You are living through the collapse of one of the great tech bubbles of our time – and it will take much of the tech and financial infrastructure down with it.
Yet, this act of creative destruction, to use Joseph Schumpeter’s phrase, is making way for one of the greatest advancements in software and national productivity since the invention of the personal computer.
What may soon appear to be the worst of times – as this frenzy unwinds, will deliver the best of times – and we will give you early glimpses here at The Black Swan Files.
The Black Swan Files is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
The upcoming slow collapse comes from the failure of hubris and pride of the tech press and a few tech oligarchs – soon to add “former” to their surnames, who tried to con an entire planet but in the end could not outrun absent revenue numbers.
Many knew collapse was coming and did all they could to cash out before the bottom fell out:
The current question is not whether we will have data centers on every farm in America to outrun the Chinese in A.I. but which A.I. company will collapse after Oracle, because Oracle is likely, in every set of predictions – to fail first.
The perfect storm of A.I. hype collapse is because so many entities – software companies, investment banks, tech press, Charles Paynes of Fox News, venture capital companies, insurance portfolios, really stupid politicos like Data Center Dave McCormick from Pennsylvania – will be swept into the sea of blame when this baby unwinds.
The initial pain is already being felt by the stalled data center projects across America – half built data centers with no customers because Anthropic and ChatGPT cannot raise the money to drive their revenue – and their revenue appears insufficient to attract the dough.
What looked like a sure 2 trillion dollar IPO 60 days ago is a maybe IPO sometime in the perhaps future – and likely to never happen for either company.
No IPO means no payback, thus the literally trillion dollars in sunk data center costs may look to Amazon warehousing or mushroom farms as a possible fall back market.
Insurance companies – today the growing source of investment – will desperately try to unwind their data center portfolios of anything relating to data center build out.
But, who will buy their shaky investments? Who will bail them out?
This unwinding is already reflecting itself in interest rates for Oracle and data centers in junk bond status.
Those safe 401(k) portfolios holding mutual funds dug into Oracle, NVIDIA and a host of other soon-to-be shriveled investments will see their monthly statements showing disappointment to middle America who invested in them.
This is the perfect storm because it is no longer a maybe. It is no longer simply “inevitable” – as in the future – sometime.
It is happening, and anyone can see from the numbers – financial catastrophe is on the horizon.
The great A.I. data center con looked great until the numbers started showing up – and the numbers turned on the lights showing simply impossible paths to payback, profitability – and the certainty of costs outrunning revenue.
Revenue is always the stickler for bubbles – when everyone realizes the rev cannot possibly cover the costs, the party ends – sometimes quickly like the 2000 DotCom collapse, ENRON, Bear Stearns and Lehman Brothers – or as is likely to happen here – an 18 month, slow deterioration of the narrative until a new administration begins to look at criminal penalties for some of the players.
What is certain is the party is over for A.I. data center madness.
What is also certain, yet hidden by the in-the-tank-for-software-companies’ tech press – is the great promise of A.I. is quite healthy and a positive revolution is happening deep under the surface of the destroyed investments burning above them.
The future is A.I. everywhere – without massive, centralized data centers – because they are not needed, are too expensive, are too widely opposed – and are central computing at a time when central computing – as in cloud – is dying.
Distributed computing – at invariant scale – a term you will learn much more about here at The Black Swan Files – invariant scale computing is making centralized processing obsolete – thus data centers.
The perfect storm is the confluence of high interest rates making fatuous investments uglier sooner than they would be at 3% and scary revenue numbers.
The Middle East war, and its stark focus of the non-viability of centralized data centers – now being attacked by cheap drones – is waking up the sleepy Pentagon to the dangers of putting all its computing eggs in one basket.
Distributed computing is the answer, everyone knows it, and the Revolutionary Guard and the Houthis are educating the Pentagon planners as to why distributed compute is maybe something they need to employ – even though their Palantir, Andruil and Oracle donors cannot get them there.
A current political administration that went all in on A.I. data centers – questioning the patriotism of anyone who fought back against having an energy-consuming behemoth destroying their farmland – is soon to take the blame for much of the coming equity destruction.
Follow us at DataCenterMadness.com
Financial destruction is now a certainty because you cannot unwind sunk costs in land purchases, political payoffs, lobbying, building construction, power line rights of way – all of which add up to a trillion – mostly lost.
More significantly – nobody can unwind the visceral, deeply angry spirit across America as politicos tried, now unsuccessfully, to sell citizens on giving up their way of life, their homesteads – for some Tech Bros billionaires getting richer.
When a trillion dollars or close to it is spent, and after being spent is gone – because it is not generating revenue in humming data centers, many related businesses, and hopes vanish with it.
That is beginning to happen and as you watch interest rates for the “neoclouds” – a fancy word for new data centers – approach 13%, you have front row seats to the fire.
What you are not seeing is this forest fire is the necessary burning of the old, accumulated deadwood on the forest floor burning in a conflagration opening the ground for a Spring of new growth – and that growth is coming.
The demise of 1980s tech companies like Oracle, the entire family of ERP SaaS vendors, the hyperscaler is needed for A.I. myth – makes way for wondrous new technologies lowering the cost of compute 90%, implemented in days, bringing A.I. benefits to anyone who wants to use them.
We sat through a demo of a major ERP system (think SAP, Oracle, Palantir, Salesforce) which provided an A.I. system nothing but a user interface, and within 24 hours, a complete, parallel system was built – without the ERP vendor software.
It ran almost 1,000 times faster, at 1/10th the cost, on a computer you can hold in your hand.
The A.I. system never looked at a line of source code, never looked at how reports were formatted, it simply evaluated the underlying data and their relationships and a parallel system was born.
Thus the dependence on SAP, Oracle, Salesforce for very costly subscriptions, maintenance and gazillions of dollars in ongoing consulting revenue – is gone.
Any such ERP system, regardless of scale, can be replicated as a digital twin, in a weekend. In the system being demonstrated, the ERP system took 9 hours to crank out a report – and it could not use ALL the enterprise’s data – the system could not handle it.
The parallel ERP equivalent ran the reports in less than 2 seconds – across all data – instantly – on a computer smaller than a shoebox.
If you thought SaaS companies were dead meat, you were right – you just had the timing wrong – it’s coming and the economics cannot be contained.
That’s why almost every demo now has a couple of hedge fund guys watching it.
We sat through a demo of a hospital revenue management system which would have taken years to build, cost tens of millions of dollars with conventional software – moved from concept to production in a quarter – with A.I and massively distributed systems.
Any hospital can reclaim 10% to as much as 17% of its lost revenue with an A.I. system which pays for itself in a quarter. Early adopter hospitals are lining up to try out the future.
We watched as a group demonstrated how to protect the entire U.S. power grid with A.I. distributed technology – which can predict any part of the grid failing well before it happens – implemented without a data center – at a cost less than a day’s spend for Anthropic.
The future of A.I. is here – growing on that forest floor and this massive capital destruction of obsolete software companies is making way for new players to deliver the promise of A.I.
That promise does not include centralized data centers.
That promise is generating revenue, gaining early customers and partners and releasing the energies of subject matter experts who could never get their ideas to market with conventional software costs.
Remove the need for centralized data centers and you enable a lot of very smart people to deliver great solutions making America the leader in A.I.
Without central data centers.
