Where does artificial intelligence (AI) live in the physical world? Not in the cloud, but in industrial vaults of silicon and steel. Its life depends on the same electrical grids that power our cities—and now it’s starting to rival them in consumption. “Generative AI changes the contract between humans and technology: before AI, interaction meant adapting to machines; after AI, machines adapt to us,” says Snowflake’s CEO, Sridhar Ramaswamy. This helps explain the dramatic surge in adoption: to command remarkably sophisticated systems, turns out all one needs is basic literacy. ChatGPT now counts more than 700 million^[1]weekly users a mere three years after launch and it is hard to find a company in 2025 not trying to incorporate AI into its operations. 

The global investment required to meet AI-data-center demand alone by 2030 is estimated at US $5.2 trillion, under McKinsey’s mid-scenario.^[2]

As Medidata’s CEO, Anthony Costello states: “The global momentum behind AI makes inaction riskier than adoption.” This powerful momentum comes with its challenges. “AI is rewriting the speed limit of software—code is being produced faster than security or governance models can keep up,” says Brian Roche, CEO of Veracode. Industries, institutions and individuals are juggling to keep up. The technology advances, we implement it, and only thereafter try to make sense of it. The challenge is clear: AI’s body will have to evolve as fast as everyone’s reliance on it.

Rooms of servers pull so much power they need thousands of fans or millions of liters of water to stay cool. These are the world’s data centers, the body of artificial intelligence. “Data centers underpin almost every digital interaction,” says Juan Font, President and CEO of CoreSite and SVP of American Tower. “Each time someone scrolls a social feed, checks their bank balance, or joins a video meeting, that request travels to a server housed in a purpose-built facility engineered for 100 percent uptime.”

David Robey and Tag Greason, co-CEOs of QTS Data Centers, say the AI boom has upended every metric. The key to meeting that demand starts not with land but with energy: “It’s no longer enough to simply acquire land and build a facility; the process must start with securing power.” Mark Moffat, CEO of IFS echoes that point: “The clearest limiting factor is energy availability. Infrastructure is being built out, but there isn’t enough energy on the grid.”

This power-first mindset mirrors how AI is reshaping industries. The U.S. Department of Energy estimates data centers already account for about 4.4 percent of U.S. electricity demand^[3], and Goldman Sachs projects global data-center power usage could rise 165 percent by 2030^[4]. As more daily tasks and routines start relying on ChatGPT, those figures do not sound far-fetched — especially if, as Moffat claims, “we’re at the very beginning of this AI cycle.” If Goldman’s projection is even close, our energy and economic models will need to change, and quickly.

Will Abbey, executive VP & CCO at Arm, said that “this exponential growth demands a smarter approach than simply adding more servers or consuming more power.” Arm’s architecture, designed for performance per watt, now powers more than 350 billion devices worldwide, with around 33 billion added each quarter.^[5] Originally built to maximise efficiency in mobile chips, the same principles now scale across many of the world’s cloud systems. This shift toward intelligent, efficiency-first design extends into the manufacturing process itself. As Dave Evans, CEO and co-founder of Fictiv, notes, “By controlling both the flow of information and the flow of materials through AI, we’re helping companies modernize an industry that has long relied on archaic methods.”

Unsurprisingly, all this upheaval is changing what data centers are. Switch argues that “what used to be a neutral ‘compute warehouse’ is now becoming an AI factory, purpose-built to handle the intensive demands of training and deploying large-scale models.” Traditional IT was built for predictable loads. AI brings spikes in computation and heat. Facilities ahead will need ultra-dense power delivery and much more efficient cooling. Data centers have, in effect, become the public utilities of the digital age: mostly invisible, essential and under heavy strain.

The great resource race

To build and fuel the ‘body’ of AI, physical resources are elemental. The language of our present has tended to conceal the most basic of all facts: hard matter makes it all possible. And those who know best about ‘matter’ are the gritty heirs of an ancient industry — they dig and dig and dig, determined to remain unyielding, certain that their toil makes the earth go round. These are the miners.

Of late, one mining asset has become particularly popular. Rare earths, which are in fact not that rare but can be unusually hard to mine, have made many headlines due to their presence in much of modern technology, from AI to defense gadgets. Robert Martin, Executive Chairman of Lindian Resources, frames the race for rare earths as nothing less than existential. Lindian’s Kangankunde project in Malawi is set to feed into Western supply chains. “It is a very large, multi-generational asset,” Martin says, “high grade, low impurity and extremely low in uranium and thorium.” The permanent magnets and precision components inside AI hardware, electric motors and even the cooling systems of GPUs are all dependent on rare earths.^[6] “Governments are even discussing floor prices to combat global price manipulation^[7],” Martin notes, hinting that the West races to secure its own sources of critical minerals. Mining projects like those of Lindian may prove as essential to digital sovereignty as the data centers they ultimately power.

If rare earths form the bodies of modern machines, then lithium is amongst the fuels that keep them going. Lithium is the essential element in the batteries that store and deliver the energy powering everything from electric cars to servers. Despite a year of falling prices, Francis MacDonald, CEO and Director of Li-FT Power, remains confident in the metal’s long game. His company’s Yellowknife project in Canada has already emerged as one of the largest undeveloped hard-rock lithium deposits in North America^[8]. MacDonald sees the current price downturn as temporary. “The long-term demand for lithium remains very strong,” he says, referring to growth drivers such as data centers and electric vehicles. As Western producers move to secure domestic supply, projects like Yellowknife remind us that even the most ethereal realms of AI and cloud computing still run on stuff dug from the ground.

AI is itself impacting the art of mining. Kevin MacLean, the CEO of MacLean Engineering, tells us of the company’s mission to increase automation. “The next challenge will be the zero-entry mine,” he says, describing a future where autonomous, battery-electric equipment operates in the extreme conditions of deep-earth mining. It’s a vision that marries the old and the new — machines learning to navigate environments once too dangerous for humans.

One metal with a subtler link to the AI economy is gold. Yes, gold does have some important industrial input in AI-related electronics, due to its conductivity. But the more interesting relationship relates to stock markets. It is unorthodox that gold prices and stock markets, now dominated by tech, should rise together. Their recent concerted surge seems to reflect the appetite for both digital expansion and the physical and financial stability offered by gold. We spoke with Jason Attew, President and CEO of OR Royalties Inc., to understand more. “Our job is risk management — financing the mines that supply the metals the world needs over the next decade,” says Attew. A royalty and streaming company, OR Royalties plays an important role in the financing of gold miners. Attew is of the belief that gold is and will remain critical to the world economy, and he is not alone. “Gold has historically been a safe haven, protecting against sudden shifts in geopolitics or financial markets. The unpredictable nature of markets today, driven by factors such as tariffs or political shifts, only highlights the importance of gold,” shares with us the CEO of Harmony Gold, Beyers Nel. Even in an increasingly digital economy, gold continues to prove the value of tangibility.

As the CEO of Endeavour Mining, Ian Cockerill, puts it: ‘Very few companies in mining can offer the combination of growth and yield—that’s our key differentiator,’ emphasizing that Endeavour maintains strong dividends even through major capital projects. In an environment where investors seek both stability and upside, this blend of resilience and return reinforces gold’s enduring role.

The other frontier

On the periphery of the AI revolution, another kind of computing is resolutely taking shape, one that could eventually redefine what “powerful” means. Quantum computing, long a category only relevant for theoretical physicists, is finally going commercial.

“Quantum computing works in a completely different way from ordinary computers,” says Niccolò de Masi, CEO of IonQ, one of the leading companies in the field. “Every time we add a single ‘qubit,’ a quantum bit, the computer’s potential power roughly doubles. That means a 64-qubit machine could tackle problems about 240 million times more complex than what our 36-qubit system can handle.” That kind of exponential scaling could change everything from materials science to drug discovery.

But today’s machines are still small and unstable. Most run with just a few dozen to a few hundred qubits, and even those can be prone to error. Reaching real-world utility, what researchers call “quantum advantage,” will require systems with thousands or even millions of reliable qubits. Companies like IonQ, IBM, and Google are racing to get there, expecting meaningful progress through the late 2020s and more dependable systems by the 2030s^[9]. Interestingly, each firm is betting on a different road to that future.

IonQ uses trapped ions, single atoms held in electromagnetic fields and controlled with lasers. Quantum computer pioneer Pasqal, led by Dr. Loïc Henriet, uses neutral atoms instead, steered by beams of light. “We’ve combined the power of GPUs with our quantum processing units to solve problems classical computers can’t handle efficiently,” Henriet explains. Trapped ions are exceptionally precise, while neutral atoms are easier to scale, and Henriet believes the first real signs of quantum advantage could appear as early as 2026.

At Xanadu, CEO Christian Weedbrook is pursuing a photonic approach. Rather than atoms or circuits, Xanadu’s qubits are made of photons, individual particles of light. “A photonic qubit is simply a quantum bit encoded in photons,” he says. “Because light can travel long distances through fiber without losing information, linking quantum computers together becomes much easier.” Weedbrook imagines a “quantum data center,” where thousands of optical racks communicate through fiber-optic networks, with light itself serving as both carrier and memory.

Fujitsu is taking yet another path. Indradeep Ghosh, CEO of Fujitsu Research of America, describes the company’s approach as hybrid, blending traditional high-performance computing with quantum technologies. “Bottom-up research is driven by market needs,” he says, “while top-down research comes from emerging technologies in academia.” Fujitsu hopes that pairing these worlds will deliver near-term benefits while paving the way for true quantum breakthroughs.

Matt Kinsella, CEO of Infleqtion, points to quantum’s potential in sensing and navigation. “We harness quantum phenomena like superposition and entanglement for products such as timekeeping devices, RF antennas and sensors.”As global GPS systems grow more vulnerable to interference, Kinsella sees a national-security use case: quantum clocks that provide local precision timing impossible to jam. Ultimately, if these projects succeed, the map of compute flips from brute-force megawatts to precision physics. Until they do, AI remains the defining industrial project of our era.

To explore more of our special business reports from around the world, visit investmentreports.co.

---

---

The content is paid for and supplied by advertiser. The Washington Post was not involved in the creation of this content.