Ultimate Assessment Program 2027/28/29 || 1:1 Mentorship Batch Admissions Open

Why in the News

The artificial intelligence (AI) boom ran on one company, Nvidia, and its chief, Jensen Huang. He has exited after reports of an affair with an alleged Chinese spy, raising the question of whether the boom is turning into a bust. A counter case holds that a “good bust” would let power grids, regulators and laws catch up with the technology.

Why did the AI boom hinge on one man?

  1. Business model: Nvidia made the chips that took buyers to the frontier of AI and financed customers to buy them. Its chief then lobbied Washington to let more buyers access the chips.
  2. Unchecked power: Nvidia reached a value of $5.7 trillion, higher than India’s gross domestic product (GDP), without even a board chair. No one could question its chief.
  3. Fallout: His exit made the company less valuable and hit the entire AI revolution it had powered and bankrolled.
  4. Bubble risk: An AI bubble means valuations rising far above what the technology earns, as in the dotcom boom. The Economist estimated a dotcom-sized pop would wipe out 8% of American household wealth.
  5. The takeaway: A boom that depended on one firm’s chips, money and chief became fragile the moment that chief fell.

What would a “good bust” deliver?

  1. Political pushback: United States (US) midterm voters rewarded candidates who resisted unchecked AI, data-centre growth and rising power bills:
    • governors in nine states promised or imposed pauses on new data centres, stalling 41% of the US data-centre pipeline;
    • Democrats, who will now control Congress, want to regulate AI, and a pause gives them time.
  2. Grid relief: Developers file power requests in several jurisdictions at once, so utilities build in parallel and customers pay for it all. A bust ends much of that spending.
  3. Reusable infrastructure: Overbuilt dot-com fibre later made internet access cheap. Idle data centres can similarly run genomic sequencing, drug discovery and climate models, and their waste heat can be reused as energy.
  4. Frugal innovation: China’s DeepSeek built capable models at a fraction of the cost without the most advanced chips. In a bust, firms may release open-source and open-weight models, which anyone can download, run and modify.
  5. Institutions catch up: A Register of AI Risks & Remedies records 103 AI risk incidents, from biased welfare algorithms to wrongful arrests. For about four in five, a compensating or deterring law already existed.

What do past technology busts show, and at what cost?

  1. Electricity revolution: Competition among Thomas Edison, George Westinghouse and Nikola Tesla drove the electricity boom. A slowdown gave regulatory bodies room to form and guard the public interest.
  2. AI winters: AI has gone through at least two “winters”, periods when funding and interest collapsed, one in the 1970s and another from the late 1980s.
  3. Machine learning: After those winters, researchers turned to machine learning, where computers learn from data instead of explicit rules. That shift produced the present AI surge.
  4. Cost of a bust: Busts cause economic upheaval in the short run. Over the longer haul they give society and its laws time to adapt to a transformative technology.

Challenges

  1. Stranded assets: Part of the AI infrastructure built during the boom will find no use, and the investment in it will be lost.
  2. Wealth shock: A dotcom-scale pop would hit household wealth in the US and spill over to economies tied to American markets.
  3. Enforcement gap: A law on the books compensates no one until regulators and courts actually apply it to AI harms.
  4. Concentration risk: Reliance on a single chip maker for frontier AI exposes the whole industry to one firm’s governance failures.

Way Forward

  1. Enforce existing law: Regulators should apply existing liability and consumer protection laws to recorded AI harms before drafting new statutes.
  2. Shared compute: Offer idle data-centre capacity to research institutions for public-interest computing.
  3. Single grid request: Utilities should accept one binding power request per data-centre project, ending parallel filings.
  4. Board oversight: Firms that dominate critical AI supply chains should separate the board chair from the chief executive.

Conclusion

The AI boom outran the grids, laws and institutions meant to govern it, and its dependence on one firm exposed that gap. What to watch is whether US data-centre pauses and congressional regulation turn a slowdown into enforceable rules.

Key numbers

  1. First AI winter: 1974 to 1980.
  2. Second AI winter: 1987 to 1993.

Matching Previous Year Question

“[2026, GS3, 15 marks] What is agentic Artificial Intelligence (AI)? Explain its working. Describe its applications with suitable examples. Discuss the advantages, risks and challenges associated with agentic AI systems.”

Click here to Download X-Factor Notes for FREE. Cover the entire Microthemes with PYQs.

Download free X-Factor Notes
Revised and updated X-Factor Notes for UPSC Mains 2026: nine subject booklets