As September 2025 unfolds, the artificial intelligence industry stands at a precipice that those who lived through previous AI winters find disturbingly familiar. The field has experienced several hype cycles, followed by disappointment and criticism, followed by funding cuts, followed by renewed interest years or even decades later. The warning signs of an impending AI winter are practically screaming from the headlines. And the road, is still a bit icy.

The Anatomy of Past Winters

The First Freeze (1974-1980)

The AI field’s first major reckoning came in the mid-1970s. Between 1956 and 1974, the U.S. Defense Advanced Research Projects Agency (DARPA) funded AI research with few requirements for developing functional projects. This golden age of unfettered funding created an environment where bold promises flourished, but practical results lagged dangerously behind expectations.

AI researcher Hans Moravec blamed the crisis on the unrealistic predictions of his colleagues: “Many researchers were caught up in a web of increasing exaggeration. Their initial promises to DARPA had been much too optimistic”. The breaking point came with the devastating 1973 Lighthill Report, which criticized the utter failure of AI to achieve its “grandiose objectives” and concluded that nothing being done in AI could not be done in other sciences.

The consequences were swift and brutal. By 1974, funding for AI projects was hard to find, with DARPA cutting funding “deeply and brutally.” This was a complete loss of faith in the field’s fundamental premises.

The Second Great Cooling (1987-2000)

Just when AI seemed to have recovered with the expert systems boom of the early 1980s, history repeated itself with even greater severity. By 1985, they were spending over a billion dollars on AI, most of it in in-house AI departments. An industry grew up to support them, including software companies like Teknowledge and Intellicorp (KEE), as well as hardware companies like Symbolics and LISP Machines Inc.

Then came 1987. Three years after Minsky and Schank’s prediction, the market for specialized LISP-based AI hardware collapsed. Desktop computers from Apple and IBM had been steadily gaining speed and power, and in 1987, they became more powerful than the more expensive Lisp machines made by Symbolics and others. There was no longer a good reason to buy them. An entire industry worth half a billion dollars was demolished.

The human cost was devastating. The case of Google’s AI workers being laid off provides a modern parallel, but in the late 1980s, entire companies vanished overnight. Not only that, but the cost to develop these systems for each company was difficult to justify. Many corporations could not afford the costs of AI systems.

Why AI Winters Happen: The Immutable Laws of Hype

Understanding why AI winters occur requires examining the fundamental dynamics that drive both boom and bust cycles. AI winters occur when the hype behind AI research and development starts to stagnate. They also happen when the functions of AI stop being commercially viable.

The pattern is remarkably consistent:

1. Over-promising: Researchers and companies make grandiose claims about AI’s imminent capabilities
2. Over-funding: Investors and governments pour money into AI projects based on these promises
3. Under-delivering: The technology fails to meet the inflated expectations within the promised timeframes
4. Backlash: Disappointment leads to funding cuts and public skepticism
5. Winter: Research stagnates as resources dry up

However, what precipitates an AI winter is the emergence of definitive evidence that this hype cannot be met. In the 1970s, it was the Lighthill Report and the failure of machine translation projects. In the 1990s, it was the collapse of expert systems and specialized hardware markets.

Today’s Bubble: Eerily Familiar Warning Signs

The parallels between current AI hype and past bubbles are striking and ominous. OpenAI’s board chair, Bret Taylor, acknowledges what many in the industry refuse to admit: “I think we’re also in a bubble, and a lot of people will lose a lot of money.” CEO Sam Altman echoes this sentiment, warning that “someone is going to lose a phenomenal amount of money.”

The Infrastructure Overspend

Today’s AI bubble has created an infrastructure gold rush that dwarfs anything seen in previous eras. Hundreds of billions of dollars are being invested in AI datacenters being constructed by Microsoft, Alphabet, Amazon’s AWS, Elon Musk’s X.ai, and Meta. OpenAI is working on its $500 billion Project Stargate data center plan with Softbank, Oracle, and other investors.

This massive capital deployment mirrors the specialized hardware investments of the 1980s. Just as LISP machines became obsolete when general-purpose computers caught up, there’s a real risk that much of today’s specialized AI infrastructure could become stranded assets if the technology fails to deliver expected returns.

The Startup Ecosystem at Risk

The current AI ecosystem has spawned thousands of startups whose entire business models depend on external API access to large language models. This creates a dangerous dependency chain remarkably similar to the expert systems companies that relied on specialized hardware in the 1980s. When the hardware market collapsed, these companies had no fallback.

As we highlighted in the Google contractor layoffs, these AI workers represent a new class of digital laborers who could disappear overnight. Workers still at the company claim they are increasingly concerned that they are being set up to replace themselves. The irony is profound: AI workers are training their own replacements while being laid off for organizing around better working conditions.

The Research Reality Gap

Recent studies are beginning to challenge the fundamental assumptions underlying current AI hype, echoing the academic skepticism that precipitated previous winters. Two recent research papers from researchers at Apple and Arizona State University have cast doubt on whether the cutting-edge AI models, which are supposed to use a “chain of thought” to reason about how to answer a prompt, are actually engaging in reasoning at all.

This mirrors the devastating critiques of earlier AI approaches, where closer examination revealed that systems weren’t actually doing what their proponents claimed.

The 2025-2026 Inflection Point

My prediction from two years ago placed the bubble’s peak around September 2025. While the timeline is extending into Q4 2025 or Q1 2026, consistent with enterprise technology adoption cycles and shareholder interests, the fundamental dynamics remain unchanged.

The trigger for the coming winter will likely be a combination of factors:

1. Earnings Reality: As companies are forced to demonstrate actual ROI from their AI investments, many will find the economics don’t work
2. Technical Limitations: The current approach may hit fundamental scaling limits, just as expert systems encountered the “qualification problem.”
3. Market Saturation: The easy AI applications have been addressed; harder problems require breakthrough innovations that may not materialize
4. Economic Pressure: Rising interest rates and economic uncertainty will force more disciplined capital allocation

The Coming Casualties and Survivors

Who Will Suffer Most

The coming AI winter will be particularly brutal for:

• API-dependent startups: Companies built on OpenAI or similar platforms with no proprietary AI capabilities
• Pure-play AI consulting firms: Without differentiated expertise, these will consolidate rapidly
• Late-stage AI investments: Venture funds that invested at peak valuations will face massive write-downs
• AI-washing companies: Firms that added “AI” to their pitch without fundamental value propositions

The Likely Survivors of an AI Winter

Drawing parallels from the dotcom crash, the survivors will likely be:

• Companies with real, measurable AI value: Those showing clear ROI and customer traction
• Platform players: Companies providing essential infrastructure that remains valuable regardless of hype cycles
• Domain-specific AI winners: Firms applying AI to solve specific, well-defined problems with measurable outcomes
• Established tech giants: Companies with diversified revenue streams and the resources to weather the storm

Lessons from the Amazon and Google Examples

The dotcom bubble’s aftermath provides crucial guidance. Companies like Amazon and Google survived and thrived by focusing on fundamental value creation rather than hype. Amazon’s stock fell 94% from its peak but recovered because the company was building real infrastructure and solving genuine customer problems.

Today’s AI survivors will be those companies that:

• Focus on solving specific, measurable problems rather than pursuing artificial general intelligence.
• Build sustainable business models not dependent on external AI APIs
• Maintain realistic expectations about AI’s current limitations.
• Invest in proprietary data and algorithms rather than just wrappers around existing models.

Preparing for an AI Winter

The larger question is how to position for survival and eventual success. For entrepreneurs, investors, and executives, this means:

1. Ruthless focus on unit economics: Can your AI application show clear, measurable value?
2. Reduced dependency on external APIs: What happens to your business if OpenAI raises prices or restricts access?
3. Conservative cash management: Can you survive 18-24 months of reduced funding availability?
4. Realistic timeline planning: Are your development milestones achievable with current technology?

The Silver Lining: Why AI Winters Matter

AI winters, while painful, serve crucial functions. They force the field to:

• Abandon approaches that don’t work.
• Focus on practical applications rather than theoretical possibilities.
• Develop more efficient methods for achieving results.
• Build stronger foundations for the next wave of innovation.

Despite the lack of direct funding for AI, advancements in microprocessors, data storage, and distributed computing during the AI Winters laid the groundwork for future breakthroughs. Today’s AI winter, when it comes, will similarly clear away unsustainable approaches and focus efforts on what actually works.

The Temperature is Already Dropping

The signs are unmistakable for those who know how to read them. The AI industry is exhibiting the same pattern of over-promising, over-funding, and inevitable under-delivery that characterized previous boom-bust cycles. The documented layoffs of workers at Google’s AI division, despite record profits and market valuations, show that even successful companies are preparing for leaner times ahead.

The smartest players in AI aren’t those building for the peak of the hype cycle. They’re those preparing for the valley that follows. History shows that AI winters are temporary, but they’re also inevitable. The companies and careers that survive the coming freeze will be those built on solid fundamentals rather than inflated expectations.

The only question now is whether we’ve learned enough from history to make this winter shorter and less devastating than its predecessors. Given the scale of current investments and the breadth of current hype, that seems unlikely.

Prepare for winter. It’s coming sooner than most people think.

---

If you find this content valuable, please share it with your network.

🍊 Follow me for daily insights.

🍓 Schedule a free call to start your AI Transformation.

🍐 Book me to speak at your next event.

Chris Hood is an AI strategist and author of the #1 Amazon Best Seller “Infailible” and “Customer Transformation,” and has been recognized as one of the Top 40 Global Gurus for Customer Experience.