SLAM Exploration Ltd. (TSXV: SXL)
‘Exploring for critical elements and precious metals in New Brunswick, Canada.’
While the world’s attention is focused on the Middle East conflict and oil price volatility, a deeper supply crisis is quietly unfolding in the silicon-based world.
A keen observation from engineer and tech investor Ben Pouladian, recently quoted in The Wall Street Journal, has struck a chord: “Everyone’s talking about oil, but I think what the world is mainly short of is tokens.”
A token is simply the unit of measurement for computing power consumed by an AI task – every query, every generated document, every autonomous agent action draws from this “basic currency of the AI economy.” The reality is that this currency is running out faster than expected.
Over the past several months, the AI industry has undergone a paradigm shift: the evolution from “conversational AI” to “agentic AI.” The latter doesn’t just answer questions but autonomously performs complex tasks like writing code, scheduling appointments, and managing multi-step workflows. The resulting computing consumption has risen exponentially.
The numbers are striking: token usage on OpenAI’s API platform surged from 6 billion per minute last October to 15 billion per minute in late March. Meanwhile, hourly rental prices for Nvidia‘s most advanced Blackwell-generation GPUs have jumped to $4.08, a 48% increase from just two months ago.
Supply-side constraints are now evident. Anthropic has announced it will ration computing access during peak weekday hours. OpenAI has scrapped its Sora video-generation app, in part to redirect computing resources to higher-priority products. The CEO of cloud infrastructure company Vultr stated: “This is a massive capacity crunch that’s unlike anything I’ve seen in more than five years of running this business.” More critically, available power capacity through 2026 is already fully spoken for – meaning new data centers cannot come online anytime soon.
Peel back one layer from the computing crunch, and you expose a deeper, more intractable problem: electricity. A detailed report released this week by Bloomberg shows that the North American Electric Reliability Corp. forecasts US power demand in summer will rise by 224 gigawatts over the next decade – roughly equivalent to adding 180 million homes. One analyst noted that the last comparable surge came during World War II. The AI buildout is a central driver of this increase.
Take the PJM Interconnection, the nation’s largest power grid. Over just the three years ending in 2028, data centers are projected to add at least $23 billion to customer bills on the PJM grid alone – an increase of more than 50%. In parts of eastern Pennsylvania, electricity prices have already risen 200% since 2020.
In response to public discontent, President Trump in his February State of the Union address told major tech companies they would “have the obligation to provide for their own power needs.” Subsequently, seven leading firms including Amazon, Google, and Microsoft signed the “Ratepayer Protection Pledge,” committing to build, bring, or buy all the power and grid infrastructure required for their data centers, with none of those costs passed to American households.
However, pledges don’t immediately translate into reality. The electricity price increases documented by Bloomberg have been building for years: PJM’s capacity prices have exploded from $28.92 per megawatt-day in the 2024-25 delivery year to $329.17 in the 2026-27 delivery year – an 11-fold increase already baked into utility rate structures long before any documents were signed at the White House. Even for self-built power plants, permitting takes two to four years, and construction takes even longer. By the time the concrete is poured, the demand it was designed to meet has often already doubled.
History has repeatedly validated a rule: in every major tech boom, the biggest winners are often not the headline-grabbing companies, but the upstream suppliers providing the “picks and shovels.”
Veteran tech investment analyst Brian Hunt has built an entire framework around this dynamic, which he calls the “AI demand shock” – when demand for a specific resource or manufactured product suddenly skyrockets, its price can soar by hundreds of percent. In 2023, the AI demand shock for advanced semiconductors sent Nvidia up 525% in under two years. The sudden need for data center cooling systems drove Comfort Systems USA up 1,000%. Meanwhile, demand for advanced optical systems drove Lumentum Holdings up 1,164% in two years. None of these are AI companies themselves; they are the “picks-and-shovels” suppliers of AI infrastructure.
Currently, Brian is focusing on a surprising sector: chemicals. From the specialty gases and ultra-pure materials used to manufacture semiconductors, to the coolants, flame retardants, and advanced polymers essential for modern data centers – every AI server relies on a long, ultra-pure chemical supply chain. As AI usage explodes, the need for both the quantity and purity of chemicals will only intensify.
Returning to the token shortage at the opening of this Digest – it is real, it is happening now, and it is the best lens through which to understand the current technology transition. Behind the token shortage lies a GPU shortage. Behind that lies a data center shortage. Behind that lies a power shortage. And behind that lies a physical infrastructure buildout that the real world simply hasn’t had time to complete. The real money is flowing to companies positioned at the center of each of these constraint points – the ones making the chemicals, cooling the servers, and supplying the power infrastructure. These are the areas where investors may want to be.
