Following the storage race, copper and optical fiber may be the next big AI-driven market.

Citi's metal strategist Charlie has been itching to make a trade in recent weeks: buying a LME copper digital call option with a strike price of $15,250 expiring in August.

He believes that since 2022, almost all of the demand growth for copper has come from energy transition and AI-related sources.

As of May 14, 2026, LME three-month copper approached $14,000 per ton, with COMEX copper opening at $6.63 per pound. Over the past 12 months, copper has risen by 41%. In the last 4 weeks, copper has risen by 10%. This is a price near its all-time high.

Over the past two years, the entire market has told the AI story as a chip story, focusing on Nvidia's market cap, TSMC's capacity, HBM yield, and CoWoS packaging bottlenecks. Almost all discussions about "AI infrastructure" have been centered on those square centimeters of silicon.

But from a perspective that many people are unaware of, AI demand is sinking from the silicon down to copper mines and down to glass.

Copper is the Oil of the AI Era

Supply and demand determine price, and the view that "copper demand is strong" is more directly visible in the market.

LME three-month copper closed at $13,943 per ton on May 11, 2026, marking a record high close for LME, up 2.7% in a single day. COMEX copper hit an intraday high of $6.58 per pound on May 12. Over the past 12 months, copper has risen by 41%. In the last four weeks, copper has risen by 10%.

At the beginning of 2025, copper was still around $9,000, broke through $12,000 by mid-year, and ended the year with a 43% gain, making it copper's best year since 2009. In January 2026, copper first broke above $13,000 during trading hours. Then, four months later, $14,000 is also within reach. The shape of this curve looks more like a rediscovered asset that is being priced according to new logic.

Trafigura is the world's second-largest metal trader. Its head of metal analysis, Graeme Train, provided a very concise breakdown of demand: out of an additional 10 million tons of copper consumption over the next decade, one-third will come from electric vehicles, one-third from power generation and transmission, and the remaining one-third from automation, manufacturing capex, and data center cooling systems.

Goldman Sachs, in a research report titled "AI and Defense Place the Power Grid at the Center of Energy Security," has provided a more incisive argument: Copper will become the oil of the AI era. Goldman Sachs estimates that by 2030, global electricity grid and power infrastructure construction will account for more than 60% of the incremental copper demand.

While this may sound somewhat exaggerated, upon closer reflection, it is reasonable.

Copper has an electrical conductivity of 100% IACS, second only to silver among all metals. However, silver is too expensive, and in almost all large-scale electrical conduction scenarios, copper is the only answer. The closest substitute is aluminum, but aluminum's electrical conductivity is only 61% of that of copper. This means that to transmit the same megawatt of electricity, aluminum wires need a larger cross-sectional area, are heavier, take up more space, and have greater heat loss. In spaces as tight as data center racks, this difference is almost unacceptable.

Thermal conductivity is even more crucial. Copper has a thermal conductivity of 401 W/(m·K), which is 5 times that of iron and 8 times that of stainless steel. The NVIDIA GB200 single-card power consumption is 1,200W, a standard rack holds 72 cards, and the total rack power exceeds 130kW. At this level of heat density, air cooling is no longer sufficient, and liquid cooling is required. In a liquid cooling system, almost every component dealing with "heat" is a copper cold plate, copper tube, or copper water block.

In other words, copper is not the "preferred material" for AI data centers; it is the "physically only choice."

The power consumption of AI is disruptive, and delivering electricity to data centers is very copper-intensive.

A 1GW AI data center requires approximately 27,000 tons of copper just for power distribution and wiring. Meta's Hyperion data center in Louisiana is a 5GW facility. Calculated, the copper requirement for just this project is close to 135,000 tons, not including the high-voltage transmission lines, substations, and grid upgrades needed to bring power to the data center site.

In our previous impression, copper was considered a readily available metal. However, recent data suggests this impression may need some adjustment.

Starting in March 2026, the US-Iran conflict disrupted sulfur and sulfuric acid exports from the Middle East. Sulfuric acid is a key input for heap leaching refined copper, and Chilean refineries were forced to reduce production. This event was the catalyst for the price surge in 2026.

A more structural and macro issue is that in the past decade, no supergiant copper mines have been discovered globally. John Meyer, an analyst at UK seller SP Angel, believes that the breakeven price for the development of the next generation of new copper mines is $13,000 per ton, already exceeding the current price of copper. The team led by Wang Jiechao of CITIC Securities estimates a global refined copper shortage of over 100,000 tons in 2026, while Citi's forecast is more aggressive at 308,000 tons.

The 2026 Fiber Optic "Great Famine"

So far, the story of copper has been a clear bullish narrative. However, if you zoom into the inner workings of AI data centers, you'll find a very subtle shift: part of the copper's demand is being replaced.

"The next generation of artificial intelligence infrastructure will require a large amount of optical connections because the computing demand is rapidly growing to the point where copper wires can no longer meet the demand," said Huang Renxun in an interview earlier this month.

As Huang Renxun mentioned, the data transmission requirements of AI clusters are pushing past the physical limitations of copper cables.

There are two fundamental constraints to transmitting high-speed signals over copper cables: signal loss sharply increases with frequency, and at high frequencies, the volume and weight of copper cables become intolerable. The interconnect bandwidth between GPU clusters is advancing from 200G, 400G all the way to 800G, 1.6T. The distance that copper cables can support has shrunk from meters to tens of centimeters. AI clusters consist of thousands of cards, spanning across racks, sometimes even data centers, which is physically impossible with copper.

But fiber optics can.

That's why this wave of fiber optics is more intense, purer, and more irreversible than copper. How exaggerated is the surge in fiber optics?

According to CRU's data: the price of China's G.652D bare fiber surged over 80% between November 2025 and January 2026. The average price in January was 31.5 RMB/core-km, with some actual transactions occurring in the range of 40 to 50 RMB, resulting in a cumulative increase of 94% to 144%.

This industrial product, which has been stable in price for the past few years, more than doubled in just three months.

By February 2026, high-end fiber optics saw an even more aggressive surge. For example, G.657.A bend-insensitive fiber increased from above 30 RMB/core-km to over 50 RMB in a month. Sun Telecom directly proclaimed the slogan "2026 Fiber Optic Great Famine." Its G.652D, which was priced at $2.20 per kilometer in 2024, rose to $3 in December 2025, and a month later surged to $4.10. The overall price of Asia's fiber optics rose by 75%, hitting a 7-year high.

The demand for fiber optics in AI data centers is a magnitude of disruption.

Rahul Puri, CEO of STL's Optical Networks Business, once mentioned a figure that even caught the attention of the author of this piece: the amount of fiber optics needed for an AI data center is 36 times that of a traditional CPU rack, marking a cliff-like leap.

The operation of a GPU cluster is completely different from that of a CPU. A multi-card training cluster requires a non-blocking, high-speed interconnection among all GPUs. This network structure is called a Scale-out architecture, with bandwidth requirements that were unimaginable in the CPU era. Additionally, data centers need DCI links to stitch together computing clusters located in different geographic locations into a supercomputer. In Meta's Hyperion data center project, the fiber optic demand has reached 8 million miles.

Returning to the economic theory that prices are determined by supply and demand, if demand is in such a state, what about the supply side?

Light Reading reported that at least one leading fiber optics manufacturer had completely sold out its 2026 inventory. Data Center Dynamics reported that delivery times for large customers had extended to 20 weeks, while small customers were nearing one year.

Why can't production capacity be expanded? Because the "glass stick," which is the core material of optical fibers, has a production expansion cycle of 18 to 24 months and an extremely complex process. Even if all manufacturers today decide to ramp up production, the earliest additional capacity would not be available until the second half of 2027. During this period, demand will only continue to rise.

Clearfield's Chief Business Officer Anis Khemakhem provided a more macroscopic figure: by 2029, the United States alone will need to add 213.3 million miles of fiber optic cable, doubling the existing 159.6 million miles to 372.9 million miles. In six years, the national fiber optic inventory will double.

The biggest winner in this story is Corning.

Founded in 1851, this is a glass company that has made the glass shells for Edison's light bulbs, the glass for television cathode-ray tubes, and the Gorilla Glass for iPhones. Many people are not even aware that this company is still alive. However, it is now the core fiber optic supplier for Meta, Nvidia, OpenAI, Google, AWS, and Microsoft. Its stock price has risen by over 75% in the past year. The story of Corning may be further detailed in a new article by Lyudong's editors, but let's not delve into it further here.

The story of copper and fiber optics seems to have just begun to attract attention in the mass market, but Lyudong's editors believe that this may be the next breakthrough track after the storage race.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia