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 US AI sector is evolving a highly complex and interwoven &#34;new financing structure&#34; and capital internal circulation system.
On March 23, according to news from the Chasing Wind Trading Desk, 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 severe 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 a capital flow diagram involving multiple parties such as Nvidia, Microsoft, Oracle, CoreWeave, Amazon, AMD, and Disney.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a15ac030-ed09-42b8-bd36-673ab4ace3ac.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a15ac030-ed09-42b8-bd36-673ab4ace3ac.png"/>
Morgan Stanley points out that the current AI boom is built on a 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 real economic leverage and risks off the balance sheet, potentially leading to &#34;sky-high contracts&#34; in the headlines being overestimated.
<h2>&#34;Suppliers support customers, customers support suppliers&#34; — A panoramic view of the circular financing structure</h2>
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:
<blockquote>
<ul>
<li>Supplier financing and favorable terms</li>
<li>Long-term procurement commitments (Take-or-Pay contracts)</li>
<li>Revenue-sharing arrangements</li>
<li>Supplier buyback agreements</li>
<li>High customer concentration</li>
<li>Third-party guarantees and endorsements</li>
<li>IP licensing in exchange for model access</li>
<li>Equity investments in exchange for computing power commitments</li>
</ul>
</blockquote>
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:
<blockquote>
These financing agreements may accelerate data center construction while also potentially overextending future demand and reallocating risks among counterparties.
</blockquote>
<h2>Core Case: The Hundreds of Billions in Capital Flow within the OpenAI Ecosystem</h2>
Morgan Stanley states that, using 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 Dominator NVIDIA
<ul>
<li>Direct Investment: In February 2026, NVIDIA finalized plans to invest $30 billion in OpenAI.</li>
<li>Supporting Core Customers: NVIDIA holds approximately 9% of cloud service provider CoreWeave (CRWV) shares (valued at about $4 billion) and added a $2 billion investment.</li>
<li>Buybacks and Leases: 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 lease obligations (mainly for data centers).</li>
</ul>
2. Microsoft's Comprehensive Penetration
<ul>
<li>Equity and Computing Power Commitment: Microsoft's $13 billion investment in OpenAI is currently valued at approximately $135 billion (about 27% on a fully diluted basis). OpenAI has committed to purchasing an additional $250 billion in Azure services.</li>
<li>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 amount to $155.1 billion.</li>
</ul>
3. Oracle and AMD's Huge Orders
<ul>
<li>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 in computing power from Oracle over approximately five years. Oracle's data center leasing commitments amount to $261 billion.</li>
<li>AMD: Announced a strategic agreement worth $100 billion with OpenAI (deploying 6GW of AMD GPUs). As part of the agreement, AMD issued warrants for up to 160 million shares to OpenAI.</li>
</ul>
4. Amazon and Disney's Cross-Industry Entry
<ul>
<li>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 uninitiated leases.</li>
<li>Disney: Invested $1 billion in OpenAI (with additional 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 power 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 vulnerabilities that this &#34;left hand to right hand&#34; capital cycle may bring when pricing for the AI frenzy.</li>
</ul>
At the same time, Morgan Stanley also clearly outlined seven potential risks arising from this cyclical structure in its research report:
<blockquote>
<ul>
<li>1. Warrants Distort True Pricing</li>
</ul>
Customers exchange long-term procurement commitments for supplier warrants, making the contract headline amounts unable to reflect the repeatable true pricing level.
<ul>
<li>2. Off-Balance-Sheet Guarantees Hide True Leverage</li>
</ul>
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 serious divergence between reported leverage and true economic leverage.
<ul>
<li>3. IP Licensing Masks True Operating Costs</li>
</ul>
Content creators authorize IP in exchange for AI model access under favorable or non-cash terms, leading to an underestimation of the true operating costs and cash needs of AI labs.
<ul>
<li>4. Supplier Equity Investments Amplify Debt Risks</li>
</ul>
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 reaction that further drives capacity expansion.
<ul>
<li>5. High Customer Concentration Amplifies Counterparty Risk</li>
</ul>
Revenue growth increasingly relies on the success of a few AI labs, significantly increasing concentration risk.
<ul>
<li>6. Revenue Sharing Arrangements Obscure True Demand</li>
</ul>
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.
<ul>
<li>7. Repurchase Agreements May Inflate Demand</li>
</ul>
Supplier repurchase agreements transfer risk back to suppliers, providing customers with downside protection, which may artificially inflate apparent demand figures.
</blockquote>
<pre><code>
The above exciting content comes from the Wind Trading Platform.

For more detailed interpretations, including real-time analysis and frontline research, please join the【 **Wind Trading Platform ▪ Annual Membership**】

![](https://wpimg-wscn.awtmt.com/3c4a713c-7a38-4582-9850-d0eabaf0e7ad.png)

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk</code></pre>

MSSM

FNCL

Morgan Stanley's latest research report points out that the capital intensity of the AI industry in the United States has reached an unprecedented level, and it is expected that AI investment will account for about 50% of the total capital expenditure of large-cap stocks in the coming years. Outside of hyperscale cloud service providers, the capital of the entire ecosystem remains constrained, giving rise to complex "new financing structures." These structures include financial mutual assistance between suppliers and customers, which may lead to overestimated contract pricing

- Morgan Stanley warns that AI industry capital intensity is reaching unprecedented levels as investment cycles accelerate.  
- The report highlights a mismatch between capital demand and AI revenue realization, leading to the emergence of complex financing structures within the ecosystem.  
- Key findings emphasize that financing mechanisms allow many participants to expand infrastructure beyond their cash flow capabilities while posing risks of overestimated contract valuations and future demand.

In-depth Analysis: The "Capital Internal Circulation" in the American AI Circle