Steel giant BlueScope's profits soar, uncovering the "resource layer" of AI computing infrastructure: industrial metals like steel, copper, and aluminum become the "second battlefield" of the AI super cycle.

Amid a surge in demand for AI computing power, the construction of AI data centers in North America is thriving, resulting in strong metal demand that is significantly boosting profits for steel manufacturer BlueScope. This latest development highlights the unprecedented increase in the demand for key industrial metals and minerals that are continuously consumed in the construction of artificial intelligence data centers. As the power density of AI GPU/ASIC compute clusters continues to rise, the expansion of AI computing will simultaneously drive the demand for physical infrastructure such as building structures, power distribution, and thermal management systems for data center projects, thereby increasing the demand for metals like copper, aluminum, and steel, rather than merely amplifying chip demand. BlueScope Steel Ltd. has seen a dramatic rise in profits, one important driver of which is the record sales growth in its North American industrial metals business spurred by the data center construction boom. AI capital expenditures are rapidly transitioning from the "silicon-based compute layer," such as GPUs and HBMs, to the physical infrastructure layer composed of steel, copper, aluminum, and power and engineering equipment. The Melbourne-based steel manufacturer reported a more than doubling of its underlying net profit to AU$851.2 million (US$604 million) for the fiscal year ending June 30. Revenue in North America increased by 9%, offsetting a 4% decline in revenue from the Asian region. From an engineering perspective, a large AI data center is primarily a highly industrialized "compute factory": steel is used in the main structure, prefabricated steel frameworks, support for equipment, racks, and peripheral facilities; copper, a crucial industrial metal, is utilized in high and low voltage power distribution, cables, busbars, grounding, transformer connections, as well as liquid cooling plates and heat exchangers. Aluminum, regarded for a long time as an alternative metal to copper, especially in cost- and weight-sensitive applications where conductivity requirements are relatively relaxed, has seen this trend accelerate in recent years due to resource security and the push for new energy industries. However, in high-reliability, high-power scenarios, copper remains irreplaceable. Industrial metals such as steel, copper, aluminum, nickel, and tin are becoming the core investment trend beyond the AI computing infrastructure chain, which includes AI GPUs/ASICs, data center CPUs, HBM/NAND/HDD storage, 2.5D/3D advanced packaging, liquid cooling thermal management systems, optical interconnect supply chains, and power supply chains for data centers. Building AI applications and data centers requires not only investment in models, chips, and high-performance AI cloud computing servers but also substantial expenditures on the underlying energy, metals, chemical products, and resource security premiums that support the expansion of AI computing infrastructure capacity. The AI data center boom has begun to "gobble up steel"! BlueScope's profits have doubled, making North America a new growth engine. BlueScope CEO Tania Archibald stated in a media interview that the North American market remains the companys "primary performance growth engine." She remarked, "We are benefiting from the demand for data centers, and we can see the strong performance of this demand." Archibald, BlueScope Steel's managing director and CEO, mentioned that the company is currently highly focused on organic growth and plans to continue enhancing overall returns for shareholders over the long term. On the other hand, Archibald expressed that as the company strives to control energy costs, BlueScope is also "highly vigilant" regarding the potential impacts that data centers may have on broader energy supply issues. In February of this year, BlueScope rejected a takeover bid from Steel Dynamics Inc. and SGH Ltd., which valued BlueScopes equity at approximately AU$15 billion. At that time, BlueScope stated that the offer undervalued the company. Since then, its share price has risen by about one-fifth. On Monday, BlueScope's stock fluctuated between gains and losses, dropping 0.8% as of 2:25 p.m. Sydney time. Archibald stated, "We have had no contact with the consortium for quite some time. We clearly rejected their last proposal because the offer did not adequately represent the fair value that BlueScope shareholders deserve." The Australian companys North Star steel mill in Ohio has also benefited from the 50% tariffs imposed by the Trump administration on imported steel, aimed at addressing the issue of excess capacity from the Asian region. On Monday, Archibald stated during an investors' conference call that BlueScope expects "the North American market to remain strong, the demand environment in Australia to be robust, and preliminary signs of recovery in New Zealand." She added, "In certain Asian markets, excess capacity continues to suppress regional steel price spreads." In a PowerPoint presentation directed at shareholders for FY2026 (the entire fiscal year of 2026), BlueScope explicitly listed "Data center and AI infrastructure build-out lifting demand" as an important support for North American demand; its North America business unit achieved an underlying EBIT profit of AU$1.034 billion for the year, a 101% increase year-on-year, with North Star Steel mill EBIT reaching AU$805 million, maintaining 100% capacity utilization. However, the profit growth was also driven by stronger steel price spreads, capacity utilization, and macroeconomic growth in North America, and thus not all of the profit doubling can be attributed to AI. The significant improvement in BlueScope's profits truly reveals that the construction process of AI data centers has transitioned from "massive purchases of GPU/ASIC cabinets" to a "massive acceleration of infrastructure at the level of Jianshe Industry Group," underscoring the critical industrial metals that are rapidly forming an "AI computing infrastructure resource layer." In other words, the endpoint of AI is not just electricity, but rather "electricity + metals + engineering capacity"when computing resources transition from silicon chip systems to GW-level infrastructure, industrial metals begin to acquire some structural AI attributes beyond traditional cyclical characteristics. Among them, steel, copper, and aluminum are the most directly related to the physical capital expenditures of AI data centers, while lithium, nickel, and tin belong to a different layer of relevance. Copper is the most typical "AI electrification metal," covering high-speed copper interconnects inside data centers, distribution systems, external grid expansion, and liquid cooling; aluminum is widely used in power distribution, busbars, cables, and some cooling and structural systems; steel corresponds to CFC/data center buildings, substations, power generation facilities, racks, and major non-residential construction. The logic for tin primarily stems from solder, PCBs, and electronics manufacturing, where its absolute scale of use is far smaller than steel, copper, and aluminum; lithium and nickel mainly benefit from UPS, energy storage systems (BESS), and the power supply chains added for data centers, making their "AI purity" notably lower than copper, and they are also more influenced by EV demand, battery chemical systems, and mining supply cycles. Goldman Sachs, a Wall Street financial giant, forecasts that global data center electricity demand will grow by 220% by 2030 compared to 2023, equivalent to adding one of the worlds top ten electricity-consuming countries. The International Energy Agency (IEA) anticipates that global data center electricity consumption will rise from approximately 415 terawatt-hours in 2024 to about 945 terawatt-hours by 2030, an average annual increase of about 15%, accounting for nearly 3% of global electricity consumption; among this, the electricity consumption of AI-accelerated servers is projected to grow at about 30% annually, with U.S. data centers increasing their electricity consumption by about 240 terawatt-hours compared to 2024, representing an approximately 130% increase, contributing to nearly half of the new power demand in the U.S. by 2030. The IEA also expects that global data center electricity usage will roughly double by 2030, with the growth of electricity consumption in dedicated AI training/inference data centers occurring even more rapidly. Meanwhile, grid bottlenecks may cause about 20% of planned data center capacities before 2030 to face interconnection delays, all suggesting that the most scarce resources for the next stage of AI-themed investments are shifting from "chips" to electricity, conductors, structural materials, and energy infrastructure.