What does the artificial intelligence boom have to do with Iron Ore?
Fourteen years ago, in 2011, a commodity boom in iron ore saw BHP’s share price hit $40 for the first time. It was the 8th of April 2011. There was great excitement, surrounding Australia being the ‘lucky’ country yet again. Peter Richardson, Morgan Stanley’s then global metals chief economist, put forward a strong investment case for the “crucial” steelmaking commodity.
We will see Morgan Stanley again soon.
By contrast, on April 11 that year we published a blog Will China demand Iron… or…?, where we wrote that iron ore prices would henceforth decline, ending the commodity boom and causing buoyant share prices to fall.
A history lesson on booms
The reason was simple, and I highlight our various arguments here;
“For what it’s worth (and I know it’s not a widely accepted view right now), I think commodities are cyclical. When prices are low, there’s precious little investment in building productive capacity; the lack of investment and long production lead times result in supply lagging demand. As prices rise, investments are proposed, delayed and then made, and this pattern causes prices to extend their rise. Then, just as prices peak, marginal operators come online, backed by Net Present Value (NPV) calculations that assume the high prices will be sustained. It’s in their interest to be bullish about the future, because their jobs and reputations – as well as the dollars they have attracted as capital – are all on the line. But eventually, supply increases and prices stop going up; if it can’t go on forever, eventually it must stop.”
“In 2010-11, world iron ore production grew 8.1 per cent or 227mt to 2.8bt. Assuming similar growth levels in 2011-12 – in a classic supply response BHP production is forecast to grow by 20 per cent, Return On Investment (RIO) by 30 per cent, FMG 25 per cent – iron ore production will grow to 3,037bt, an increase of 237mt.
“And assuming China consumes 60 per cent of global production again (highly optimistic), their demand would increase by 136.2mt. However moderating [economic] growth means current estimates for China’s iron ore requirements are half this level. With few other countries growing or competing heavily with China, who will pick up that supply overhang in a low-growth environment?
“By 2015 we estimate that two entire Pilbara regions (700mt) in supply terms will come onto the market. It’s a far stretch to expect China to absorb 420mt (60 per cent) of that. The impact we expect is pressure on iron ore prices.”
By January 2016 BHP’s share price had fallen 66 per cent to $13.59.
Never ask a barber whether you need a haircut!
Why the history lesson?
It’s standard practice for analysts to focus on individual companies. They speak to managers, CEO, CFOs and Investor Relations representative of individual companies to formulate a forecast of the company’s revenues, margins and earnings. That’s their job.
Less frequently however do they ask; “If we aggregate the production or revenue for every company we are covering are the resulting numbers even possible?” Or, “Can the proposed aggregate volumes and revenues be supported by the existing pool of buyers?”
That’s how, in 2011, we concluded the iron ore price would decline and made a decision to reduce our holdings of companies exposed to the boom. The predicted volume of iron ore production was equivalent to two entire Pilbaras, and it would be impossible for China to absorb it, especially as its economic growth was slowing.
When we sum the individual companies, we sometimes find that reasonable individual estimates become unreasonable in aggregate.
That’s when you know you’re in a bubble – by swapping the microscope for a telescope.
Thinking aloud
Today’s frenzy isn’t iron ore, its artificial intelligence (AI). And, who knows, Large Language Models (LLMs) might become a commodity too, but that’s not the point.
Just as China couldn’t support the growing aggregate output of Australia’s iron ore giants, a recent analysis by Dallas, Texas-based St James Investment Company has me suspecting the world can’t buy enough AI tools to support a reasonable return on investment (ROI) for today’s AI super spenders.
Massive sums are being funnelled into data centres to power artificial intelligence. Analysts at Morgan Stanley (there they are again!) predict that total spending on these facilities could reach a staggering $3 trillion by 2028 – not even accounting for skyrocketing energy costs. McKinsey takes it further, projecting $5.2 trillion by 2030.
Meanwhile, stock market investors seem convinced this is money well spent. Ever since ChatGPT launched in late 2022, the big six U.S. tech giants – NVIDIA, Microsoft, Apple, Amazon, Meta, and Alphabet – have seen their collective market caps balloon by almost $12 trillion.
According to McKinsey, again, about 60 per cent of this AI-related data centre spending will go toward semiconductor chips and other hardware. These assets typically wear out and are depreciated over roughly 5.5 years. If we assume they don’t produce any real economic value after that – and factoring in that much of Morgan Stanley’s $3 trillion estimate is back-loaded toward the later years – the hardware and semiconductor investments alone would require generating more than $500 billion in net cash flow just in 2028 to cover the basic cost of capital on the equipment investment.
To make the math work for data centre operators aiming for a healthy 20 per cent free cash flow margin (which would validate their lofty stock prices), they’d need to pull in around $2.5 trillion in revenue annually.
OK, now assume the data centres’ customers – the businesses and consumers buying these AI services – also want to generate a 20 per cent margin. They would need to spend about $3.1 trillion on AI services like agents and LLMs.
And here’s the problem; St James notes that $3.1 trillion is equivalent to roughly 10 per cent of the entire U.S. gross domestic product (GDP) right now. To put that figure in context, total U.S. military spending is about 3.46 per cent of GDP. According to U.S. Treasury, total federal government spending in FY 2025 was $7.3 trillion, which was 23 per cent of GDP.
St James, provided some other comparison to understand the optimistic expectations implied by the required aggregate estimates for AI spending and revenue generation. Netflix, with its army of about 300 million subscribers, is on track for around $39 billion in revenue this year. Microsoft’s Office 365 suite, a staple for businesses worldwide, brings in about $95 billion. Even OpenAI, boasting the biggest AI chatbot user base, is generating roughly $13 billion annually. AI’s transformative potential is undeniable, and its growth story is just getting started. Yet, the sheer scale of revenue and spending on AI tools – $3.1 trillion – needed to recoup today’s capital outlays highlights AI’s requirement to monetise fast to justify the hype.
As investors, we need to remember that bubbles occur when disconnects become unreasonable. Some companies in the AI space will succeed, but many will prove to be ‘me-too’ businesses that fall behind and just burn capital. Some, for example, will commit to data centres where the power is too expensive, others will buy a generation of chips and be unable to capture all of the value, before the next generation of chips supersede those they acquired. Unfortunately, in a race for funding and market share, AI tech companies don’t have the option of pausing for a minute and letting the race pass them by. They will raise billions, but they will prove to be dead-end investments, and many billions will be lost.
There is no question that AI is the biggest technology development in our lifetimes. As I have pointed out many times, it will profoundly transform humanity. But as I have also said, often new technologies – think the motor vehicle, commercial air-travel and television – benefit consumers more than investors, especially when investors pay frenzied prices created by the promises for humanity.
Roger Montgomery is the Founder and Chairman of Montgomery Investment Management. Roger has over three decades of experience in funds management and related activities, including equities analysis, equity and derivatives strategy, trading and stockbroking. Prior to establishing Montgomery, Roger held positions at Ord Minnett Jardine Fleming, BT (Australia) Limited and Merrill Lynch.
He is also author of best-selling investment guide-book for the stock market, Value.able – how to value the best stocks and buy them for less than they are worth.
Roger appears regularly on television and radio, and in the press, including ABC radio and TV, The Australian and Ausbiz. View upcoming media appearances.
This post was contributed by a representative of Montgomery Investment Management Pty Limited (AFSL No. 354564). The principal purpose of this post is to provide factual information and not provide financial product advice. Additionally, the information provided is not intended to provide any recommendation or opinion about any financial product. Any commentary and statements of opinion however may contain general advice only that is prepared without taking into account your personal objectives, financial circumstances or needs. Because of this, before acting on any of the information provided, you should always consider its appropriateness in light of your personal objectives, financial circumstances and needs and should consider seeking independent advice from a financial advisor if necessary before making any decisions. This post specifically excludes personal advice.
