13F File Revealed: Institutional holdings of Bitcoin ETFs and mining companies have increased, and AI funds have not been withdrawn

According to Woofun AI, D1 Capital Partners (D1 Capital Partners)'s latest 13F filing shows that the weight of Bitcoin-related assets in its portfolio has risen significantly, but that doesn't mean the agency is withdrawing from the field of artificial intelligence. Instead, the data revealed a more complex capital allocation logic: while maintaining huge exposure to tech giants, capital is flowing to hybrid targets that have both Bitcoin mining capabilities and AI computing power infrastructure attributes.

This 'dual attribute' investment strategy blurs the boundaries between traditional crypto assets and technology growth stocks, showing that institutional investors are not simply switching tracks, but are looking for a crossroads where they can simultaneously benefit from rising digital asset prices and burgeoning demand for high-performance computing. The market lacks empirical support for the single narrative of “capital flows from artificial intelligence to Bitcoin.” The real trend is for capital to seek a better risk-return ratio between the two high-growth fields, diversifying investment portfolios and increasing resilience to risks through entities with dual business models.

To thoroughly analyze the theoretical background of this phenomenon, it is necessary to face up to the essential limitations of Form 13F as a regulatory disclosure tool. Arthur Hayes once proposed that the huge capital absorption effect in the field of artificial intelligence is crowding out the liquidity that could have flowed into the Bitcoin and Ethereum markets; Michael Sayle described this pressure as a temporary 'suction effect', predicting that as AI investment matures and benefits are redistributed, some of the capital will return to the crypto market.

However, these macro-views cannot be directly verified through Form 13F, because the document only reflects long positions in specific stocks in the US market at the end of the quarter, completely ignoring short sales, swaps, private investments, overseas positions, and intra-quarter trading trends. As a result, although Decainer reduced its holdings of Intel (INTC.US) and Micron Technology (MU.US) — two companies that are not pure artificial intelligence companies but are deeply involved in the supply chain — and increased its holdings of four publicly listed Bitcoin mining companies in the same quarter, the filing documents failed to prove that the funds obtained from the holdings reduction were directly used for crypto investments.

This data gap makes it extremely misleading to infer 'capital transfer' based solely on changes in positions. Investors must combine more dimensional information to restore the actual capital flow path.

Details of Decana's holdings further confirm this complexity. The company's exposure to the technology industry not only remained unabated; its Amazon (AMZN.US) holdings surged from 45,800 shares to 541,600 shares, purchased 336,300 new Alphabet (GOOGL.US) shares, and increased investments in TSM.US (TSM.US) and STM.US (STM.US).

Meanwhile, its four new mining companies — Riot (RIOT.US), Hut 8 (HUT.US), Bitdeer (BTDR.US), and IREN (IREN.US) — all showed a strong trend of 'de-pure encryption' and actively transitioning to artificial intelligence infrastructure. Riot (RIOT.US) signed the first data center lease agreement with AMD (AMD.US) to expand the high-performance computing business; Hut 8 (HUT.US)'s Beacon Point campus has 352 megawatts of electricity supply capacity and has commercialized dedicated artificial intelligence data centers; Bitdeer (BTDR.US) disclosed in its May operating report that its business combining Bitcoin mining and artificial intelligence cloud services has annual recurring revenue of about 69 million In dollars, GPU utilization is as high as 90%; IREN (IREN.US) also deeply integrates mining operations with artificial intelligence cloud services based on GPU infrastructure.

According to data compiled by Woofun AI, this position structure shows that Decana is not “choosing one of the two,” but is betting on hybrid enterprises that can utilize existing power and data center resources while also serving the needs of digital asset mining and artificial intelligence computing. These companies not only provide indirect exposure to the Bitcoin price, but also have a substantial business foundation to participate in the artificial intelligence wave, thus forming a unique dual support in valuation logic.

From a broader institutional perspective, Tudor Investments, Zhenjie, and UBS (UBS.US)'s 13F documents also revealed an increase in Bitcoin spot ETF holdings, but these increases were not accompanied by a reduction in AI-related assets. According to Tudor Investments' holdings in the second quarter, the number of Hellyder's iShares Bitcoin Trust (IBIT.US) common shares held rose from 579,083 shares in the first quarter to 688,529 shares, an increase of about 18.9%.

However, the company still holds a large number of IBIT (IBIT.US) options: the number of shares subject to call options fell from 998,000 to 148,000 shares, and the number of shares subject to put options fell slightly from 725,000 to 715,000 shares.

This complex derivatives structure shows that Tudor's investment strategy is not simply one-sided speculation, but involves a comprehensive arrangement of hedging and risk management. The situation of Jane Street is even more special. As the main market maker, the market value of its holdings in the five ETFs IBIT (IBIT.US), FBTC (FBTC.US), ARKB (ARKB.US), BITB (BITB.US), and GBTC (GBTC.US) increased from about US$438.4 million in the first quarter to around US$1.01 billion in the second quarter.

This increase is due in large part to market capitalization fluctuations rather than new capital inflows, and the purpose of holding positions is probably more to meet clients' trading needs, provide liquidity, or carry out arbitrage rather than to express long-term optimism. UBS (UBS.US) IBIT (IBIT.US) common stock holdings increased by about 11.9% from 364,371 shares to 407,890 shares, but its call option holdings were larger, surging from 80,000 shares to around 1.95 million shares. These banks' motives for holding positions may involve customer service, structured product construction, or hedging operations, rather than simple proprietary investments, so they cannot simply be interpreted as an institution's collective optimism about Bitcoin.

Indirect investment channels and futures market data provide another layer of supporting evidence. Strategy (MSTR.US) became the largest holding project in the $4.5 billion GRNY ETF (GRNY.US) managed by Tom Lee, which enabled the fund to indirectly hold assets closely linked to Bitcoin reserves through equity.

However, this also does not prove that GRNY (GRNY.US) funds bought Strategy (MSTR.US) shares by selling artificial intelligence holdings. Following the 13F data as of June 30, the CME (CME.US) standardized Bitcoin futures market position data released by the US Commodity Futures Trading Commission (CFTC) showed that net long positions in the asset manager category rose from 2,000 contracts (4,754 longs minus 2,754 bears) to 3,698 contracts on September 1 (4,837 long minus 1,139 shorts).

This change is mainly due to a decrease in short positions rather than a sharp increase in bulls, reflecting the weakening of bearish sentiment on Bitcoin in the market, but there has been no large-scale influx of speculative bulls. Meanwhile, asset managers' net holdings in NASDAQ 100 futures rose from 68,195 contracts to 72,886 contracts.

Although the Nasdaq index can be used as a reference for the technology market, it is not directly equivalent to investing in artificial intelligence, and there is a significant difference in size and risk attributes from Bitcoin futures. As a result, asset managers have reduced their bearish exposure to Bitcoin while maintaining and increasing their long positions in indices, mainly technology stocks, which further confirms that capital is not being transferred in a zero-sum game between the two markets.

Taken together, the available evidence suggests that institutional investors' interest in Bitcoin ETFs, mining companies, and regulated futures products is indeed picking up, but this recovery is not at the expense of investments in artificial intelligence. The operating model of Decainer, Tudor Investments, Zhen Street, and UBS (UBS.US) shows that capital is seeking a more refined allocation strategy: by holding mixed targets with both computing power and mining attributes, or using derivatives for risk management, investors can obtain potential benefits from the digital asset market without breaking away from the main line of technological growth. In the future, as the Bitcoin price fluctuates and the AI business becomes more divided, the position adjustments of these institutions will provide more clues to understand their underlying investment logic. The current market structure is not simply a 'lose it', but rather shows the characteristics of a 'selective allocation', that is, capital flows to assets that can cross traditional industry boundaries and provide multiple value drivers.

This trend indicates that in the future process of integrating Web3 and technology, enterprises with dual attributes will become the focus of institutional capital attention, and assets that simply rely on a single narrative may face greater valuation pressure.