為了應對科技巨頭龐大的資本支出需求,摩根士丹利設計出結合傳統專案融資保護機制與廣泛發行債券特性的混合型金融商品。透過讓具備優良資產負債表的科技巨頭(如Google)提供擔保,企業能以更低的利率籌集資金,同時吸引保險公司與退休基金等更廣泛的投資人參與。這種模式成功地將巨額的人工智慧運算合約轉化為證券,大幅擴展了可用於基礎設施的資金池。
這套融資模式正從資料中心建設延伸至直接為圖形處理器等硬體設備融資,例如近期為CoreWeave安排的數十億美元聯貸案。然而,隨著貸款對象逐漸從資金雄厚的科技巨頭轉向直接消耗算力的人工智慧實驗室,底層信用風險也隨之增加。分析師警告,這是一波史無前例的巨大資本支出週期,市場目前缺乏歷史經驗可供借鑒,因此參與企業的財務健康狀況將是未來關注的潛在風險。

Morgan Stanley has successfully overtaken its long-time rival Goldman Sachs in the first half of the year, with its debt and equity capital market fees growing to $2.3 billion to rank second globally, driven by innovative financing structures for the AI boom. The bank has become Wall Street's primary architect and dominant adviser in AI infrastructure financing, funneling tens of billions of dollars into data center construction. These deals are not only changing the trajectory of the technology sector but also dramatically reshaping the operations of capital markets.
To address the massive capital expenditure needs of tech giants, Morgan Stanley designed hybrid financial instruments that combine the protections of traditional project loans with the broad distribution of bonds. By having hyperscalers with pristine balance sheets, such as Google, guarantee the leases, companies can raise funds at much lower interest rates while attracting a wider pool of investors like insurers and pension funds. This model successfully packages massive AI computing contracts into securities, significantly expanding the pool of capital available for infrastructure.
This financing model is now expanding beyond data centers to directly fund hardware like graphics processing units, as seen in the recent multibillion-dollar syndicated loan arranged for CoreWeave. However, as lending moves away from well-capitalized tech giants toward the AI labs directly consuming the computing power, the underlying credit risk increases accordingly. Analysts warn that this is an unprecedented massive capital expenditure cycle with no historical playbook, making the financial health of participating companies a critical potential risk to monitor.