NVIDIA locks in 3.2 billion in production capacity! Is this the breakout moment for this industry?

Recently, NVIDIA and Corning reached a strategic cooperation.

According to the agreement, Corning will increase its optical connection manufacturing capacity in the United States by 10 times, with fiber optic production expanding by over 50%.

To achieve this goal, Corning will build three advanced manufacturing plants in North Carolina and Texas.

To support Corning’s capacity expansion, NVIDIA also signed a securities purchase agreement with Corning, granting NVIDIA the right to invest up to $3.2 billion in Corning, including $500 million from pre-paid warrants, and an option to purchase an additional $2.7 billion worth of Corning stock.

After the announcement, Corning (NYSE: GLW) stock price surged significantly.

This cooperation is not just a simple commercial procurement; it marks NVIDIA’s first long-term multi-billion-dollar order to lock in upstream fiber optic capacity, signaling that fiber optics have evolved from AI infrastructure components into strategic scarce resources.

01 Who controls the pricing power in the fiber optic industry?

The fiber optic cable industry follows a strict “rod-fiber-cable” pyramid structure, where technological barriers, profit margins, and pricing power decrease step by step from top to bottom. The true determinant of industry dynamics is at the top of the pyramid—the fiber preform rod segment.

In terms of value distribution, the fiber rod is the absolute core of the entire industry chain, accounting for about 70% of fiber optic production costs and capturing around 70% of industry profits.

The midstream fiber drawing process has lower technical barriers and profit share, representing about 20% of the industry’s profit; the downstream cable assembly segment has the lowest technical barriers and most competitive market, accounting for only about 10% of profit.

The core of technological barriers lies in fiber rod manufacturing. Producing fiber rods involves complex chemical vapor deposition processes, with impurity levels controlled at parts per billion. Slight deviations can cause fiber loss to fail standards, making it impossible to meet communication transmission requirements.

More critically, expanding fiber rod capacity requires customized, self-developed equipment, strict hazardous chemical approvals, and the construction of ultra-clean, Class 100 cleanrooms. Even with fully mature technology, the full capacity expansion and yield ramp-up cycle can take 18-24 months.

This means that even if fiber prices surge now, manufacturers cannot quickly release new capacity in the short term, creating a hard constraint on supply.

Globally, the fiber industry shows a pattern of “China-led capacity, significant concentration at the top.” The global nominal annual fiber capacity is about 550-600 million core kilometers, with over 60% of capacity in China.

Domestic giants include Long Fiber, Hengtong Optoelectronics, Zhongtian Technology, and FiberHome, which together hold over 70% of the domestic market share and nearly 50% of the global market.

In overseas markets, companies like Corning, Fujikura, and Sumitomo Electric dominate. Corning leads in high-end specialty and hollow-core fibers, firmly holding the North American market; Fujikura and Sumitomo have stable shares in high-end telecom and data center markets.

Notably, overseas manufacturers are more restrained in capacity expansion. Even amid current supply-demand tightness, the additional capacity from Corning’s cooperation with NVIDIA will only gradually come online by the end of 2027 and reach full production in 2028. Effective global supply will hardly see significant short-term increases.

The industry’s competitive logic has shifted from cost and scale competition to a comprehensive contest over control of the entire industry chain.

Only companies with independent control over fiber rod technology, capable of large-scale production of special fibers, and able to secure long-term orders from global cloud giants will hold the advantage in this cycle. The “winner-takes-all” effect becomes increasingly evident.

02 AI-driven demand continues to widen the supply-demand gap

This cycle of fiber optic industry prosperity is not just short-term price fluctuations but a long-term trend driven by fundamental changes in demand structure and supply-demand mismatch.

The primary engine is the exponential growth in demand driven by AI computing infrastructure.

Unlike traditional data centers, AI computing centers handle “east-west traffic” of parameter synchronization between GPUs, requiring 1:1, non-blocking network architectures, which leads to an exponential increase in fiber consumption.

This demand surge is already reflected in the capital expenditures and orders of cloud providers.

According to capital expenditure guidance from Microsoft, Google, Amazon, and Meta for FY2026, total capital spending will jump from about $380 billion in 2025 to approximately $650 billion in 2026, a year-over-year growth of 70%. Most of this increase will be invested in AI infrastructure.

Based on publicly available industry data, in 2026, the fiber demand from data centers in China and the US alone will exceed 50 million core kilometers, accounting for over 10% of the global traditional demand market.

More importantly, high-end fibers required for AI scenarios have prices and gross margins far higher than ordinary fibers. To pursue higher profits, manufacturers are shifting limited fiber rod capacity toward specialty fibers, further squeezing supply of regular fibers and intensifying industry tightness.

The second major driver is the demand from fiber-guided drones, an often-overlooked incremental market.

In military applications, fibers have absolute advantages such as resistance to electronic interference, high bandwidth, low latency, and no electromagnetic signal exposure, making them the core consumables for fiber-guided drones.

A single drone typically consumes 5-20 kilometers of fiber, and the fibers are non-recoverable after mission completion, turning fiber from a “century-long infrastructure product” into a “disposable military consumable.”

Industry estimates suggest that the current global annual demand for fiber in fiber-guided drones has reached about 50 million core kilometers, roughly 10% of total annual fiber demand. As penetration of fiber-guided drones continues, this demand will further grow.

Traditional demand is also rebounding. The increase in procurement prices for telecom operators’ bulk purchases confirms a fundamental shift in supply-demand dynamics. In 2026, domestic telecom operators’ procurement prices for ordinary cables are generally over 30% higher than in 2025, shifting industry bargaining power toward sellers, with a focus on “ensuring supply, delivery, and quality” rather than just lowest prices.

03 What core opportunities should be seized during the fiber optic prosperity cycle?

In the AI wave of investment, the most reliable opportunities are those “core infrastructures that everyone must use regardless of who wins.”

Fiber optics are exactly such infrastructure—whether it’s NVIDIA or AMD GPUs, Meta or Microsoft AI data centers, they all rely on fiber for high-speed data transmission.

Many ask, isn’t the fiber industry a typical cyclical sector? What’s different about this cycle compared to the past?

The answer is simple: the underlying driving logic of this cycle is completely different from previous ones.

The 2016-2017 fiber bull market was driven by a one-time surge in demand from domestic fiber-to-the-home (FTTH) construction, mainly concentrated among China’s three major operators. When household broadband penetration peaked, the industry fell into years of price wars and overcapacity.

This cycle, however, is driven by the long-term trend of global AI computing infrastructure. Demand is no longer limited to domestic operators. On the supply side, after a deep capacity cleanup in the last cycle, leading manufacturers are cautious about expansion, and the rigid supply constraints from fiber rod capacity make short-term supply gaps difficult to fill.

Considering the industry’s technological barriers, demand elasticity, and growth certainty, we identify three main investment themes worth focusing on in this cycle.

The first is industry leaders with full vertical integration of fiber rod, fiber, and cable.

Fiber rods are the industry’s core barrier and the biggest beneficiary of recent price increases. Only manufacturers with fully autonomous fiber rod technology can fully enjoy the profit elasticity from this price rise.

The second is companies with technological advantages in specialty and hollow-core fibers.

Specialty fibers command higher prices and margins, representing high-value-added segments. Domestic pioneers in mass production and commercialization of hollow-core fibers will gain first-mover advantages in next-generation technology.

The third is companies with strong international expansion capabilities and the ability to circumvent trade barriers.

Due to anti-dumping and tariff policies, domestic nominal capacity is difficult to directly translate into effective capacity in Europe and America. Companies that have preemptively established overseas production bases are more likely to benefit from the current tight global supply-demand situation.

For ongoing industry tracking, three key indicators are essential: first, spot prices of loose fiber; second, expansion moves of leading manufacturers; third, the progress of overseas long-term contracts, which verify the industry’s demand sustainability.

Of course, potential risks include underwhelming global AI investment, faster-than-expected fiber rod capacity expansion, intensified industry competition, and technological shifts like CPO routes.

04 Conclusion

Looking back at the development history of the optical communication industry, every technological revolution has spawned a group of companies with long-term competitiveness. Fiber optic cables not only support the core neural network of global AI infrastructure but also open new growth spaces in military drones, next-generation hollow-core fibers, and more.