--- title: "In-depth Analysis: The \"Capital Internal Circulation\" in the U.S. AI Circle" type: "News" locale: "en" url: "https://longbridge.com/en/news/280076233.md" description: "Morgan Stanley believes that the AI sector in the United States has formed a \"capital internal circulation of mutual investment between suppliers and customers\" to support the massive expansion of computing power. This model conceals real leverage and systemic risks through off-balance-sheet guarantees, warrants, and other means. The firm, centered on the OpenAI ecosystem, reveals the hundreds of billions of dollars in cross-investments, computing power leasing, and IP bundling among giants such as NVIDIA, Microsoft, Oracle, AMD, Amazon, and Disney" datetime: "2026-03-23T01:04:54.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/280076233.md) - [en](https://longbridge.com/en/news/280076233.md) - [zh-HK](https://longbridge.com/zh-HK/news/280076233.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/280076233.md) | [繁體中文](https://longbridge.com/zh-HK/news/280076233.md) # In-depth Analysis: The "Capital Internal Circulation" in the U.S. AI Circle Morgan Stanley's latest warning states that as the investment cycle in artificial intelligence computing accelerates, the capital intensity of the AI industry is reaching unprecedented levels. However, outside of hyperscale cloud service providers, the capital across the entire ecosystem remains constrained. To support the massive infrastructure expansion, **the U.S. AI sector is evolving a highly complex and interwoven "new financing structure" and capital internal circulation system.** On March 23, according to news from the Wind Trading Platform, Morgan Stanley stated in its latest research report that in the coming years, AI-related investments will account for about 50% of the total capital expenditures of large-cap stocks, with its capital intensity surpassing that of the previous internet bubble period. **The scale of this investment is enormous and highly front-loaded, resulting in a serious mismatch between recent capital demand and AI revenue realization. It is this mismatch that has spurred the emergence of various new financing structures throughout the ecosystem.** The research report uses OpenAI as a core case study, mapping out the capital flow involving multiple parties such as NVIDIA, Microsoft, Oracle, CoreWeave, Amazon, AMD, and Disney. Morgan Stanley points out that **the current AI boom is built on the foundation of suppliers and customers providing each other with funding, warrants, and off-balance-sheet guarantees.** While this model has greatly accelerated the expansion of infrastructure, it may also exaggerate the apparent contract pricing and hide the real economic leverage and risks off the balance sheet, potentially leading to "sky-high contracts" being overestimated in the headlines. ## "Suppliers support customers, customers support suppliers" — A panoramic view of the circular financing structure The core finding of the report is that **the AI ecosystem has formed a highly interconnected structure where suppliers fund customers, and customers reciprocate by supporting suppliers.** These arrangements include: > - Supplier financing and favorable terms > > - Long-term procurement commitments (Take-or-Pay contracts) > > - Revenue-sharing arrangements > > - Supplier repurchase agreements > > - High customer concentration > > - Third-party guarantees and endorsements > > - IP licensing in exchange for model access > > - Equity investments in exchange for computing power commitments > Morgan Stanley believes that these arrangements are essentially financing mechanisms that allow numerous ecosystem participants to expand infrastructure to levels that their own cash flows cannot support. The report explicitly warns: > These financing agreements, while accelerating data center construction, may also prematurely overdraw future demand and redistribute risks among counterparties. ## Core Case: The Hundreds of Billions in Capital Flow within the OpenAI Ecosystem Morgan Stanley states that taking the currently most disclosed OpenAI ecosystem as an example, we can clearly see the astonishing scale of this capital internal circulation: **1\. The deep binding of computing power giant NVIDIA** - **Direct Investment:** In February 2026, NVIDIA finalized plans to invest $30 billion in OpenAI. - **Supporting Core Customers:** NVIDIA holds approximately 9% of cloud service provider CoreWeave (CRWV), valued at about $4 billion, and has added an additional $2 billion investment. - **Repurchase and Leasing:** NVIDIA agreed to repurchase up to $6.3 billion of unsold computing capacity from CRWV and signed a $1.3 billion agreement to lease back the GPUs it supplied to CRWV over four years. Additionally, NVIDIA has $22.7 billion in future leasing obligations (primarily for data centers). **2\. Microsoft's Comprehensive Penetration** - **Equity and Computing Commitment:** Microsoft's $13 billion investment in OpenAI is currently valued at about $135 billion (approximately 27% on a fully diluted basis). OpenAI has committed to purchasing an additional $250 billion in Azure services. - **Revenue Sharing and Hardware Procurement:** There is a revenue-sharing agreement between the two parties, expected to generate $6.1 billion in revenue for Microsoft in the 2026 calendar year. Microsoft anticipates spending about $45 billion on NVIDIA chips in fiscal year 2026 and has signed a $10 billion server leasing agreement with CRWV. Microsoft's additional leasing agreements total up to $155.1 billion. **3\. Oracle and AMD's Huge Orders** - **Oracle:** Signed a contract worth approximately $40 billion with NVIDIA (to purchase about 400,000 chips) and a total of about $5.7 billion in GPU deployment contracts with AMD. OpenAI has agreed to purchase $300 billion worth of computing power from Oracle over approximately five years. Oracle's data center leasing commitments total up to $261 billion. - **AMD:** Announced a strategic agreement worth $100 billion with OpenAI (deploying 6GW of AMD GPUs). As part of the agreement, AMD issued up to 160 million stock warrants to OpenAI. **4\. Amazon and Disney's Cross-Industry Entry** - **Amazon:** Committed to investing $50 billion in OpenAI (initially $15 billion, with an additional $35 billion contingent on milestones). The scale of the partnership has expanded by $100 billion over eight years. Amazon currently has $96.4 billion in unstarted leases. - **Disney:** Invested $1 billion in OpenAI (with additional stock warrant options) in exchange for the right to use OpenAI models, while OpenAI received a three-year license to use Disney IP. This model of exchanging IP for computing power is, in fact, a form of financing that does not require immediate cash outlay. Morgan Stanley stated that in this AI capital network woven together by hundreds of billions of dollars, new funds often only cover a portion of the total computing commitments, with the fulfillment of remaining contracts highly dependent on future revenue growth or new rounds of financing. **Investors must be wary of the systemic fragility that this "left hand to right hand" capital cycle may bring when pricing for AI enthusiasm.** At the same time, Morgan Stanley has clearly listed seven potential risks arising from this cyclical structure in its research report: > - **1\. Warrants Distort True Pricing** > > Customers exchange long-term procurement commitments for supplier warrants, making the contract headline amounts unable to reflect the repeatable true pricing level. > > - **2\. Off-Balance-Sheet Guarantees Hide True Leverage** > > Third-party guarantees provided by cloud computing suppliers support data center construction but are often not reflected on the guarantor's balance sheet, leading to a significant divergence between reported leverage and true economic leverage. > > - **3\. IP Licensing Masks True Operating Costs** > > Content creators authorize IP under favorable or non-cash terms in exchange for access to AI models, resulting in an underestimation of the true operating costs and cash needs of AI labs. > > - **4\. Supplier Equity Investments Amplify Debt Risks** > > Supplier equity investments provide cash flow backing for other suppliers selling to the same customer, allowing them to take on more debt, creating a chain effect that further drives capacity expansion. > > - **5\. High Customer Concentration Amplifies Counterparty Risk** > > Revenue growth increasingly relies on the success of a few AI labs, significantly increasing concentration risk. > > - **6\. Revenue Sharing Arrangements Obscure True Demand** > > Revenue sharing arrangements among multiple parties may allow several entities to record the same revenue under U.S. GAAP, making it difficult to assess true AI demand. > > - **7\. Repurchase Agreements May Inflate Demand** > > Supplier repurchase agreements transfer risk back to suppliers, providing customers with downside protection, which may artificially inflate apparent demand figures ### Related Stocks - [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/en/quote/SMH.US.md) - [YieldMax MSFT Option Income Strategy ETF (MSFO.US)](https://longbridge.com/en/quote/MSFO.US.md) - [State Street® SPDR® S&P® Smcndctr ETF (XSD.US)](https://longbridge.com/en/quote/XSD.US.md) - [iShares Global Tech ETF (IXN.US)](https://longbridge.com/en/quote/IXN.US.md) - [GraniteShares 2x Long AMD Daily ETF (AMDL.US)](https://longbridge.com/en/quote/AMDL.US.md) - [Vanguard Information Technology ETF (VGT.US)](https://longbridge.com/en/quote/VGT.US.md) - [The Technology Select Sector SPDR® ETF (XLK.US)](https://longbridge.com/en/quote/XLK.US.md) - [OpenAI (OpenAI.NA)](https://longbridge.com/en/quote/OpenAI.NA.md) - [iShares Expanded Tech-Software Sect ETF (IGV.US)](https://longbridge.com/en/quote/IGV.US.md) - [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/en/quote/SOXX.US.md) - [Direxion Daily NVDA Bull 2X Shares (NVDU.US)](https://longbridge.com/en/quote/NVDU.US.md) - [Invesco Semiconductors ETF (PSI.US)](https://longbridge.com/en/quote/PSI.US.md) - [Invesco PHLX Semiconductor ETF (SOXQ.US)](https://longbridge.com/en/quote/SOXQ.US.md) - [Direxion Daily Semicondct Bull 3X ETF (SOXL.US)](https://longbridge.com/en/quote/SOXL.US.md) - [GraniteShares 2x Long MSFT Daily ETF (MSFL.US)](https://longbridge.com/en/quote/MSFL.US.md) - [Defiance Daily Target 2X Long ORCL ETF (ORCX.US)](https://longbridge.com/en/quote/ORCX.US.md) - [Direxion Daily AMD Bull 2X Shares (AMUU.US)](https://longbridge.com/en/quote/AMUU.US.md) - [Microsoft Corporation (MSFT.US)](https://longbridge.com/en/quote/MSFT.US.md) - [T-REX 2X Long NVIDIA Daily Target ETF (NVDX.US)](https://longbridge.com/en/quote/NVDX.US.md) - [Invesco S&P 500® Equal Weight Tech ETF (RSPT.US)](https://longbridge.com/en/quote/RSPT.US.md) - [GraniteShares 2x Long AMZN Daily ETF (AMZZ.US)](https://longbridge.com/en/quote/AMZZ.US.md) - [Oracle Corporation (ORCL.US)](https://longbridge.com/en/quote/ORCL.US.md) - [Global X Cloud Computing ETF (CLOU.US)](https://longbridge.com/en/quote/CLOU.US.md) - [State StreetSPDRS&PSftwr&SvcsETF (XSW.US)](https://longbridge.com/en/quote/XSW.US.md) - [GraniteShares 2x Long NVDA Daily ETF (NVDL.US)](https://longbridge.com/en/quote/NVDL.US.md) - [Invesco AI and Next Gen Software ETF (IGPT.US)](https://longbridge.com/en/quote/IGPT.US.md) - [Amazon.com, Inc. (AMZN.US)](https://longbridge.com/en/quote/AMZN.US.md) - [YieldMax NVDA Option Income Strategy ETF (NVDY.US)](https://longbridge.com/en/quote/NVDY.US.md) - [Direxion Daily MSFT Bull 2X Shares (MSFU.US)](https://longbridge.com/en/quote/MSFU.US.md) - [NVIDIA Corporation (NVDA.US)](https://longbridge.com/en/quote/NVDA.US.md) - [T-Rex 2X Long Microsoft Daily Target ETF (MSFX.US)](https://longbridge.com/en/quote/MSFX.US.md) - [Advanced Micro Devices, Inc. (AMD.US)](https://longbridge.com/en/quote/AMD.US.md) ## Related News & Research - [Microsoft-Backed OpenAI to Reportedly Start Showing Ads to Free, Go Users of ChatGPT in US](https://longbridge.com/en/news/280141022.md) - [AMD Eyes Major AI Win with 10,000-Chip Deal in Korea](https://longbridge.com/en/news/280108451.md) - [Persistent Systems launches AI-powered generative molecules and virtual screening solution powered by Nvidia](https://longbridge.com/en/news/279394554.md) - [Planet to build world’s first GPU-native AI engine for planetary intelligence with Nvidia](https://longbridge.com/en/news/279330399.md) - [LPU: Nvidia popularizes a new AI acronym](https://longbridge.com/en/news/279662346.md)