The Loop and the Black Hole III

Semiconductors, AI, and the Arithmetic of Exuberance

IV. The Loop and the Black Hole

The metaphor that best captures the dynamics of the industry is not that of a conventional bubble. It is that of a closed loop — or, in astrophysical terms, a black hole with accretion disk, that is, a region of extreme gravity from where nothing escapes, and a spiral of gas and dust that, before being swallowed, due to friction, it overheats and emits a spectacular shine.

The basic circuit is as follows: OpenAI, Microsoft, Google, and Meta need GPUs to train increasingly larger models. These GPUs are almost exclusively from Nvidia, which charges gross margins above 70%. Big Tech companies fund these purchases using existing revenues or investment rounds driven by the promise of future AI. Nvidia’s market cap rises, validating the narrative. Seeing Nvidia’s stock soar, Big Tech companies convince the market that the investment is justified, raise more capital, and buy more GPUs. The output of one (GPUs) is the input of the other (models), and the financial validation of one supports that of the other. In this circuit, there is no flow of value that ends up in a bakery, a textile factory, or a hospital in a transformative way. Money circulates among the same agents: it is a financial resonance that can be real without the economic radiation being proportional.

This loop contrasts with previous major technological leaps. The electrification of the early 20th century brought generators and motors to hundreds of different industries—textiles, steel, railroads—and to domestic consumers. The internet of the 1990s–2000s, despite the dotcom bubble, left fiber infrastructure and tools that were absorbed by millions of SMEs and created new markets. The semiconductor-plus-AI complex concentrates a disproportionate share of investment in a handful of companies that sell to one another and to their own cloud operations. The "spillover" to the rest of the economy is slow, narrow, and mediated by APIs. Never has a technology been simultaneously so central and so enclosed.

The black hole metaphor fits in five dimensions:

Event horizon. The valuations (41.7× GVA) only make sense within the loop itself. An outsider sees a gravitational anomaly. An insider finds a coherent internal logic: each asset appears justified because they all validate one another. The entire system may be overvalued without any single company appearing so.

Accretion of matter. The loop absorbs capital (investment, liquidity via balance-sheet expansion—a variant of QE), talent (scarce engineers and PhDs), and physical resources (electricity, rare earths, ultrapure water for fabs). The rest of the economy may wither, much like a companion star losing mass to a central object. Total factor productivity data—which have slowed in most developed countries since 2005 despite digital advances—suggest that the loop’s absorption of resources does not translate into broad-based growth.

Internal singularity. At its center lies the promise of Artificial General Intelligence (AGI). The entire spiral of value converges on a point where AI is supposed to solve every problem and generate infinite profits. For now, this point is a mathematical singularity—unobservable and potentially unattainable.

Spaghettification. As concentration increases, any disturbance stretches the value chains to their limit.The systemic fragility is enormous because there is no redundancy.A single dominant GPU designer (Nvidia), a single extreme lithography supplier (ASML), a single high-end manufacturer at scale (TSMC). A shock to the weakest link propagates throughout the entire loop.

The accretion disk. By maintaining low interest rates and injecting liquidity into the system, central banks provide the energy that keeps the disk glowing. With zero or negative real interest rates, investors are forced to seek returns in long-duration assets—quintessentially, tech stocks. This compresses the discount rate and inflates multiples, fueling the loop. Halting the model of managed structural liquidity would mean stripping matter from the disk: the black hole would reveal its dark core.

The "Magnificent Seven" already account for approximately 30% of the S&P 500. Their implosion would impact pension funds, insurance companies, and global financial stability. In this sense, the loop is "too big to fail"—not because the state wants to save it, but because it cannot bear the losses that would result from letting it collapse.

V. Three Evaporation Scenarios

Given the dynamics described, what are the realistic scenarios for a breakdown? Three are identified—with distinct triggers, sequences, and consequences.

Scenario 1 — The Disappearing Singularity (Stagnation of Scaling Laws)

Trigger. The next generation of large language models does not deliver disruptive gains in capability. Marginal investment in GPUs no longer yields proportional increases in revenue. Big Tech companies are subtly acknowledging that "AGI via raw scaling" is a mirage.

Sequence. Demand for GPUs contracts. Nvidia’s revenue stagnates at ~$45 billion (down from the current $60 billion), and its net margin falls to 30% due to competitive pressure—specifically from Google’s proprietary chips, AMD, and specialized startups. Annual profit drops to ~$13.5 billion. Without the narrative of 20% annual growth, the market reprices the stock: a P/E ratio of 25 (still generous for a cyclical sector) implies a market cap of ~$337 billion—an 89% decline from the current $3 trillion. If we project this same multiple contraction across the entire semiconductor sector—where the market cap would retreat from $5 trillion to $350 billion—we see a 93% drop.

An engineer earning $300,000 a year—half in stock—sees the value of their compensation package plummet by over 40%. In such a scenario, these professionals can no longer cover their fixed expenses (which can range from $120,000 to $150,000 annually in the Bay Area). The best among them migrate to sectors where cash is king—defense, energy, and traditional finance. The innovation that sustained the cycle withers for lack of talent. Index funds and pension funds, saturated with tech stocks, lose value. The negative wealth effect triggers a consumer recession in the US. The market cap-to-salary ratio—currently 58.1x—aligns with that of the transportation sector: ~2.5x. The stock-as-currency model vanishes.

Scenario 2 — Hawking Radiation (External Technological Leap)

Trigger. Just as the "bodies" emitted by the loop lose mass until evaporation in Hawking radiation, a radically new architecture—optical computing, third-generation neuromorphic chips, or a breakthrough in quantum computing with commercial applications—renders traditional GPUs obsolete. The holder of the new technology is not Nvidia.

Sequence. Nvidia loses its monopoly on AI hardware. Its GPUs become commodities relegated to residual tasks, much like mainframes did. Big Tech firms migrate to the new hardware. TSMC and ASML lose relevance and falter because the new paradigm dispenses with EUV lithography. The loop breaks—yet capital shifts to a different corporate constellation, just as energy escapes a black hole via Hawking radiation, particle by particle, until it evaporates.

This is the least painful scenario from a systemic perspective. Talent is rapidly absorbed by the new industry, which pays high salaries in its own stock. Central banks can accommodate this transition without triggering an aggregate demand crisis. It is Schumpeterian creative destruction in its purest form—and perhaps the only scenario in which the promises of the "loop" are fulfilled, albeit in different hands.

Scenario 3 — The Collapse of the Accretion Disk (Monetary and Energy Choke)

Trigger. A geopolitical crisis (such as a blockade of the Taiwan Strait) or persistent inflation driven by supply shocks forces central banks to raise benchmark rates to 4.5–5% and to halt—fully or partially — liquidity injections via balance sheet expansion. This scenario is not without recent historical precedent: the 2022–2023 cycle already provided the template—the Fed raised rates from 0.25% to 5.25% over 16 months (the most aggressive tightening cycle in forty years), and the Nasdaq fell 33% in 2022. What this scenario adds to the template is a relapse: a new inflationary wave that prevents monetary easing before the "loop" narrative re-emerges, as occurred in 2024–2025.

Sequence. With US public debt exceeding 120% of GDP, a benchmark rate of 4.5–5% marks the threshold where sovereign debt service begins to structurally compete with any other fiscal priority — leaving the Fed cornered between fighting inflation and avoiding a collapse of public finances. Markets interpret this predicament as an inability to return to the "monetary morphine" of the past; it is this fracture in the narrative — rather than the absolute level of interest rates — that triggers the adjustment. The discount rate implicit in long-duration growth assets, artificially compressed to around 8% during the years of balance-sheet expansion, rises to 10–11%. In a Gordon growth model, the equilibrium P/E ratio for an asset with 5% terminal growth drops from 33x (at 8%) to 17–20x (at 10–11%). For Nvidia, whose valuation premium demands multiples of 35–50x reported earnings, the adjustment is non-linear: it is not a gradual multiple compression but an abrupt re-evaluation of a monopoly premium that no longer finds a buyer at the previous price. Semiconductor market capitalization corrects to $700 billion–$1 trillion—an 80–85% contraction from the current $5 trillion—and the shockwave ripples through the cloud, SaaS, and digital advertising sectors.

Companies cannot sustain astronomical payrolls without overvalued stock. Layoffs cease to be isolated "restructuring" events and become structural cuts affecting 20–30% of the workforce. The loss of stock market wealth and rising unemployment among highly skilled workers trigger a contraction in consumption that extends far beyond the tech sector. Central banks' initial reaction will be to cut rates and resume balance-sheet expansion—but confidence in the narrative of eternal expansion has been shattered. "Too big to fail" proves to be "too big to save": the State can avert bankruptcy, but it cannot restore the illusion of exponential profits.

The economy is trapped in technological stagflation: low rates sustain "zombie" firms, yet aggregate productivity fails to rise because innovative investment requires optimistic equity—and that has evaporated. The wage share—at 72% of Gross Value Added—once a bastion of strength, has turned into a trap: fixed costs associated with skilled labor render companies rigid and unable to adapt to a slow-growth regime.

Methodological note:

¹ Market capitalization figures refer to the leading US-listed companies in each sector (mid-2024): Nvidia, Broadcom, AMD, and Intel for semiconductors; Alphabet, Meta, Microsoft (partially), and Netflix for internet/software; Eli Lilly, Pfizer, and Merck for pharmaceuticals; ExxonMobil, Chevron, and NextEra for energy; ADM, Coca-Cola, PepsiCo, and Tyson for agri-food; UPS, Union Pacific, and Delta for transportation. Gross Value Added (GVA) and wage figures are derived from the Bureau of Economic Analysis (US, 2022, current prices), using the corresponding NAICS codes (3344 for semiconductors; 519+5112+5415 for internet/software; 3254 for pharmaceuticals; 211+324+22 for energy; 11+311+312 for agri-food; 48-49 for transportation). All figures are rounded.