A portfolio manager at Franklin Templeton just told Nvidia to explain its capital deployment plans. Publicly. That's not a casual comment. That's a shot across the bow from one of the world's largest asset managers, a firm that oversees more than $1.6 trillion in assets. The market barely reacted.
The lack of reaction is the problem.
Nvidia's market cap sits above $3.5 trillion. Its price-to-earnings ratio is hovering around 50x. This valuation does not exist in a vacuum—it is built on the assumption that Nvidia will continue to dominate the AI infrastructure boom for years. Yet the company's own financial disclosures are opaque about how it will allocate its capital to defend that position. That's a structural mispricing. And no one is pricing in the risk.
Franklin Templeton's Sara Araghi asked Nvidia to clarify its capital deployment plans. The request is simple, logical, and overdue. Nvidia has over $30 billion in cash sitting on its balance sheet. The company is generating enormous free cash flow from its data center business, which posted $30.8 billion in revenue in Q3 of FY2025, up 112% year-over-year. That's not a small amount of money.
The market's response has been to treat this as noise. I don't. Based on my years of tracking supply chains and on-chain flows, this is the kind of request that reveals the cracks in a narrative long before the narrative breaks.
The Context: This Isn't About Marketing
Franklin Templeton isn't a retail investor. They don't buy the keynote speech. They buy the 10-K. And their public request suggests the internal modeling is hitting a wall.
Here's what the balance sheet is showing. Nvidia's business is overwhelmingly dependent on data center revenue, which accounts for over 87% of total sales. The gross margins on those sales are extraordinary—somewhere between 73% and 75%. But maintaining those margins requires a massive, ongoing capital commitment to the ecosystem.
Nvidia is transitioning from the Hopper architecture to Blackwell. This isn't a small refresh. It's a full supply chain overhaul involving advanced packaging (CoWoS) and HBM3e memory. Capital deployment priorities determine whether Nvidia can lock in supply agreements with TSMC and SK hynix ahead of competitors.
But Nvidia hasn't provided the granularity the market needs. How much is being spent on vertical integration? How much is going to strategic investments in AI startups like CoreWeave or potentially OpenAI? How much is allocated to the CUDA software ecosystem, which has a longer return cycle?
These aren't idle questions. The market is currently pricing Nvidia as a pure chip monopoly, but the company is moving toward becoming a capital-intensive AI infrastructure platform. Those are two very different valuation models. The transition between them requires absolute transparency about capital allocation. Without it, the valuation is just a guess.
The Core: The $200 Billion Question
This isn't just about Nvidia. The company's capital deployment is a macro signal for the entire AI industry.
Hyperscalers—Microsoft, Google, Amazon, Meta—are planning to spend over $200 billion in combined capex this year. A significant portion of that flows directly into Nvidia GPUs. These purchasing decisions depend on Nvidia's ability to deliver, which depends on Nvidia's own capex.
Here's the breakdown of what needs to be clarified.
First, the supply chain. GPU lead times have stretched from 3 months to 6-12 months. That's a logistical bottleneck. If Nvidia isn't publicly committing to expanding CoWoS capacity or HBM supply through long-term contracts, the market cannot accurately predict when supply will catch up with demand. This directly impacts cloud compute prices and the profitability of AI startups.
Second, the competitive landscape. The window of Nvidia's dominance is not closed. AMD is pushing its MI300 series, reaching 80-90% of H100 performance in some benchmarks, at lower price points. Google and Amazon are accelerating the adoption of their own silicon (TPU v5p and Trainium2). They're all working to reduce their dependency on Nvidia. If Nvidia's capital deployment signals uncertainty, it will accelerate the shift toward alternatives.
Third, the software ecosystem. CUDA has over 5 million developers, which is a strong moat. But this is only maintained through continuous investment. The capital allocation between hardware and software is not fully disclosed. If the market doesn't know how much is being spent on the software ecosystem, it's impossible to assess the durability of that moat.
Fourth, the data center strategy. Nvidia is pushing the "AI factory" concept. Are they planning to build their own massive data centers, like Google and Amazon? Or will they continue to partner with cloud providers? The answer fundamentally changes their margin structure. Data centers are high capex, high overhead, and lower margin than selling chips. Investors need clarity on this trajectory.
From my audit and trading experience, I've seen this pattern before. When a company's growth relies on a self-reinforcing narrative—GPU shortage, demand explosion, ecosystem lock-in—the cracks appear when that narrative hits a capital constraint. The constraints usually start with questions from institutional investors, not from retail.
The Contrarian View: The Real Risk Is Capital Mispricing
Here's what the market is missing. The biggest risk to Nvidia's valuation isn't a drop in demand. It's the misallocation of capital.
The bear case isn't that AI is a bubble. The bear case is that Nvidia will be forced to spend heavily in areas that don't immediately generate returns—like vertical integration or AI startup investments—to maintain its dominance. This will pressure free cash flow and squeeze the margins. If Nvidia's internal investments don't translate to tangible revenue growth, the 50x P/E will look irrational.
There's another possibility. Institutional demand for clarity could be a proxy for a broader concern about the entire AI infrastructure market's capital efficiency. If Nvidia's own capital planning is opaque, it signals that the entire industry's massive investment cycle might not be as well-reasoned as it appears. This is a systemic concern, not just a company-specific one.
The potential for an antitrust review is also being ignored. Nvidia's strategy of investing in downstream AI companies, like CoreWeave and Together AI, could be seen as anti-competitive. Regulatory focus on these investments could slow the growth strategy and impose additional costs. This is a low-probability, high-impact event that's not being priced in.
The Takeaway: A Catalyst for Repricing
Here's what I'm watching. Nvidia's next earnings call and GTC 2026. If they deliver a clear capital allocation framework—showing concrete plans for supply chain, software, and the AI factory model—the current uncertainty will resolve itself. The stock will continue to move forward.
But if the answer is a vague "we're investing in all the right areas," the market will interpret that as weakness. And then, this not just an Nvidia problem. It's an AI infrastructure problem.
I don't trade on hope. I trade on the supply chain and the capital flows. And right now, the market is betting on certainty that hasn't been delivered yet. The market doesn't care about the keynote. It cares about the capital structure.
Franklin Templeton isn't asking for a favor. They're asking for the data they need to calculate the fair price. Nvidia's response to this question will determine whether the AI bull case has a solid foundation or if it's just building on a massive, well-marketed claim.
I'm watching the earnings call. Not the buzz. The numbers.
If the clarity isn't there, the repricing risk is real. And the market won't be able to say it wasn't warned. This is the season where capital allocation is the alpha. Everything else is just noise.