Silicon’s Circular Moment: NVIDIA, OpenAI, Intel, and the Next Boom-Bust Cycle

(Op-Ed Analysis) The chip king bankrolls the hungriest AI lab; the lab commits to buy the chip king’s hardware at industrial scale; both sprinkle equity across the supply chain.

It’s audacious, elegantly aligned—and inherently fragile. We’ve seen this movie in railroads, telecoms, and dot-coms: the technology endures, the cycle doesn’t.

NVIDIA’s plan to invest up to $100 billion in OpenAI—paired with supplying millions of data-center GPUs to power roughly 10 gigawatts of new compute—turns AI infrastructure into a systems bet. Equity plus hardware isn’t just a purchase order; it’s an attempt to lock in the economics of the frontier.

Tightening the loop: suppliers, clouds, capital

OpenAI is stitching multi-gigawatt capacity with partners like CoreWeave, Oracle, and SoftBank. CoreWeave, in turn, orders NVIDIA systems while NVIDIA owns a stake there. On paper, incentives align and time-to-capacity compresses.

In practice, risk concentrates: if model monetization lags, the same network that amplifies growth can transmit stress—vendors, customers, and financiers pulled down in a single downdraft.

Enter Intel: hedge, rescue, or both?

NVIDIA’s roughly $5 billion stake in Intel, alongside government and outside capital, gives Intel time and NVIDIA optionality.

A successful Intel turnaround in foundry and accelerators becomes a friendly capacity hedge; a stumble still stabilizes a strategic U.S. supplier at modest cost. Either way, the optics—state support, cross-holdings, potential exclusivities—invite scrutiny.

The bull case: productivity tides

Generative and agentic systems look like general-purpose technologies. As inference embeds in code, design, support, and science, unit economics improve; subscriptions bundle AI; compute demand proves stickier than skeptics expect.

Valuations can compress without killing the build-out—like fiber after the dot-com bust. In this reading, circular ownership is a feature that speeds deployment and lowers transaction friction.

The bear case: circular finance meets hard limits

One vendor dominating a chokepoint magnifies any supply, legal, or security shock. Demand elasticity may bite: if enterprise ROI trails promises, pre-booked capacity turns into ballast.

Antitrust risk rises as cross-holdings blur lines between partner and competitor. And the physical world intrudes—power, land, cooling, and grid interconnects can’t scale at the speed of term sheets. Circular financing papers over frictions—until it can’t.

A necessary counterpoint

Critics add a simpler worry: feedback loops dull discipline. When capital, customers, and suppliers are partially the same entities, price signals blur and risk is underpriced.

That’s how sectors overbuild: tracks laid side-by-side in the 19th century, redundant fiber in the 1990s, cloud overspend in 2022–23. The assets often prove useful later; the equity holders pay the tuition.

What the last man in remembers

Every cycle mints real infrastructure—and punishes late capital. The lesson isn’t to sit out the revolution; it’s to price it correctly.

Separate compute that unlocks durable revenue from capacity built for leaderboard theater. Watch leverage. Assume power and networking, not chips alone, become the scarcest inputs.

A prudent stance for exuberant times

Policy should accelerate deployment—onshoring, diversified supply, cheaper inference—while resisting entanglements that cement single-vendor dependency.

Investors should favor cash-generative workflows over speculative scale. Operators should build for constraints: watts, racks, routes.

AI’s future looks inevitable; its cash flows won’t be linear. The circular deals of 2025 could be brilliant accelerants—or the kindling for the next reset. The smart play is to ride the technology—and hedge the cycle.