online.
That gap between rapidly growing demand for AI compute and slower memory supply has pushed memory prices higher, creating a risk of bottlenecks that could slow some AI projects. Nammour cited a Morgan Stanley projection that hyperscaler capital expenditure could reach levels last seen in the dot-com era and account for about 40% of Russell 1000 cash capex over 2026–28, a total above $2 trillion.
Memory-chip stocks have risen strongly this year, with Micron and SK Hynix reaching about $1 trillion in market value. Nammour continues to hold sizable positions in SK Hynix because he expects continued demand from cloud companies, while he sold Nvidia and trimmed overall market exposure because of concentration and leverage in parts of the tech sector.
On aerospace, Nammour retains holdings he views as more stable. He described the industry as concentrated, with long product cycles and persistent demand from rising global travel, noting that roughly 80% of the world has never flown and that aircraft and engine markets can sustain profitable cycles lasting up to 15 years.
The fund uses automated tools across research, including news scrapers, Claude-powered models and proprietary sentiment trackers to speed analysis and monitor market perception. Nammour warned many of these AI tools are now widely used and do not automatically improve decision-making; he looks for clear cyclical turning points to guide investment choices.
The S&P 500 has posted recent record highs and several weeks of gains driven in part by inflows into large cloud companies. Nammour’s actions — selling Nvidia while keeping memory suppliers and aerospace exposure — reflect his adjustments to those market conditions.








