AI software has driven unprecedented hardware investment, fueling massive AI data centers and a surge of start-ups. Enthusiasm peaked as costs, geopolitics, regulation, and skepticism grew, raising fears of a bubble. Yet tens of thousands of engineers continue pushing innovation. Even if many efforts fail or consolidate, the scale of activity ensures lasting progress. AI’s potential impact rivals past transformative technologies, provoking both optimism and concern about its future trajectory.
The software drives the hardware. It’s been true for the last 65 years. That means the software creates the need and, thus, the demand for hardware to make the software work—and work faster. AI, in its many manifestations and promise, and usually based on a demonstration of the latest R&D, has excited companies to get in the race and build huge AI data centers filled with AI supercomputers. Fed enormous amounts of precious water, electricity, and seemingly endless supplies of cash, AI has been the biggest thing ever (including wars) to rock the world. But it hasn’t gone completely the way many thought it might—reality caught up with enthusiasm. Talk of a bubble bubbled up, and bean counters got nervous. Geopolitical games developed between ego-driven leaders who could barely spell AI. And governments, i.e., politicians, jumped in with their infinite wisdom and desire to do the best for… well…. But the promise wouldn’t abate, and fear of missing out loomed large as a line item in BOD meetings, so investment and bold statements continued. Meanwhile, pundits, investors, and other smart people made predictions.
Gary Marcus, who made 17 prescient predictions in early 2025, discovered by the end of the year that most of them were on the money, so he made a few new ones for 2026, some of which are listed here:
These and other developments and promises about GAI, LLMs, agents, agentic, and neural networks have kept the fires high and encouraged the 109 AI processor start-ups to spend the $28 billion they raised to go up against the 29 much larger, more powerful public companies making AI processors.
It’s easy to dismiss the misallocation of resources and presumably unwarranted enthusiasm of the entrepreneurs and their backers, and call this a fool’s gold rush. But the reality is that there are over 30,000 very bright, very sleep-deprived engineers, architects, coders, and visionaries at start-ups doing stuff. And some of them are going to have a break, though, and probably get acquired, but the overall industry is going to leap forward. You can’t put that much energy and effort into something and not get an explosive result.
Others recognize this, and the consequences of AGI frightens them, given humanity’s record in handling dangerous new developments. If AI were just a bubble, why are so many people worried about it?
If you thought the steam engine, electricity, communications, atomic energy, semiconductors, and the Internet were explosive new developments that forever changed the world, you’ve got a front-row seat to the most significant economic, social, political, and dangerous new development in evolution since the discovery of fire. And almost every prediction you read is going to be true, simultaneously.
Epilogue
AI is the ability to recognize patterns. Patterns in speech, in images, in sound. Patterns that predict and, in some cases, promise.
If an AI system is a pattern recognition system, why can’t we use AI to predict AI?
AI is being used in war games, cancer research, drug development, and semiconductor design. Why not use it to predict how AI is going to develop?
The conspiracist in us says, they already have, and that’s why there’s so much enthusiasm and investment.
WHAT DO YOU THINK? LIKE THIS STORY? TELL YOUR FRIENDS, TELL US.