Trillion-Dollar AI Arms Race: Moody's Warning—Bear Trap or Value Destruction?

The most valuable piece of information in this article is: Moody's rating action redefines AI capital expenditure from a "growth story" to a "credit risk event," providing secondary market investors with an anchor point to reassess the valuation logic of tech giants from a balance sheet perspective.

As PE practitioners, when we look at a company, we never just focus on revenue and profit growth rates, but on the ability to generate Free Cash Flow (FCF) and capital allocation efficiency. The issue pointed out by Moody's is precisely the core contradiction selectively ignored by the market in the AI narrative over the past two years: When investing more than 10% or even 20% of annual revenue into AI infrastructure each year, does the "quality" of these cash flows still support current P/E ratios above 30x?

1. Cash Bleeding: Repricing Valuation from "Money Printer" to "Cash Burner"

First, let's look at a set of data comparisons I organized (based on 2025 annual reports and sell-side consensus):

Company	Annualized AI CapEx (USD Billions)	% of Operating Cash Flow	Free Cash Flow (FCF) Trend	Current EV/EBITDA
Amazon	~90-100	55%	Declined quarter-over-quarter for 4 consecutive quarters	22x
Google	~70-80	45%	FCF expected to shrink by 30% in 2026	25x
Microsoft	~60-70	40%	CapEx exceeds D&A for the first time	28x
Meta	~40-50	50%	FCF halved YoY in 2025	20x

Behind these numbers lies the same logic: To capture the "trillion-dollar revenue opportunity" of AI, these companies are consuming their own cash at a pace unprecedented in corporate history. The "unprecedented AI spending" mentioned by Moody's essentially shifts the capital structure from "operating leverage" to "capital leverage"—assuming AI infrastructure brings future cash flows, but confirming hundreds of billions in depreciation and interest expenses every quarter.

For mature platforms like Amazon and Google, the valuation premium granted by the market over the past decade largely stemmed from their "asset-light" cash flow characteristics: Once investment stabilized, advertising and cloud businesses could generate nearly 100% marginal profit. Now, AI data center construction requires significant debt financing (Moody's focus on credit ratings is for this reason), causing the proportion of fixed assets on the balance sheet to skyrocket from around 10% to over 30%. This means Enterprise Value (EV) composition has shifted from "discounted cash flow" to "discounted asset replacement cost," the latter being extremely sensitive to interest rates and cost of capital.

2. Differentiation of Competitive Barriers: Who is Building a "Moat" and Who is Digging a "Trap"

From a PE perspective, AI capital expenditure is not synonymous with competitive barriers. The key is to look at three points:

• Is the CapEx Exclusive?: Microsoft invested in OpenAI, building vertical integration of algorithms + compute. Most of its CapEx is used for custom chips and private data centers, forming a "data flywheel" effect. Meta's "open-source LLM" strategy, however, essentially trades CapEx for ecosystem voice, but the replicability of open-source models lowers barriers, resulting in very low user switching costs.
• Depreciation Cycle vs. Monetization Speed: The depreciation cycle for GPUs is about 3-4 years, while the monetization speed of AI applications (like Copilot, Gemini) is far below expectations. Amazon's AWS has a clear revenue path on the AI inference side (token-based billing), but the incremental revenue brought by improvements in Meta's AI ad system currently cannot cover hardware investments.
• Balance Sheet Resilience: Amazon and Google have cash reserves exceeding $50 billion, and their cloud businesses generate stable cash flows, so a downgrade from AAA to AA might not be fatal. However, Meta's advertising revenue has entered a low-growth phase. After a significant reduction in FCF, if capital markets tighten, its refinancing costs will rise significantly.

My judgment is that the true competitive barrier is not "who built the most data centers," but "who can achieve a Return on Invested Capital (ROIC) on AI infrastructure that exceeds the cost of capital." Currently, only Microsoft and Amazon show signs of ROIC > WACC at the cloud service level. Google and Meta's AI investment returns remain below 10%, while the current cost of financing (Weighted Average Cost of Capital) is around 8%, leaving extremely thin margins.

3. Investment Judgment: Repricing Exit Paths

Returning to what PE cares about most: exit paths. For secondary market investors, the exit path is "the logic behind stock price movements in the next 6-12 months." Moody's warning provides a contrarian view: When credit quality becomes the market's focus, the valuation anchor for companies will shift from "revenue growth rate" to "cash flow generation capability." This implies:

• The market will pay more attention to metrics like "FCF margin" and "CapEx efficiency," rather than "

Original Link: https://www.cnbc.com/2026/07/24/moodys-ai-spending-credit-quality-amazon-meta-alphabet.html