Business Opportunities After the AI ​​Bubble Bursts

It seems that the collapse of the AI bubble has become almost inevitable. However, this does not mean the “death of AI,” but rather a financial crisis triggered by the exposure of overinvestment. It is highly likely that those affected will be retail investors and pension funds that have nothing to do with AI. Of course, infrastructure companies that have used NVIDIA GPUs as collateral for their loans will likely be bought out at a bargain by well-funded major corporations.

There is no longer any point in worrying about “whether or not the AI bubble will burst,” so it seems it is time to start thinking about the post-bubble world. A hint for this can be found in asking: what was trending during the collapse of the dot-com bubble back in 2001?

The dot-com bubble era was characterized by overinvestment in infrastructure and expectations, where the primary goal was simply to “build websites and gather traffic (page views).” After the bubble burst, companies shifted toward pragmatism, asking “how to directly connect the gathered traffic and infrastructure to profits (ROI).” This shift gave birth to the data mining and CRM (Customer Relationship Management) boom, which focused on digging deeper into customer data.

When the current AI bubble (fueled by massive GPU investments and the performance race of general-purpose, ultra-large language models) bursts, it is highly likely that the focus will similarly shift from the “AI intelligence race” to the “recovery of ROI through AI.” At that time, the following pragmatism-oriented booms are expected.

4 Expected Booms After the AI Bubble Bursts

1. The Rise of Small Language Models (SLMs) and Autonomous Agents

A shift will occur from “huge, expensive AI that can answer anything” to “AI that handles specific tasks cheaply and reliably.”

• Agentic Workflows: Instead of humans giving step-by-step instructions via chat, we will see the widespread adoption of tools where AI autonomously completes entire business processes, such as expense settlement, initial contract review, and first-line customer support.
• SLMs (Small Language Models): Rather than giant models that consume massive computing resources, lightweight and cost-effective small models specialized for specific industries or internal company use will become the mainstream.

2. AI Governance and Compliance (AI Trust & Safety)

The phase of “being able to build and converse with AI” will end, and “how to safely integrate it into business operations” will become the top priority.

• Security and Auditing: “Security tools for AI” that monitor and control AI outputs—such as detecting AI hallucinations (falsehoods), checking for copyright infringement, and preventing confidential data leaks—will become a massive market.
• Accountability solutions (XAI: Explainable AI) and compliance management platforms designed to meet global AI regulations will likely be adopted by all companies, much like antivirus software was in the past.

3. “Data Engineering 2.0” for Unstructured Data

While data mining in 2001 targeted “structured customer behavioral logs,” the next boom will focus on organizing the “unstructured data” sleeping within companies.

• To customize AI (such as through RAG technology) for exclusive company use, internal data scattered across PDFs, meeting minutes, videos, and chat histories must be organized and integrated into a format that AI can read.
• Businesses that support the “cleansing and pipeline construction of internal data to be fed into AI” will generate explosive demand, arguably even more so than the actual “implementation of AI” itself.

4. Power Efficiency of Inference Infrastructure and Edge AI

Once the AI training race settles down, the operational costs and power consumption required to actually use (infer) AI will become the next major hurdle.

• Edge AI: Technology that processes AI locally on devices—such as smartphones, PCs, cars, and home appliances—without relying on cloud servers will become widespread.
• Green AI: Technologies that reduce the operational costs and environmental impact of AI through software optimization (like quantization) and power-efficient dedicated chips (NPUs) will experience a boom.

Historical Analogy: The Bubble and the Subsequent Pragmatic Phase

	2001 (Post Dot-com Bubble)	Future (Post AI Bubble)
Target of Overinvestment	Telecom infrastructure, servers, PV acquisition	GPUs, development of giant foundation models (LLMs)
Post-collapse Shift	Quantity of PVs → Improvement of LTV (Lifetime Value)	Model versatility → Cost reduction and automation of specific tasks
Driving Booms	Data mining, CRM, One-to-One Marketing	Specialized AI agents, AI governance, data preparation
Subsequent Development	Cloud, SaaS, Social Media, Smartphones	Invisible AI (AI blended into UI), autonomous robotics

The bursting of a bubble does not mean the end of a technology; it signifies a healthy transition from an “era of frenzy” to an “era of practical use and widespread adoption.” The unglamorous, gritty phase of “business implementation”—making AI take root in society as a true infrastructure—will be the epicenter of the next massive boom.

Which of these four upcoming trends do you think will have the most immediate impact on your industry?