AI’s Power Demand Is Rewiring Global Markets

Array
(
    [0] => Array
        (
            [0] => ((BHP))
            [1] => ((SFR))
            [2] => ((AIS))
            [3] => ((29M))
            [4] => ((A1M))
            [5] => ((HGO))
            [6] => ((MAC))
            [7] => ((RIO))
            [8] => ((PDN))
            [9] => ((BOE))
            [10] => ((DYL))
            [11] => ((BMN))
            [12] => ((NXG))
            [13] => ((SPN))
            [14] => ((NXT))
            [15] => ((IFT))
            [16] => ((MAQ))
            [17] => ((DGT))
            [18] => ((GMG))
            [19] => ((S32))
            [20] => ((WDS))
            [21] => ((STO))
            [22] => ((PLS))
            [23] => ((IGO))
            [24] => ((LYC))
        )

    [1] => Array
        (
            [0] => BHP
            [1] => SFR
            [2] => AIS
            [3] => 29M
            [4] => A1M
            [5] => HGO
            [6] => MAC
            [7] => RIO
            [8] => PDN
            [9] => BOE
            [10] => DYL
            [11] => BMN
            [12] => NXG
            [13] => SPN
            [14] => NXT
            [15] => IFT
            [16] => MAQ
            [17] => DGT
            [18] => GMG
            [19] => S32
            [20] => WDS
            [21] => STO
            [22] => PLS
            [23] => IGO
            [24] => LYC
        )

)

This story features BHP GROUP LIMITED, and other companies.
For more info SHARE ANALYSIS: BHP
The company is included in ASX20, ASX50, ASX100, ASX200, ASX300 and ALL-ORDS

Analysts are increasingly toying with the idea that the AI infrastructure build out might just herald the next super cycle for commodities.

• AI, truly, is changing the world
• Impacts are no longer confined to technology, but increasingly include energy, metals, and utilities
• The world might be witnessing the early stages of another commodities super cycle
• The ASX offers multiple ways for investors to get on board

By Omega Ukama

Only a few forces in history have reshaped global commodity markets with the intensity of a major technological paradigm shift.

The steam engine demanded coal. The automobile demanded oil. Now, artificial intelligence is demanding an unprecedented convergence of copper, the infrastructure of the electrical grid and even uranium, aluminium, gas, lithium, and rare earths.

What began as a narrative dominated by Silicon Valley algorithms and large language models has transformed into something far more tangible.

The AI boom has become a physical endeavour requiring gigawatt-scale campuses, miles of copper cabling, and a fundamental restructuring of how the world generates and consumes electricity.

According to the International Energy Agency (IEA), global data centre electricity consumption stood at approximately 460 terawatt-hours (TWh) in 2024 and is projected to more than double to over 1,000 TWh by 2030, potentially reaching 1,300 TWh by 2035.

Goldman Sachs estimates the increase at as much as 175% above 2023 levels by the end of the decade, effectively adding the electricity consumption of a top-ten nation to global demand.

In the United States alone, data centres could account for up to 9% of total electricity consumption by 2030, up from 4% in 2023.

The bottleneck for AI progress has moved down the stack. The primary constraints are no longer advanced chips, but the physical capacity of the electrical grid and the raw materials needed to build massive “AI factories”.

The central question now is whether this AI-driven demand surge will ignite a prolonged period of structurally higher prices driven by supply constraints and demand, comparable to the industrialisation of the early twentieth century, or as some would put it, a new commodities supercycle.

Industry analysts at Jones Lang LaSalle (JLL), taking into account the world’s demand for AI infrastructure and data centres, have described the current moment as “the beginning of one of the largest infrastructure investment supercycles in the modern era”.

The evidence, drawn from the latest market data and analyst forecasts, suggests the foundations are already being laid.

The Copper Conundrum

Of all the commodities caught in AI’s gravitational pull, copper stands the most exposed.

Data centres are copper-intensive. Estimates vary by facility type. AI-focused centres require between 27 and 50 tonnes of copper per megawatt of capacity, significantly more than traditional enterprise data centres, encompassing power delivery wiring, switchgear, cooling systems, and internal server connections that keep massive GPU clusters operational.

BloombergNEF projects copper demand specifically from data centres will reach 572,000 tonnes annually by 2028, roughly equivalent to adding a new top-tier mining nation to global supply in under four years.

Looking further ahead, Wood Mackenzie estimates global copper demand will surge 24% by 2035, yet the industry has allocated only US$76bn in mining investments over the past six years, far short of the US$210bn required to meet projected needs.

Mine development cycles of approximately 17 to 18 years mean supply cannot simply respond to price signals. Morgan Stanley analysts project a cumulative global supply deficit of -10m tonnes by 2040.

This structural imbalance is already reflected in prices. Benchmark copper on the London Metal Exchange has surged more than 100% since the end of 2019, with prices in early 2026 frequently trading above US$13,000 per tonne, having reached a record US$14,527.50/t in January this year.

The implications for Australian-listed copper producers are profound.

BHP Group ((BHP)), the world’s largest producer (on reported volumes), stands as a primary beneficiary. In its 2026 half-year result, copper earnings surpassed iron ore for the first time in the company’s history.

With the Escondida mine in Chile, the world’s largest copper operation, and an expanding portfolio across Australia and Peru, BHP is uniquely positioned to capture what analysts at Wedbush Securities have termed the “AI physical supercycle”.

Sandfire Resources ((SFR)) has evolved from a single-asset Australian copper miner into a genuinely diversified global producer. That transformation positions it well for the structural demand shift underway.

Sandfire’s two core operating hubs, the Matsa polymetallic complex in Spain and the Motheo copper mine in Botswana’s Kalahari Copper Belt, give it geographic diversification across two of the world’s most prospective copper regions.

Aeris Resources ((AIS)) represents a more targeted, higher-conviction copper play concentrated in Australia’s Cobar Basin, a region that has quietly become one of the most active copper development corridors on the continent.

Aeris’ cornerstone Tritton underground copper operation in central New South Wales forms the foundation of a production hub that is now being substantially expanded.

Note: the ASX currently offers potential exposure to seven copper producing mining companies, also including 29Metals ((29M)), AIC Mines ((A1M)), Hillgrove Resources ((HGO)), and MAC Copper ((MAC)). Plus Rio Tinto ((RIO)), of course.

Uranium’s Second Act

Perhaps the most dramatic pivot in the AI-energy nexus is the resurgence of nuclear power.

For decades, uranium was a controversial and politically fraught commodity. Today, it is being presented as the strategic bedrock of reliable, carbon-free baseload power for hyperscale data centres.

The logic is compelling. A single generative AI query can consume up to ten times more electricity than a conventional web search. Multiplied across billions of daily interactions, the cumulative load demands power that is not just abundant but uninterrupted. Solar and wind cannot alone provide that consistency.

Meta has signed agreements securing up to 6.6 gigawatts of nuclear capacity to support its AI expansion. In February, Meta and Oklo separately announced a 1.2 gigawatt nuclear campus in Pike County, Ohio.

Constellation Energy is leading the restart of the former Three Mile Island Unit 1 to supply 835 megawatts exclusively to Microsoft, while NextEra Energy has partnered with Alphabet to restart the Duane Arnold plant in Iowa.

These “behind-the-meter” arrangements are becoming the standard for the industry’s elite.

The uranium market is responding accordingly. Spot prices surged 25% in January, briefly surpassing US$100 per pound for the first time in two years.

A global survey of more than 600 investors conducted for Uranium.io found 63% believe AI-related power consumption will become a critical factor in nuclear planning over the next decade, with more than 85% expecting further price rises in 2026, projecting a range of US$100 to US$135 per pound.

Analysts project the uranium sector could grow to more than US$60bn by 2030.

For Australian investors, Paladin Energy ((PDN)) occupies the most direct position in this theme. Its Langer Heinrich Mine in Namibia recommenced operations in March 2024, with production targeting up to 4.4m pounds of U3O8 for FY26.

With 85% of reserves either uncontracted or on market-related pricing, Paladin carries significant leverage to further price appreciation. Its ongoing acquisition of Fission Uranium enhances longer-term production capacity through the Patterson Lake South project in Canada.

Boss Energy ((BOE)), operating from its Honeymoon mine in South Australia and holding a 30% stake in the Alta Mesa project in the United States, offers complementary geographic diversification, but also operational uncertainty.

Boss Energy’s recent problems have centred on Honeymoon, where operating data and drilling showed the project’s geology and recovery characteristics were less favourable than assumed in the 2021 enhanced feasibility study.

The company said reduced continuity of mineralisation and weaker leachability raised doubts about sustaining capital intensity and its ability to achieve nameplate capacity. 

Updated FY26 guidance and higher cost expectations triggered a sharp market reaction in July 2025.

Boss later said its Honeymoon review found a material deviation from the 2021 study and formally withdrew that feasibility work as a guide to future performance. Importantly, Boss has maintained Honeymoon remains on track for FY26; the larger concern is the mine’s FY27-plus production and cost profile.

Deep Yellow ((DYL)) is the only ASX-listed company with two advanced projects in Tier-1 mining jurisdictions. Its flagship Tumas project in Namibia carries a completed Definitive Feasibility Study, a 20-year mining licence, and a potential production capacity of 3.6m pounds of uranium per annum with a mine life exceeding 30 years.

A second asset, the Mulga Rock project in Western Australia, the only fully permitted uranium project in the state, adds a further 3.5m pounds per annum of potential capacity and exposure to critical minerals, plus rare earth elements as a byproduct.

Bannerman Energy ((BMN)) brings sheer scale. The Etango Uranium Project in Namibia’s Erongo Region holds a uranium mineral resource endowment of 207m pounds of contained U3O8, making it one of the largest undeveloped uranium deposits in the world, located in the same established mining corridor as Paladin’s Langer Heinrich and Rio Tinto’s Roessing operations.

The project’s development pathway was dramatically de-risked in February 2026 when Bannerman announced a landmark financing agreement with CNOL, a subsidiary of Chinese state nuclear giant CNNC.

Equally listed on the ASX is NexGen Energy (Canada) ((NXG)), a Canadian uranium developer whose investment case centres on the 100%-owned Rook I project in Saskatchewan, which recently received final federal approval and is moving towards construction.

The company describes Rook I as the largest development-stage uranium project in Canada.

The Grid as Ultimate Constraint

Beneath the narratives of copper mines and uranium deposits lays a more fundamental infrastructure story, the electrical grid itself.

Goldman Sachs estimates approximately US$720bn will need to be spent on grid upgrades through 2030, with analysts warning transmission projects can take many years to permit and build, creating a structural bottleneck if regions fail to act proactively.

Siemens Energy’s order backlog has swelled to a record EUR146bn, driven by demand for gas turbines and grid equipment, while Hitachi Energy has committed US$1bn to scale up US transformer and high-voltage equipment production, explicitly citing data centres as the core demand driver.

The fiscal consequences are already flowing through to households. In the PJM electricity market, stretching from Illinois to North Carolina, data centres accounted for an estimated US$9.3bn price increase in the 2025–26 capacity market, with average residential bills expected to rise by US$18 a month in western Maryland and US$16 a month in Ohio.

A Carnegie Mellon University study estimates data centres and cryptocurrency mining combined could push average US electricity bills up by 8% by 2030, and by more than 25% in the highest-demand markets.

In Northern Virginia’s “Data Center Alley”, data centres already consume 25% of regional electricity, with projections suggesting this could reach 50% of Virginia’s total.

Some analysts estimate the United States may require as many as three million miles of new transmission lines by the end of the decade to support AI-related load growth.

The mismatch in development timelines is acute. A data centre can be constructed in 18 to 24 months, while a new transmission line or power plant typically takes many years.

In response, developers are increasingly deploying on-site behind-the-meter generation and locating facilities in remote regions such as Southern Texas, rural Louisiana, and Wisconsin, where low-cost power matters more than proximity to users.

The federal government has initiated “fast-track” rulemaking to prioritise interconnection of large-scale AI loads, reflecting a broader national security priority.

For ASX-listed companies, the grid buildout creates opportunities beyond traditional mining. Sparc Technologies ((SPN)) is developing AI-driven corrosion assessment software that digitalises and accelerates critical materials testing for infrastructure assets, a capability of increasing value as utilities race to expand and maintain grid infrastructure.

Australia itself is expected to attract approximately $26bn in data centre investment by 2030 benefiting domestic operators including NextDC ((NXT)), which is recording strong demand for high-density AI racks, and Infratil ((IFT)), which holds a stake in CDC Data Centres.

Among smaller sized market participants, we also find Macquarie Technology ((MAQ)) and DigiCo Infrastructure REIT ((DGT)), while ASX Top20 member Goodman Group ((GMG)) is a true global market leader when it comes to building data centres the world around.

Away from the ASX, Sydney-based Iren Energy (NASDAQ:IREN), which pivoted from bitcoin mining to AI compute infrastructure, has seen its fortunes rise sharply as demand for dedicated processing power surges.

Extra note: FNArena publishes a weekly update dedicated to the ins and outs of the global uranium sector, including ASX-listed producers and explorers.

This week’s update: https://fnarena.com/index.php/2026/04/14/uranium-week-buyers-turn-to-term-market/

Beyond Copper, Uranium and the Grid

The copper and uranium narratives capture the most immediate material consequences of AI electrification, but they represent only the first layer of a far broader commodity demand cascade.

Analysts increasingly argue AI is not simply creating a mining supercycle, it is creating a multi-layered electrification supercycle, where demand ripples outward across energy, metals, and infrastructure simultaneously.

Aluminium is perhaps the most overlooked beneficiary. Cheaper and lighter than copper, it is the dominant material for high-voltage transmission lines and is playing a growing role inside data centres themselves, where it is increasingly substituted for copper in power distribution applications as operators seek cost efficiencies at scale.

According to commodity research group CRU, the data centre sector represented approximately 7% of total North American cable demand in 2025.

It is expected to grow at a compound annual growth rate of approximately 17% between 2026 and 2030, with aluminium projected to account for an increasing share of the data centre cable product mix as operators seek more cost-efficient power distribution solutions.

The commercial opportunity has not been lost on major producers. In March 2026, Rio Tinto and cable manufacturer Prysmian announced a five-year supply agreement and joint trial to develop low-carbon aluminium cables specifically targeting the data centre market.

This strategic partnership validates both the scale of the opportunity and Rio Tinto’s positioning within it. The LME aluminium price reached its highest quarterly average since mid-2022 in the fourth quarter of 2025, supported by expectations of a tighter global balance in the first half of 2026.

South32 ((S32)), with its significant alumina and aluminium operations, offers complementary exposure to the same structural demand shift.

Natural gas occupies a strategically critical, if less glamorous, position in the AI energy mix. The challenge confronting data centre operators is fundamentally one of timing as nuclear and large-scale renewables cannot be built fast enough to meet demand that is accelerating now.

Gas-fired generation is filling that gap. Nearly 75% of the power equipment planned for on-site use at data centres is natural gas, according to recent industry analysis.

The IEA projects approximately 40% of additional energy consumption by data centres through 2030 will still be supplied by gas- and coal-based sources, owing to the inherent variability of renewables and their mismatch with the flat, 24/7 load profile of AI workloads.

At the LNG 2026 conference in Doha, the leaders of Shell, ExxonMobil, TotalEnergies, ConocoPhillips, and QatarEnergy reached a collective conclusion that AI, data centres, electrification and population growth are all pulling the energy system to a new scale, with demand rising faster than grids, infrastructure, and policy frameworks can adapt.

For ASX investors, Woodside Energy ((WDS)) is the most direct beneficiary. Its CEO has explicitly linked the company’s US$17.5bn Louisiana LNG project to future energy demand from the US technology sector, framing LNG exports as a direct feed into the gas turbines powering American AI infrastructure.

Santos ((STO)) provides complementary exposure across a diversified gas portfolio spanning Australia, Papua New Guinea, and the United States.

Lithium’s role in the AI energy story is increasingly material. Battery Energy Storage Systems, large-format batteries used to smooth intermittent renewable supply and provide reliable backup power for data centres, saw demand surge 51% in the most recent year, compared with 26% for electric vehicles, with BESS now accounting for approximately 23% of total global lithium demand.

As AI hyperscalers simultaneously commit to net-zero targets and wrestle with the intermittency of wind and solar, battery storage has become the critical buffer between clean energy generation and the uninterrupted 24/7 power that AI workloads demand.

Annual BESS additions could approach 800 gigawatt-hours by the 2030s, underpinned by a -43% decline in lithium-ion battery costs since 2022, with government incentives and growing needs for grid optimisation and backup power for AI-driven data centres reinforcing this growth.

Spodumene prices responded dramatically: between December 2025 and February 2026, spodumene concentrate prices tripled from their lows, with January 2026 spot prices reaching approximately A$2,500 per tonne compared with A$600 per tonne in July 2025.

PLS Group ((PLS)), formerly Pilbara Minerals, operator of the world-class Pilgangoora lithium-tantalum project in Western Australia, is the most leveraged producer on the ASX to a continued BESS-driven demand recovery.

Its P2000 feasibility study, which would double concentrate production at Pilgangoora, is due in the December 2026 quarter, representing a significant potential production catalyst.

IGO Limited ((IGO)), as co-owner of the Greenbushes Mine in Western Australia, widely regarded as the highest-quality hard-rock lithium operation in the world, offers a lower-risk entry into the same structural theme.

The deepest layer of the AI commodity demand stack, and the one that most investors have yet to fully price in, is rare earths.

The high-efficiency motors that cool data centre GPU clusters, the wind turbines generating renewable power for AI campuses, and the robotics increasingly integrated into AI-enabled supply chains all depend on permanent magnets built from neodymium, praseodymium, and dysprosium.

AI, in other words, is an indirect but powerful driver of rare earth demand through the full electrification and automation ecosystem it is accelerating.

Lynas Rare Earths ((LYC)), the largest producer of separated rare earths outside China, is expected to enter a stronger growth phase in FY26, with total rare earth oxide production forecast to rise 53% year on year to 16,100 tonnes, and average realised prices projected to climb 47% to $72.50 per kilogram, supported by tighter global supply and improving demand.

Paying subscribers can explore these themes further through FNArena Windows: https://fnarena.com/index.php/financial-news/fnarena-windows/?sector=43

A Supercycle Takes Shape

The evidence increasingly supports the supercycle thesis. Moody’s projects at least US$3trn in global data centre investments over the next five years, with the sector expanding at a 14% compound annual growth rate and roughly 100 gigawatts of new capacity expected online between 2026 and 2030.

AI-related investment categories accounted for 39% of total US GDP growth across the first three quarters of 2025, while AI-related trade topped US$272bn in the first half of 2025, up 65% year-on-year.

Yet commodity markets tied to the physical layer of that buildout have not yet fully priced in the structural demand shift ahead.

Crucially, unlike previous commodity cycles driven primarily by Chinese industrialisation, this cycle is multi-polar.

The US Inflation Reduction Act and CHIPS Act, Europe’s Green Deal, and Asia’s digital infrastructure programmes are converging to create synchronised demand growth across major economies, creating a sustained and geographically distributed demand base that previous cycles lacked.

Risks and Reality Checks

Significant risks remain. Moody’s analysts have flagged growing concerns about an AI investment bubble, with capital spending on computing infrastructure outpacing revenue generation from AI applications.

If return on investment does not manifest more clearly in corporate earnings by late 2026, there is a risk infrastructure spending could create overcapacity.

Environmental and social constraints are also mounting. In data centre hubs like Abilene, Texas, and New Albany, Ohio, community resistance is growing amid concerns about electricity rates and water consumption, a phenomenon now termed the “Data Center Rebellion”.

Advances in chip efficiency could also mitigate some projected demand growth, though analysts note the historical tendency for efficiency gains to be absorbed by greater utilisation rather than reduced overall consumption.

The Bottom Line

For investors navigating this landscape, the AI infrastructure buildout has fundamentally altered valuation models across sectors. The “AI trade” is no longer confined to technology stocks, it is transforming industrials, materials, and utilities.

The key takeaway is that the market is rewarding vertical integration and strategic positioning.

As the data centre industry evolves toward gigawatt-scale campuses and nuclear-powered computing clusters, the success of the AI era will be measured not just in floating-point operations per second, but in megawatts and metric tons.

The comparison to early twentieth century electrification, a buildout that lasted decades and remade commodity markets, may not be hyperbole after all.

Read more about AI and its multiple aspects and developments through FNArena’s dedicated section: https://fnarena.com/index.php/tag/gen-ai/

Find out why FNArena subscribers like the service so much: “Your Feedback (Thank You)” – Warning this story contains unashamedly positive feedback on the service provided.

FNArena is proud about its track record and past achievements: Ten Years On

Click to view our Glossary of Financial Terms