AI 1.0 Has Failed, the Data Is Undeniable

Stephen Klein

Is Tech's Next Bubble Near?

If so, then what's in store for humanoids, industrial robots & converged AI/robotics?

Stephen Klein’s insightful columns constantly skewer AI as highly vulnerable and potentially short lived. His latest calls AI 1.0 as failed, and needing an AI 2.0 as a necessary transition

by Stephen Klein:

“AI 1.0 promised to replace human thinking. The data says it failed. AI 2.0 promises to elevate human thinking. The data says it works.”
—Stephen Klein: Co-Founder & CEO at Curiouser.AI

Here’s Klein’s AI 1.0 Has Failed, the Data Is Undeniable, and our infographic to better visualize the transition. Plus…a look at the potential AI 1.0 bubble bursting and what it could mean for robotics:

Yesterday, Fortune reported on a new NBER [National Bureau of Economic Research] study of 6,000 CEOs across four countries.

90% of firms said AI had zero impact on productivity over the past three years.

Zero.

Economists are resurrecting Solow’s Productivity Paradox from 1987: “You can see the computer age everywhere but in the productivity statistics.”

Here’s the full picture:

MIT found 95% of enterprise AI pilots deliver no measurable P&L impact.
Microsoft Copilot has penetrated just 3% of its 450 million paid seats.
Gartner says only 6% of enterprises moved AI past pilot.

S&P Global: 42% of companies scrapped most AI initiatives in 2025, up from 17% the year prior.
Deloitte: only 20% of leaders report actual revenue growth from AI. 74% still “aspire” to it.

78% of organizations report active AI use. Only 1% consider themselves mature.

That’s not an adoption problem. That’s a paradigm problem.

The Bank of England warned of a global market correction from AI overvaluation. The IMF compared it to the dot-com bubble.

A top economist says three of four “horsemen” of a bubble are present. The only missing piece is an IPO wave.

OpenAI plans to go public in Q4 2026.

$1.1 trillion in projected mega-cap AI spending through 2029. And 90% of CEOs say it hasn’t moved the needle.

This isn’t a technology failure. It’s a premise failure.

Here's a Graphic Look at the Data from Above

AND WHAT ABOUT BUBBLE'S IMPACT ON ROBOTICS?
HERE'S WHAT THE FORECAST LOOKS LIKE TO 2030

Humanoid shake-out
Humanoid and AI‑robotics startups are already flagged as entering a speculative zone, with many prototypes, few deployments, and real challenges in dexterity, reliability, and cost. A sharp AI correction would almost certainly crush late‑stage valuations here, kill weaker players, and stretch timelines for the survivor.

Workhorse robotics industry—industrial, logistics, and service systems with clear ROI—probably experiences a funding rotation toward it, tighter discipline on AI claims, and slower but more sustainable AI integration into existing platforms.

Strategic and corporate funding. Automation/robotics funding is already tilting away from pure VC toward corporate balance sheets and debt. In a bust, hyperscalers, OEMs, and large integrators will still fund AI‑infused robotics that protect their labor, logistics, and supply‑chain positions, even as financial investors pull back..

Central‑bank work comparing AI to the 1990s telecom boom suggests that, even if there is short‑run over-investment and a correction, long‑run demand for AI‑related infrastructure and embedded capabilities is likely to catch up, benefiting “workhorse” automation more than the frothiest frontier plays.

Talent and cost tailwinds
If general AI hiring and cloud projects slump, robotics gets access to cheaper ML talent and under‑utilized compute, easing constraints that have kept some real‑world robotics projects on the back burner.

Enterprise surveys show that 42% of companies abandoned most of their AI initiatives in 2025, up from 17% in 2024, with nearly half of AI proofs‑of‑concept scrapped before production due to weak business impact.

Net impact: painful at the edge, constructive in the core
Analysts estimate hundreds of billions in economic loss from failed AI projects, reinforcing board‑level pressure to cut experiments that don’t deliver measurable value.