Is the AI Bubble About to Burst? Understanding the 2025 Market Wobble
Recent headlines warning of an “AI bubble burst” have dominated tech news as AI-linked stocks experience significant volatility. But before we panic, we need to understand: what does history tell us about technology booms, busts, and the real value that emerges afterward?
AI Winters: A Brief History of Artificial Intelligence Cycles
Artificial intelligence isn’t new. The AI field officially began at the 1956 Dartmouth Workshop, emerging from early 20th-century concepts of thinking machines. Since then, we’ve witnessed two “AI Winters”—periods when funding dried up, public interest waned, and expectations crashed after intense hype cycles.
This historical context raises a more nuanced question “What kind of market correction should we expect, and what comes after?”
Two Historical Mirrors: Dot-Com Boom and Railroad Expansion
To understand where AI is headed, two historical parallels offer valuable insights:
- The late-1990s dot-com bubble
- The 19th-century U.S. railroad build-out
Both periods featured genuine technological breakthroughs alongside speculative excess. Neither ended with the death of the technology—instead, they reshaped winners and losers while redefining how value was created.
Today’s AI landscape reflects a fusion of these patterns: exuberant capital flows, massive infrastructure builds, and uneven yet real utility.
The Dot-Com Era: Eyeballs First, Profits Later
During the dot-com boom, the prevailing wisdom was “eyeballs first, profits later.” Companies attracted users but lacked viable business models. Sound familiar?
Similarly, the railroad boom featured promotional claims and proliferating charters that raced ahead of actual demand. AI’s current phase echoes both scenarios: valuations price in world-changing impact now, while operational benefits will arrive later as enterprises re-engineer processes, data infrastructure, and governance frameworks.
Periodic reality checks are inevitable but that’s not the same as a technology winter.
The Infrastructure Overbuild: Dark Fibber and Redundant Rails
The dot-com excess left behind “dark fibber” vast bandwidth capacity that sat unused until applications caught up years later. Railroads laid redundant, parallel lines that were eventually rationalized.
AI’s equivalent is the compute and energy build-out. Hyperscalers are spending at historic levels on:
- Data centers
- GPUs and specialized chips
- Advanced networking infrastructure
- Power generation and distribution
This investment represents the “railway tracks” of the AI era. A capital expenditure digestion phase is likely, where capacity may briefly outstrip demand before utilization rises—just as happened with railways and fibre optic networks.
After the Crash: Consolidation and Durable Winners
Dot-Com Survivors Became Tech Giants
After the dot-com crash, a handful of survivors, Amazon, Google, Microsoft, and Oracle grew into the backbone of the web economy. Many railroads failed, merged, or were restructured, yet the network that remained powered a century of U.S. economic growth.
Expect a similar AI trajectory:
- Consolidation among model start-ups and thin-moat applications
- Durable platform infrastructure providers will compound
- Foundation models with defensible moats will dominate
- Integrators with strong distribution channels will thrive
When Does Value Match Valuation?
A common misconception in every tech boom is that utility must match valuation immediately. Historical precedent suggests otherwise.
The internet’s productivity dividend became clear only after broadband, digital payments, logistics networks, and SaaS business models matured. Railroads created outsized economic gains through standardization: standard gauge tracks, coordinated timetables, and interchange agreements unlocked powerful network effects.
AI’s larger payoff will follow similar patterns:
- Standardized data pipelines
- Governance guardrails and compliance frameworks
- Robust change management processes
- Interoperability standards (the standard gauge of AI)
Where Will Value Accrue?
As the dot-com hardware stack commoditized, pricing power shifted to platforms and software ecosystems. Railroads saw freight rates compress as competition intensified and weaker lines failed.
AI is poised for a comparable transition:
- Compute scarcity will ease with increased chip supply, better schedulers, open-weight models, and inference optimization
- Margin pressure will move upstream in the value chain
- Value will accrue to companies turning usage into dependable revenue and measurable outcomes
- Distribution, data advantages, and workflow embedding matter more than novelty
Regulation and Consolidation: The Rules That Shape Winners
The web economy only stabilized after clearer regulations around privacy, payments, and competition created predictable playing fields. Railroads eventually faced the Interstate Commerce Commission and state regulations that constrained abuses while accelerating consolidation.
AI will follow the same path:
- Safety regimes and auditability requirements
- Data provenance and transparency standards
- Competitive oversight and antitrust considerations
- Higher fixed costs of participation favouring scaled, compliant operators
This regulatory framework will punish under-capitalized or non-differentiated entrants while rewarding those with robust compliance infrastructure.
Rolling Corrections vs. The Big Pop
The dot-com bust felt like a single dramatic collapse (2000–2002), though recovery was selective and prolonged. Railroads experienced waves of failures, mergers, and refinancing as market conditions evolved.
AI is likely to experience rolling corrections rather than a single catastrophic crash:
- Initial corrections in overheated niches (certain infrastructure suppliers, app layers with weak moats)
- Stabilization as real deployments compound
- Customer demand for tangibly better outcomes driving market discipline
- Gradual separation of hype from genuine value creation
AI Strategy for New Zealand Organizations
For Aotearoa New Zealand, these historical parallels offer particularly instructive lessons. Railroads and fibre optic networks were national enablers, but the winners paired infrastructure with relevant services, effective governance, and public trust.
AI success in New Zealand requires:
- Aligning deployments with local data realities and cultural expectations
- Focusing on sector-specific outcomes that convert global infrastructure into domestic productivity
- Robust change management processes
- Data readiness and governance frameworks
- Moving beyond hype to measurable results
Organizations chasing novelty without proper change management or data readiness will discover the harsh limits of hype-driven strategies.
The 2025 AI Market
AI in late 2025 resembles late-1999 dot-com meets 1870s railroads: speculative capital building something genuinely foundational.
The right historical lesson isn’t bubble, then bust, then nothing. Instead, expect:
- Overbuild – Excess capacity in compute, models, and infrastructure
- Correction – Market reality checks and valuation adjustments
- Consolidation – Survival of the strategically positioned and well-capitalized
- Durable value creation – Long-term productivity gains and economic transformation
Networks, standards, and complementary capabilities will eventually catch up. Expect froth to be skimmed and market players to change. However, the “tracks” now being laid—compute infrastructure, foundation models, data pipelines, and governance frameworks—are likely to carry the next decade of software and services innovation.
Building for the Post-Correction World
The strategic imperative now is to build for the post-digestion world:
- Standardise the gauge – Establish interoperable systems and data standards
- Maximize asset utilisation – Ensure infrastructure investments deliver measurable returns
- Focus on real, repeatable outcomes – Move beyond proof-of-concept to production value
- Invest in change management – Technology alone doesn’t drive transformation
- Build defensible moats – Distribution, data advantages, and workflow integration create lasting value
Key Takeaways: AI Bubble or AI Evolution?
- AI market volatility is normal for transformative technologies, not a sign of failure
- Historical parallels with dot-com and railroads suggest rolling corrections, not catastrophic collapse
- Infrastructure “overbuild” today becomes the foundation for tomorrow’s productivity gains
- Consolidation will favour platforms, defensible models, and strong distribution channels
- Real value creation requires standardization, governance, and change management—not just technology
- New Zealand organizations must align AI deployments with local realities and measurable outcomes
The AI revolution is real. The bubble concerns are valid. But history suggests the outcome won’t be bust it will be transformation, selective success, and the foundation for a decade of innovation.
