The $350 Billion Question: Why Franklin Templeton Is Forcing Nvidia's Capital Allocation into the Light

Sara Araghi, a voice from Franklin Templeton, just put Nvidia on notice. It's not about chip performance. It's not about the Blackwell roadmap. It's about capital. The call is simple: Nvidia must clarify its capital deployment plans. When a $1.6 trillion asset manager publicly urges the world's most valuable company to explain where the cash goes, this stops being a suggestion. It becomes a market signal. The market cap has crossed $3.5 trillion. The trailing P/E is north of 50. Every valuation model assumes 30% annualized growth for the next five years. That growth is not a product of engineering alone. It is a product of capital allocation. And right now, the allocation is a black box. The market has priced in the output of the machine but has no clear line of sight into the inputs. That mismatch is the story. Let's get the context right. Nvidia is not just a chip designer anymore. It is the financial engine of the AI infrastructure boom. Data center revenue hit $30.8 billion in Q3 FY2025, up 112% year-over-year, accounting for over 87% of total revenue. Gross margins are hovering in the 73% to 75% range. These are staggering figures. But they come with a hidden weight: the capital intensity of this business is unlike anything in the history of semiconductors. The scale of the build-out is unprecedented. This is where the tension lives. Nvidia's product cycle is in transition, moving from the Hopper architecture to the Blackwell platform. The next-generation chips, Blackwell Ultra and the Rubin platform, are set to define the next cycle of dominance. But the supply chain requires lock-step coordination with TSMC for CoWoS advanced packaging and SK hynix for HBM3e memory. Every one of these dependencies is a physical constraint on future revenue. Capital deployment is the mechanism that secures those constraints. If the deployment is unclear, the supply chain is uncertain, and the revenue is at risk. Franklin Templeton's public nudge suggests a deeper dissatisfaction with how Nvidia communicates. The pattern is clear: heavy on marketing, light on disclosure. It's been a common complaint since the GTC event in 2024. The institution is not asking for charity. It is asking for visibility. The question is not whether Nvidia will invest in AI. The question is whether the capital allocation will be efficient enough to sustain the valuation. The difference between those two things is the spread where value is made or lost. The market's focus is still on the revenue line. The smart money is focused on the balance sheet. The capital deployment plan will define the company's ability to defend its moat, not just its growth. I have seen this dynamic before in other contexts. When a company's growth is heavily dependent on external constraints, like supply chain access or fabrication capacity, the clarity of the capital plan is the true strategic document. The deck for the quarterly earnings call is just marketing. Here's the core of the issue. Nvidia's strategy is transitioning. It's no longer just a chip designer. It's building an "AI factory" model, moving from a pure hardware supplier to a platform for AI infrastructure. That transition requires massive, sustained capital. The question of whether to build its own data centers, like Google or Amazon, or continue to partner with cloud providers, is a binary choice with massive implications. If Nvidia owns the infrastructure, it competes with its own customers. If it doesn't, it leaves the profit center to the hyperscalers. The capital allocation decision is the answer to that question. Then there's the software play. The CUDA ecosystem, with over 5 million developers, is the deepest moat in tech. But moats require maintenance. The investment into software, libraries, NIM microservices, and the monetization of that stack is a long-term play. Investors are increasingly asking about the path to software monetization. The hardware sales are clear. The software revenue is opaque. This is another block in the black box. There's also a strategic investment angle. The reports suggest Nvidia is exploring investments in AI-native companies, potentially including a direct stake in OpenAI. The "compute-for-equity" model is a structural strategy, but it has a cost. It's not a pure investment. It's a way to lock in demand and keep the ecosystem tied to the CUDA stack. But the capital efficiency of that model is unclear. Are they buying revenue or buying a liability? Here's the contrarian angle. The market is treating this as a governance story, but it's not. It's a story about the limits of a single company's balance sheet. We assume that Nvidia can solve any problem with its cash. But the scale of the problem is the issue. Let's look at the math. The hyperscalers are spending over $200 billion combined per year on capex. Most of that flows to Nvidia. To maintain its dominant position, Nvidia must deploy billions in a coordinated manner to secure supply. If it fails to do so, the constraints are already visible. Consider the competitive landscape. AMD's MI300 series has reached 80-90% of H100 performance in some benchmarks, with a lower price. The self-designed chips from cloud providers, Google's TPU v5p, Amazon's Trainium2, and Microsoft's Maia 100, are starting to divert the high-end AI training workloads. The market share is not guaranteed. It is a function of the pace of innovation, and the pace of innovation is a function of capital. The faster they deploy, the faster the lead is extended. The slower the deployment, the wider the window for competitors to close the gap. Sentiment is the invisible ledger of value. And the sentiment is shifting from raw revenue growth to capital efficiency. The market has already priced in the growth. Now it's pricing in the delivery. The inability to clearly state the capital deployment plan is, in effect, a discount on the future. The market is beginning to realize that the story of AI is a story of capital velocity, not just chip architecture. The market will be watching the next earnings call. The signals are clear. We need to see the capital expenditure guidance. We need to see the share repurchase plan. We need to see the investment in AI startups. If Nvidia can articulate its capital deployment plan with the same precision it applies to its chip roadmap, the market will reward it. If not, the valuation will start to price in the uncertainty. We are at a critical juncture where the network of trust is built on code, not character. Nvidia's code is strong, but the capital architecture is its character. The current ambiguity creates a vulnerability. The risk is not that the company fails. The risk is that the market simply pays less for the future because it cannot see the path. The discount rate goes up. The multiple contracts. This is the new front line in the battle for the AI century. It's not just about the chip. It's about the chip, the software, the supply chain, and the capital that binds them together. The question is whether Nvidia can provide the clarity required to keep the institutional capital locked in. The market is waiting. The next earnings call is the delivery date. The question is whether the company can deliver more than just a chip. The question is whether it can deliver the certainty that the market demands. Speed is the only currency that never depreciates. But so is clarity. In the end, the market is not just pricing the product. It's pricing the plan.