To The Moon

Hedge funds pile into the next frontier of AI compute and orbital infrastructure

The latest batch of 13F filings, covering the second quarter of 2026, reveals an unusual degree of consensus among the industry’s most closely watched managers. Across multi-strategy giants, long-short equity specialists and concentrated growth funds, a clear thematic tilt has emerged: a decisive pivot towards the physical infrastructure required to sustain the artificial intelligence boom, alongside a fresh embrace of newly public space and advanced semiconductor plays.

Citadel Advisors, whose equity book swelled to $875bn, disclosed fresh stakes in SpaceX (now trading as SPCX following its mid-year IPO), Cerebras Systems (CBRS) and Quantinuum. Coatue Management, Altimeter Capital, Point72 and Appaloosa all showed meaningful exposure to the same constellation of names. Taiwan Semiconductor, Micron Technology, Advanced Micro Devices and related memory and foundry plays featured prominently in the top holdings or largest additions of David Tepper, Philippe Laffont, Brad Gerstner and several quant-heavy books. Even Berkshire Hathaway, while remaining true to its quality bias, markedly increased its Alphabet positions—a company whose own strategic interest in SpaceX has become one of the market’s more conspicuous institutional footprints.

The overlaps are not accidental. After years in which the AI trade concentrated almost exclusively in a handful of hyperscalers and Nvidia, the second quarter marked a broadening and a deepening of the thesis. Managers appear to have concluded that the next phase of value creation will be determined less by software models alone and more by the scarcity of specialised silicon, high-bandwidth memory and the energy and connectivity infrastructure needed to run them at scale. Cerebras, with its wafer-scale engines optimised for inference workloads, and Micron, whose high-bandwidth memory remains a critical bottleneck, fit neatly into that narrative. SpaceX, freshly public and already integrating AI ambitions through its xAI acquisition and Starlink constellation, offers a dual exposure: orbital broadband that could underpin global data demand, and a platform for the next generation of compute and sensing applications.

The timing is telling. Many of these positions were initiated or expanded during a period of elevated volatility in the broader semiconductor complex and after SpaceX’s own post-IPO share-price swings. That suggests conviction rather than momentum-chasing. Multi-strategy platforms such as Citadel and Point72, which can size positions dynamically and hedge systematically, used the quarter to add both directional exposure and selective new ideas in frontier technology. Concentrated funds such as Altimeter and Coatue treated the newly listed Cerebras and SpaceX as high-conviction growth vehicles, while value-oriented and hybrid managers layered in the more established foundry and memory names as relative-value expressions of the same theme.

Why the synchronised move? Three forces appear to be at work. First, the cash-flow realities of the AI build-out have become clearer: hyperscalers continue to spend heavily, yet the pure-play enablers further down the supply chain are beginning to generate more tangible returns. Second, the public-market debut of previously private assets—SpaceX in particular—has given managers liquid vehicles through which to express views that were previously confined to venture or secondary markets. Third, the broader market regime has shifted from narrow mega-cap leadership towards greater dispersion. In such an environment, skilled capital gravitates towards differentiated exposures that still sit inside a recognisable secular story.

There are, of course, risks. Valuation discipline remains uneven; some of the newer names carry the exuberance that accompanies any high-profile IPO. Geopolitical sensitivity around Taiwanese foundries and the capital intensity of both space launch and advanced chip fabrication leave little margin for execution error. Yet the breadth of participation—from Ken Griffin’s sprawling book to the more focused portfolios of Tepper and Gerstner—suggests that a significant cohort of professional investors has decided the risk is worth taking.

What the filings do not yet reveal is the degree of hedging or the private-market overlays that sit alongside these public positions. Nor do they capture activity after 30 June. Still, the public record is unambiguous: a critical mass of the industry’s most successful managers is looking in the same direction. They are not abandoning the established AI leaders so much as extending the trade into the specialised hardware, memory and orbital infrastructure that will determine whether the current cycle has further to run. In the language of the market, the smart money is moving from the application layer deeper into the stack—and, increasingly, beyond the atmosphere.