The Ghost in the GPU: Tobin’s q and the Illusion of Investment

These articles are not designed to offer definitive answers or fixed positions. Instead, they are explorations—reflections grounded in history, data, and evolving thought. Our aim is to surface questions, provide context, and deepen understanding. We believe education thrives not in certainty, but in curiosity.

Cold Start in a Hot Factory

In the summer of 2024, a private message circulated inside a clean room in Hsinchu, Taiwan. The email, curt and catastrophic, came from an executive at ASE Group: “No CoWoS substrate availability until Q1 next year. NVIDIA H100 delayed.”

At NVIDIA’s headquarters in Santa Clara, urgency turned to paralysis. Factory lines in Oregon, Singapore, and Frankfurt slowed to a crawl. Cloud executives, whose data centers were being retrofitted for generative AI, found themselves staring at empty server racks. The missing piece wasn’t the high-performance chip itself, but a specialized resin film—made by just one supplier in Japan: Ajinomoto. No ABF substrate, no interposer. No interposer, no GPU.

It wasn’t war. It wasn’t sabotage. It was capital misalignment—a story James Tobin had seen long before silicon began thinking.

A Mind Torn Between Markets and Machines

The year was 1965, and James Tobin, then a Yale professor, was haunted by a question: Why did firms often fail to invest, even when flush with capital?

Tobin wasn’t an ideologue. He was a technician—a man who once likened macroeconomic modeling to “trying to do surgery with a saw.” But by the mid-60s, even his precise mind couldn’t ignore the disconnect between Wall Street optimism and Main Street stagnation.

He had watched the Kennedy tax cuts push corporate profits skyward. He had seen the rise of financial conglomerates and watched as capital zipped across continents. Yet, real investment lagged. Plants weren’t being built. Machines weren’t being upgraded. The gears of productive capacity were rusting while stock tickers gleamed.

To understand this dislocation, Tobin went to war with a dominant idea: that firms were indifferent to how they financed themselves.

Standing on the Shoulders of Ghosts

The Modigliani-Miller theorem, born in the late 1950s, had hypnotized the economics profession. It argued that under certain assumptions—no taxes, no transaction costs, rational actors—a firm's value was independent of its capital structure. Debt or equity? It didn’t matter. What mattered was expected cash flow.

To Tobin, this elegant irrelevance smacked of unreality.

He had spent time in the Navy, watched factories turn into engines of war, and lived through the Great Depression. He knew that where capital came from—and where it was allocated—mattered profoundly. He was steeped in the intellectual residue of Keynes, whom he admired not for precision, but for judgment.

Tobin wanted a model that didn’t just describe equilibrium—it needed to inspire investment. He believed that markets had a duty not just to reflect value but to guide capital to its most productive use.

The Equation That Emerged From the Fire

• If q > 1, then the market values a firm’s assets more than it costs to reproduce them. New investment should follow.

• If q < 1, then it's cheaper to buy existing assets than build new ones. Firms will hold back. Investors will speculate. Bubbles may form.

Housing Analogy

• If the market price of a house is more than it costs to build (q > 1), then it makes sense to build more houses—you’ll make a profit.
   → Companies should invest and expand.
  

• But if the market price is less than it costs to build (q < 1), then it’s cheaper to buy an old house than build a new one.
   → Companies stop investing. Speculators step in. Bubbles can form.

Tobin’s equation was more than a valuation tool—it was a diagnostic for capitalism’s pulse. When investment failed to respond to high q, something was broken: trust, incentives, or capacity.

From Yale to Hsinchu: The Ghost of Tobin’s q

Fast forward six decades, and Tobin’s fears have become prophetic. Consider the NVIDIA H100—a chip that has become the heart of the AI revolution.

The H100’s value is sky-high. It sells for $30,000–$40,000 on the open market. Demand outpaces supply by an order of magnitude. Market valuations for NVIDIA and its ecosystem—TSMC, ASML, SK Hynix—have soared.

And yet, capacity lags.

🌐 Let’s break down the supply chain:

Each node is a potential fracture.

• TSMC and ASE control 80% of advanced packaging.

• Ajinomoto supplies the only resin used in interposer substrates.

• ASML is the sole maker of EUV tools needed for advanced nodes.

• HBM3 memory has a 52-week lead time.

• Geopolitical exposure in Taiwan, Japan, and Korea introduces existential fragility.

Even as market value soars, replacement capacity crawls. Capital isn’t flowing fast enough to relieve pressure points. There is no functional “q” guiding real investment. The ratio has become a relic—observed, not obeyed.

Inflated Prices, Hollow Capacity

In Tobin’s framework, when q > 1, the logic is clear: build.

But in today’s world, firms often don't.

Why?

• Geopolitical Risk: No firm wants to double down on fabs in Taiwan with war clouds overhead.

• Regulatory Bottlenecks: Building a new lithography line takes years—and geopolitical clearance.

• CapEx Fatigue: TSMC and SK Hynix operate at wafer-thin margins and already run near full utilization.

• Software Lock-in: NVIDIA’s CUDA stack locks the AI ecosystem into a proprietary platform. Even if chips could be built, they’re not portable.

So instead of building, we bid up. Valuations rise. Expectations inflate. But the real economy remains stranded.

Tobin’s Quiet Rage—and Final Hope

James Tobin didn’t live to see GPUs or deep learning. He died in 2002, just before Moore’s Law hit the wall. But he would have seen in the AI boom the same pathology that plagued the 1960s: a market unmoored from its physical base.

He would have asked:
Why are firms not investing in redundancy?
Why are governments not subsidizing capacity?
Why does the stock price rise when a firm cancels its CapEx plan?

He would not have been surprised.
But he would have been disappointed.

Resolution: The Ratio We Ignored

Tobin’s q was never meant to be an academic toy. It was a call to stewardship—a reminder that the market’s job isn’t to guess future value, but to guide it.

In an age of AI, fragility, and speculation, we have forgotten the signal in the noise.

It’s time to remember the ratio that saw the future.

🧠 Further Reading & References

1. Tobin, J. (1969). A General Equilibrium Approach to Monetary Theory. An essential read where Tobin outlines the intellectual foundations of q in a macro context. Link

2. Modigliani, F., & Miller, M. (1958). The Cost of Capital, Corporation Finance and the Theory of Investment. The origin of the capital structure debate.

3. Keynes, J.M. (1936). The General Theory of Employment, Interest and Money. Tobin’s intellectual anchor.

4. Deloitte (2023). AI Semiconductor Supply Chain Fragility Report. Offers a current lens on GPU bottlenecks.

5. Semianalysis (2024). TSMC, NVIDIA, and the Future of AI Packaging. A deeply technical dive into modern chip supply. Link