Guest Post by Chris MacIntosh
Numbers don’t lie, though naturally the people spending them often do.
IBM’s CEO Arvind Krishna recently did something refreshingly rare: he mentioned basic math. One gigawatt data center costs $80 billion to fill.
The AGI crowd is chasing 100 gigawatts of compute. That’s $8 trillion in capex. The numbers are far too big to be funded by equity, which means it’s debt, which means you need roughly $800 billion in annual profit just to service the interest on capex spend like that.
Honestly, I want to see the business model for this stuff.
OpenAI is burning cash at a rate that would make a Weimar central banker blush, and HSBC says they won’t turn a profit before 2030 — at minimum. They’ll need another $200 billion in debt or equity just to keep the lights on. Meanwhile, their valuation has “soared to unprecedented levels” on the back of what the Wall Street Journal delicately describes as the absence of a “clear financial model for profitable AI.” Hahaha!
Translated: they’re lending against a religion, not a business. This entire thing feels extraordinarily cult-like.
Anyway, Krishna put the AGI thesis plainly — he called it a belief system. He gives current technology a 0-1% chance of actually getting there. Not zero, to his credit, but close enough that no rational capital allocator should be writing trillion-dollar checks against those odds.
Generative AI is useful. It will probably unlock real productivity. Krishna said as much. But “useful enterprise software” and “justify $8 trillion in speculative infrastructure spend” are not the same sentence, and conflating them is how you end up with the dot-com wreckage circa 2001 — lots of real technology, catastrophically mispriced.
We recall that dot-com era well. Sure the internet was transformative and has changed the world, but that didn’t stop most of the companies in the space from blowing away in the wind. Even the few that managed it through took decades to get back to breakeven. And that is before accounting for inflation, which if we did so makes things even worse.
Now, here’s where it gets interesting… and where the real money is likely being made.
Every gigawatt of compute needs power. Serious, uninterrupted, industrial-scale power. Not the kind you’ll get from intermittent wind, solar or unicorn farts. The kind, only found in fossil fuels. The hyperscalers are already signing 20-year offtake agreements with nuclear operators, buying stranded gas assets, and lobbying hard to keep coal plants online that were scheduled for retirement.
Microsoft inked a deal to restart Three Mile Island. Amazon is circling nuclear sites across the Midwest. Google is funding next-generation geothermal. I’m old enough to remember when these very same companies were all chastising these same industries for killing the spotted owl and participating in the climate change hoax for brownie points at the ESG podium. In any event, reality is now reasserting itself.
The AI capex bonfire is, functionally, a massive subsidy to energy infrastructure.
Whether or not OpenAI ever posts a profit, the energy still gets consumed. Whether or not AGI materialises in our lifetimes, or at all, the data centres still need cooling, copper, transformers, and baseload generation that renewables alone cannot provide. The speculative layer lies in the software valuations, the model companies, the GPU darlings — that’s where the bubble is. But the physical substrate underneath it that allows it to exist at all is where scarcity exists, and where few are looking.
Relating it to the dot-com era and bust, recall that it destroyed Pets.com and made Cisco’s customers — the ones who actually owned the fibre — permanently richer.
Editor’s Note: The AI bubble may collapse, but the forces inflating it pose a much greater danger than falling technology stocks. Decades of debt creation and currency debasement now threaten the dollar and the purchasing power of your savings.
Legendary investor Doug Casey calls what lies ahead the most dangerous event of the 21st century. In a special video warning, he explains the threat, reveals the “inflation-proof” asset he favors, and shows you how to prepare.
Click here to watch Doug Casey’s free video warning.
Tweet
Continue reading...
[ H/T The Burning Platform ]
Numbers don’t lie, though naturally the people spending them often do.
IBM’s CEO Arvind Krishna recently did something refreshingly rare: he mentioned basic math. One gigawatt data center costs $80 billion to fill.
The AGI crowd is chasing 100 gigawatts of compute. That’s $8 trillion in capex. The numbers are far too big to be funded by equity, which means it’s debt, which means you need roughly $800 billion in annual profit just to service the interest on capex spend like that.
Honestly, I want to see the business model for this stuff.
OpenAI is burning cash at a rate that would make a Weimar central banker blush, and HSBC says they won’t turn a profit before 2030 — at minimum. They’ll need another $200 billion in debt or equity just to keep the lights on. Meanwhile, their valuation has “soared to unprecedented levels” on the back of what the Wall Street Journal delicately describes as the absence of a “clear financial model for profitable AI.” Hahaha!
Translated: they’re lending against a religion, not a business. This entire thing feels extraordinarily cult-like.
Anyway, Krishna put the AGI thesis plainly — he called it a belief system. He gives current technology a 0-1% chance of actually getting there. Not zero, to his credit, but close enough that no rational capital allocator should be writing trillion-dollar checks against those odds.
Generative AI is useful. It will probably unlock real productivity. Krishna said as much. But “useful enterprise software” and “justify $8 trillion in speculative infrastructure spend” are not the same sentence, and conflating them is how you end up with the dot-com wreckage circa 2001 — lots of real technology, catastrophically mispriced.
We recall that dot-com era well. Sure the internet was transformative and has changed the world, but that didn’t stop most of the companies in the space from blowing away in the wind. Even the few that managed it through took decades to get back to breakeven. And that is before accounting for inflation, which if we did so makes things even worse.
Now, here’s where it gets interesting… and where the real money is likely being made.
Every gigawatt of compute needs power. Serious, uninterrupted, industrial-scale power. Not the kind you’ll get from intermittent wind, solar or unicorn farts. The kind, only found in fossil fuels. The hyperscalers are already signing 20-year offtake agreements with nuclear operators, buying stranded gas assets, and lobbying hard to keep coal plants online that were scheduled for retirement.
Microsoft inked a deal to restart Three Mile Island. Amazon is circling nuclear sites across the Midwest. Google is funding next-generation geothermal. I’m old enough to remember when these very same companies were all chastising these same industries for killing the spotted owl and participating in the climate change hoax for brownie points at the ESG podium. In any event, reality is now reasserting itself.
The AI capex bonfire is, functionally, a massive subsidy to energy infrastructure.
Whether or not OpenAI ever posts a profit, the energy still gets consumed. Whether or not AGI materialises in our lifetimes, or at all, the data centres still need cooling, copper, transformers, and baseload generation that renewables alone cannot provide. The speculative layer lies in the software valuations, the model companies, the GPU darlings — that’s where the bubble is. But the physical substrate underneath it that allows it to exist at all is where scarcity exists, and where few are looking.
Relating it to the dot-com era and bust, recall that it destroyed Pets.com and made Cisco’s customers — the ones who actually owned the fibre — permanently richer.
Editor’s Note: The AI bubble may collapse, but the forces inflating it pose a much greater danger than falling technology stocks. Decades of debt creation and currency debasement now threaten the dollar and the purchasing power of your savings.
Legendary investor Doug Casey calls what lies ahead the most dangerous event of the 21st century. In a special video warning, he explains the threat, reveals the “inflation-proof” asset he favors, and shows you how to prepare.
Click here to watch Doug Casey’s free video warning.
Tweet
Continue reading...
[ H/T The Burning Platform ]
