Leopold Aschenbrenner has repositioned his $200 billion fund, cutting exposure to major semiconductor firms and placing roughly $9 billion in short bets against companies including NVIDIA, ASML, and Oracle. At the same time, nearly 20 percent of the portfolio has been redirected toward AI developer Anthropic and businesses tied to data centers, memory, and power generation.
The move signals a broader shift away from chipmakers and toward the physical systems required to sustain AI expansion.
NVIDIA bond sale highlights broader financing trend
The repositioning comes after NVIDIA issued $25 billion in bonds, with maturities ranging from two to thirty years. The offering was oversubscribed almost four times, despite the company holding more than $12 billion in cash and approving $80 billion in share buybacks alongside a dividend increase.
Across the sector, companies such as Amazon and Google have also tapped debt markets, reflecting a wider strategy of using low-cost financing to fund large-scale AI infrastructure growth.
Infrastructure seen as next bottleneck in AI growth
Aschenbrenner’s strategy is based on the view that constraints in AI are shifting away from chip supply toward energy availability, memory capacity, and data network build-out. His fund, which has reportedly expanded from $2 billion to $200 billion in about 18 months, now holds its largest exposure in energy-related assets.
The emphasis reflects growing logistical challenges in power distribution, land development, and regulatory approvals, all of which are slowing the pace of new data center construction.
Rising demand for copper and fiber
The portfolio has also increased focus on materials critical for data transmission, particularly copper and optical fiber. Copper continues to dominate short-distance, high-speed connections, while fiber is gaining ground in longer-distance and high-temperature environments.
Demand for both is rising, with copper prices remaining strong through mid-2026 as bandwidth requirements accelerate.
Anthropic surge lifts fund valuation
Anthropic has become a central holding, with its valuation climbing from about $60 billion in early 2025 to nearly $965 billion. The roughly fifteenfold increase has significantly boosted the overall size of the fund, placing it alongside long-established firms of similar scale.
Capital rotates toward physical build-out
Additional holdings in companies such as Marvell, CoreWeave, and Iron underscore a focus on cloud infrastructure and optical networking. These firms provide the physical capacity required for advanced AI model training, where supply chain and labor constraints remain significant.
Data center construction spending reached $49.5 billion by April 2026, up sharply from $13.6 billion a year earlier. At the same time, costs have surged, with standard facilities averaging $11.3 million per megawatt and AI-optimized sites exceeding $20 million per megawatt.
Short positions reflect valuation concerns
The fund’s short stance against companies like Oracle has aligned with recent market moves, including a drop in Oracle’s shares after announcing a $40 billion capital raise that pushed its free cash flow deep into negative territory.
Similarly, the bearish view on ASML suggests expectations that valuations among key chip suppliers may have peaked, even as demand shifts toward other bottlenecks in the supply chain.
Power emerges as decisive factor
Energy availability has become the primary constraint in data center development, surpassing location as the key factor in site selection. This shift is driving capital toward utilities, grid expansion, and renewable-powered infrastructure.
At the same time, related sectors such as optical networking components and copper production are positioned to benefit. Some projections estimate the optical networking market could expand to $154 billion within two years, while copper prices have risen more than 30 percent year over year.
Trillions required for AI infrastructure expansion
The strategy reflects a broader thesis that the next phase of the AI economy will be defined by physical infrastructure rather than software alone. Analysts estimate that nearly $3 trillion in investment will be required by 2028 to support global AI development.
In this view, opportunities are shifting toward the digital equivalents of industrial inputs such as power, materials, and construction capacity, as these become the limiting factors for continued growth.
Explore how AI is transforming trading itself with our in-depth guide: AI copy trading explained for investors.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
