The scale of investment in artificial intelligence is extraordinary.
According to a recent Wall Street Journal analysis, nine of the world’s leading technology companies have accumulated more than $3 trillion in off-balance-sheet commitments, much of it associated with the AI infrastructure buildout.
That’s on top of roughly $600 billion in disclosed capital expenditures.
And the money isn’t just going into software.
It’s going into semiconductor fabs, advanced packaging, memory, data centers, networking, power generation, electrical equipment, cooling systems and construction.
AI may be digital, but the infrastructure required to support it is enormously physical.
The Information Age is beginning to look like an industrial age.
The obvious question is whether all this investment will generate economic returns sufficient to justify its extraordinary cost.
What does history teach us?
The railroads transformed economies, but railroad booms also produced enormous speculation, overbuilding and financial failures.
The Internet transformed the world, but the dot-com bubble destroyed hundreds of companies and billions of dollars of investment.
Both were genuine technological revolutions.
Both also produced bubbles.
Every industrial revolution creates winners and losers. To the winners, it looks like a revolution. To many of the losers, it looks like a bubble.
AI will be no different.
Some of today’s enormous investments in chips, data centers and power infrastructure may produce extraordinary returns. Others will never earn their cost of capital.
Some AI companies will become some of the most important companies in the world. Many others will probably disappear.
AI doesn’t have to justify every dollar being invested in it to change the world.
The technology can transform the global economy even if some of today’s investments prove wildly excessive.
The revolution can be real. The bubble can be real too.
The challenge is figuring out which investments belong to which.
#ArtificialIntelligence #AI #Technology #Semiconductors #DataCenters