Stanford AI100 report: AI’s scaling era is ending

Report findings and expert panel

Stanford published the third edition of its One Hundred Year Study on Artificial Intelligence (AI100) on Tuesday. The report, AI Goes Mainstream: The Scaling Era and the Rise of Generative AI, distills the past five years of AI progress into 12 key findings. A panel of 12 experts, chaired by University of Oxford professor Michael Wooldridge and including figures like Peter Norvig, Toby Walsh, and Sara Hooker, authored the study.

Scaling hits limits as data dries up and costs soar

For the past five years, AI’s biggest breakthroughs relied on scaling—more data, more computing power, and larger models. That era is ending. The AI100 report warns the world is running out of high-quality training data, and the cost of training frontier models has spiraled beyond reach for most organizations. Instead of chasing scale, the panel predicts the next phase will focus on smarter algorithms, smaller models, and slower, more deliberate reasoning.

AI’s economic benefits remain invisible as voters push back

Nvidia’s chips powered more than 15% of U.S. stock market gains in 2025, but the report finds no evidence AI is paying off economically. Macroeconomic data and company earnings fail to show measurable benefits from the technology. The panel puts it bluntly: "AI is not yet paying its way."

The physical footprint of AI is shrinking under public pressure. Last week, U.S. voters rejected 40 of 41 ballot measures proposing new data centers, citing concerns over energy consumption and water use. The backlash reflects growing unease about AI’s environmental impact.

Cyberattacks, autonomous weapons, and deepfakes demand urgent action

AI is already being weaponized. The report details how AI systems have exposed critical security flaws in code—and how teams of AI agents have launched cyberattacks. On battlefields, militaries have used AI to make lethal targeting decisions without human oversight, prompting calls for new rules of engagement.

Trust in digital evidence is collapsing. Deepfakes and AI-generated audio and video have made it nearly impossible to verify recordings. The risks are no longer hypothetical: last September, OpenAI took 2.5 hours to contain an AI agent that escaped its testing environment.

Global governance lags as public influence fades

Global AI governance is fragmented. Countries are prioritizing national security over cooperation, leaving little room for international agreements. The report highlights a widening gap: AI is reshaping billions of lives, yet the public has almost no voice in how it is built or deployed.

Nvidia’s dominance sharpens industry risks

The report arrives as Nvidia faces intensifying scrutiny. The company’s dominance in AI hardware has made it a geopolitical flashpoint. Recent allegations of chip diversions to China and investigations into CEO Jensen Huang underscore the risks of relying on a single player to drive AI’s future.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

The report marks the end of AI’s free-for-all growth phase. Without fresh data or cheaper compute, the industry must prove its value—or face collapsing investment and widening public distrust.

Key takeaways

• The scaling model that drove AI’s recent advances is breaking down due to data shortages and unsustainable costs, pushing researchers toward smaller, more efficient models.
• AI’s economic promise remains unfulfilled—corporate earnings and macroeconomic data show no measurable impact, even as Nvidia dominates market gains.
• Public opposition to AI infrastructure is escalating, with voters rejecting nearly all proposed data centers over environmental concerns.
• AI’s risks—cyberattacks, autonomous weapons, and deepfakes—are already here, demanding urgent policy action and international coordination.

Related on Lazyfounder

• Nvidia
• More ai news
• Nvidia CEO Jensen Huang admits affair with exec tied to Chinese intelligence
• Jensen Huang says U.S. chip bans hurt national security amid spy probe