The AI Bubble (Or Not)
- Eniokos
- 5 days ago
- 10 min read
The AI bubble refers to concerns that AI companies are overvalued and overhyped.
Investors have poured (and are continuing to pour) billions into AI startups and tech giants, betting on massive future profits. Stock prices have soared based on expectations rather than current earnings.
Many economists and AI critics worry this mirrors past tech bubbles, like the dot-com crash, where excitement exceeded reality. Companies may not deliver the revolutionary changes promised, or profits may take much longer to materialize than expected.
Should we care?
If the bubble bursts, it could trigger stock market losses, job cuts, and reduced funding for legitimate AI innovation. Your market investments or job could be affected. Remember 2008?
I thought it would be interesting to use AI to further understand this situation. Here is what ChatGPT told me:
How to Tell if an AI Boom Is a Bubble
Below are practical signals economists and investors use to test whether a boom is a bubble.
1. Revenue vs Valuation Gap
A classic red flag:Companies with tiny or no revenues get massive valuations.
Example pattern (hypothetical):
A start-up earning £2 million a year is valued at £3 billion→ This mismatch suggests hype-led pricing.
Today, some AI firms show this pattern, but not all. Large companies like NVIDIA, Microsoft, and Amazon have real revenue matching their valuations, so they are not bubble-like.
2. High burn rate, low profitability
Many AI companies spend:
billions on hardware
billions on talent
billions on model training but have uncertain paths to profit.
If companies must keep raising money just to survive, it signals bubble-like fragility.
3. Too many start-ups doing the same thing
When investors fear missing out (FOMO), they fund:
identical chatbots
identical AI tools
identical model labs
This mirrors the dot-com era, where 100 companies built online bookstores.
Oversupply often precedes collapse.
4. Unclear business models
A sector becomes fragile if the answer to “How will this make money?” is vague or delayed.
Today, many AI companies rely on:
selling compute
selling API calls
enterprise subscriptions
But long-term profitability is still being tested.
5. Investment driven by fear of missing out
If money flows into a sector because “everyone else is investing,” it is usually a bubble.
VC behaviour in 2023–2025 has some of this. Rapid commitments with minimal due diligence are a warning sign.
6. Hardware constraints
AI development depends on:
GPUs
data centres
electricity
cooling
These costs keep rising. If infrastructure becomes too expensive, investment slows and a correction follows.
When Does a Bubble Usually Pop?
No one can predict exact timing, but bubbles tend to pop when three conditions align:
1. Capital dries up
When investors stop funding loss-making companies, the weakest firms collapse. AI start-ups with no revenue would fall first.
2. Regulation arrives
If governments impose:
safety rules
compute taxes
data restrictions
it can slow down growth and trigger corrections.
3. Failed expectations
If AI cannot deliver:
true autonomy
reliable reasoning
cheap running costs
as fast as promised, money exits and valuations drop.
Historically, bubbles pop 1–3 years after peak hype if fundamentals are weak.
How Sustainable Is Current AI Investment?
The answer is mixed — some parts are solid, others are fragile.
Sustainable components
These areas have long-term real utility:
1. Infrastructure
chips
data centres
cloud services
These earn revenue even if software start-ups fail.
2. Enterprise AI adoption
Banks, hospitals, logistics companies, and tech firms are integrating AI steadily.
This creates stable demand.
3. AI in productivity software
Products like Microsoft Office integrated with AI have millions of paying users. This is real income, not speculative.
4. Robotics + automation
Industrial robots, warehouse automation, and autonomous inspection systems are growing reliably.
Fragile or Bubble-like components
1. Foundation model companies (small and medium ones)
Many are burning money faster than they earn it.
Without sustainable business models, they may not survive the next 2–3 years.
2. AI start-ups promising unrealistic capabilities
If a company promises:
AGI timelines
self-driving Level 5 within months
robot butlers
instant productivity revolutions
these become high-risk investments.
3. Consumer AI tools with no moat
Tools that can be cloned easily may not last. Competition is too high.
4. Excessive valuations in private markets
When investors value a company at £50 billion without revenue to justify it, a correction is likely.
So, Is It a Bubble?
Partially yes, partially no.
The infrastructure and enterprise adoption side is strong and sustainable.
The speculative start-up and AGI hype side is bubble-like and may collapse sharply.
A more accurate statement is:The AI sector has a bubble inside a real technological revolution.
Likely Outcome (flagged: partly speculative)
Weak start-ups will fail first.
Foundation model competition may consolidate to a few giant players.
Infrastructure providers (NVIDIA, AMD, TSMC, cloud giants) will remain strong.
AI will become a normal layer of software, like cloud computing or mobile apps.
Prices for compute may fall once competition increases or regulation adjusts.
The hype will collapse, but the technology will remain and grow.
Comparision of AI boom vs dot-com bubble
Big similarities
You can think of both as tech waves with a speculative layer on top of a real shift.
1. Strong narrative and fear of missing out
Dot-com: any site with a dot com in its name got funding.
AI: any pitch with generative AI or foundation models gets attention. Analyses of the current boom highlight stretched valuations, rapid venture funding and classic bubble signs such as overinvestment and over-ownership.
2. Concentration in a few star stocks
Dot-com era was heavily concentrated in a handful of tech names on Nasdaq.
Today, a small cluster of AI-heavy firms and chip makers carry much of the index return.
3. Massive infrastructure build-out
Dot-com funded fibre networks, data centres and hosting.
AI is driving huge spend on GPUs, data centres and power infrastructure, with central banks now flagging AI-linked valuations and investment as a financial stability risk.
4. Classic late-cycle signs
Commentators point out that AI now shows four common bubble traits: overinvestment, overvaluation, over-ownership and rising leverage in parts of the system.
So yes, the pattern rhymes with the late nineties.
Important differences
This is why many economists say it is not a one-to-one repeat.
1. More real revenue and profits this time
In 1999, only a small fraction of big tech listings were actually profitable; many were pre-revenue.
Today, a lot of the AI boom is anchored in already profitable giants (cloud providers, chip makers, software suites). A recent US Federal Reserve speech explicitly argued that AI-linked stock gains are less likely to be a repeat of the dot-com crash because today’s firms have stronger earnings and less reliance on debt.
2. Actual usage is already mainstream
Internet in the nineties had huge promise but limited daily use outside early adopters.
AI tools are already embedded in office software, customer support, coding tools, design tools and search. Studies and industry reports show widespread adoption across sectors such as healthcare, finance, manufacturing and education.
3. Different stress factors
The AI era faces constraints that did not exist in the dot-com era:
hardware bottlenecks in high-end chips
export controls and geopolitical tension around semiconductors
very high compute and energy costs slowing commercialisation of many ideas
4. Less obvious single trigger
The dot-com crash coincided with:
Y2K spending fading
rate rises
realisation that many business models were not viable
Analysts note that the Y2K boost was a one-off that magnified overcapacity, something that does not have a direct parallel in AI.
5. Science and innovation patterns look different
One recent academic study comparing the two eras finds that research and patenting patterns in AI do not line up neatly with dot-com signals, so the same early warning indicators may not apply in a simple way.
Aspect | Dot-com era (late 1990s) | Current AI era |
Main story | Internet will change everything | AI will change everything |
Typical company | Young, unprofitable web start-up | Mix of mega-caps plus start-ups |
Adoption at the time | Early but patchy | Already broad across industries |
Infrastructure | Fibre and basic data centres | GPUs, advanced data centres, power |
Profitability | Low at the index level | Much higher for leading firms |
Systemic leverage | High in some equity and margin accounts | Lower so far, but rising in credit |
Policy maker view | Saw it mostly in hindsight | Actively watching AI risks now |
Will the AI bubble pop, and when?
Honest answer: timing is unknowable.
We can say:
There are clear bubble-like pockets in speculative AI start-ups and very high multiple stocks.
Central banks see elevated valuations and heavy AI infrastructure investment as risks, but not yet a systemic crisis trigger on their own.
What is most likely (speculative, flagged):
Some overvalued firms and copy-paste start-ups will be wiped out in the next downturn.
Core infrastructure and genuinely useful AI services will remain and keep growing, just as Amazon and a few others emerged stronger after the dot-com crash.
So we are probably in:Real long-term shift plus a frothy layer that will get corrected.
Sectors likely to benefit most from AI over the next decade
1. Healthcare and life sciences
Why it is strong:
Diagnostics support from imaging and pattern recognition.
Predictive models for risk scoring and triage.
Drug discovery and molecule design, where AI can cut discovery timelines and costs significantly.
What survives a correction
Tools that make clinicians faster or more accurate.
Software that automates routine admin and documentation.
Drug discovery platforms that actually lead to new, approved molecules.
2. Banking, financial services and insurance (BFSI)
Why it is strong:
AI is already standard for fraud detection, credit scoring, anti-money-laundering checks and algorithmic trading.
Indian and global banks are steadily increasing AI budgets because the cost savings and risk reduction are clear.
What survives
Risk and fraud analytics.
Personalised financial advice and customer service.
Better underwriting in insurance.
3. Manufacturing and industrial automation
Why it is strong:
Predictive maintenance and fault detection.
Quality control via computer vision.
Process optimisation and robotics.
What survives
Anything that saves downtime, reduces scrap or energy use. These are direct, measurable returns, so adoption continues even if investors fall out of love with AI buzzwords.
4. Transport, logistics and supply chain
Why it is strong:
Route optimisation and dynamic pricing.
Warehouse automation and demand forecasting.
What survives
Fleet management and route planning that cut fuel and time.
Supply chain risk tools that help firms keep inventory lean without being fragile.
5. Retail, e-commerce and marketing
Why it is strong:
Recommendation engines, personalised offers and chat-based support.
Better targeting of ads and content, with lower campaign costs.
What survives
Systems that materially raise conversion rates or basket size.
Tools that let small businesses run campaigns that previously required a full agency.
6. Education and training
Why it is strong:
Personalised practice, tutoring and language learning.
Automated marking and feedback for routine work, freeing teacher time.
What survives
Tools that actually improve learning outcomes or teacher workload, not just shiny apps.
Niche verticals such as corporate training, coding upskilling and continuous professional education.
7. Agriculture and food systems
Why it matters, especially in countries such as India:
Crop yield prediction, pest and disease detection from satellite and drone imagery.
Smarter irrigation and fertiliser use.
What survives
Services tied to farmer income, yield stability and input cost savings.
Decision support for co-operatives, agri-fintech and supply chains.
8. IT services, cloud and developer tooling
Why it is strong:
AI is becoming part of every software stack: code assistants, automated testing, observability, security.
For India, IT services firms that help global clients adopt AI are already positioning themselves here.
What survives
Platforms that make developers faster and reduce bugs.
Consulting and integration work to bring AI into legacy systems.
9. Public sector, security and smart cities
Why it matters:
AI for traffic management, citizen services, fraud detection in welfare schemes, and cyber-security.
What survives
Systems that help governments do more with limited staff and budgets, while passing basic transparency and fairness checks.
10. Energy and utilities
Why it is emerging:
Grid balancing for renewables.
Demand forecasting and predictive maintenance on critical assets.
What survives
Anything that stabilises grids, reduces outages and avoids costly over-capacity.
Why OpenAI may be over-promising
Gigantic spending vs uncertain returns
Recent reports suggest OpenAI is burning through billions annually to fund model-training, infrastructure and operational costs. (The Guardian)
Some analyses (e.g., from HSBC) argue that projected revenue is far smaller than the capital it needs to scale — implying a large funding gap. (Fortune)
In effect they are betting on a future where demand for compute and AI services remains huge, indefinitely. If that doesn’t materialise, costs will overwhelm returns. (The Guardian)
Circular deals and over-dependence on hype
OpenAI’s deals with chip-makers (like Nvidia and sometimes AMD) have drawn scrutiny: critics call them “vendor financing” — a red flag historically associated with bubbles. (The Guardian)
The business model rests heavily on continued investor enthusiasm, rising infrastructure investment, and growing enterprise/consumer adoption. If sentiment changes, or competition intensifies, the fragility becomes exposed. (SiliconSnark)
Unclear path to profitability and ROI for many use-cases
Many advanced AI systems remain expensive to run (compute, electricity, maintenance) compared to the revenue they bring. (GARP)
Critics argue AI hype often overpromises on what generative models can deliver: realistic improvements in productivity, trust, safety, or business value remain uneven. (ResearchGate)
There is a risk that some flagship promises — scientific breakthroughs, universal AI assistants, global digital transformation — may not happen soon enough to justify the capital invested.
Reputation, governance and ethical risks
As a formerly “open” research organisation turned profit-seeking firm, some critique that OpenAI may prioritise scaling and revenue over safety and transparency. (The Times of India)
If AI misuse or safety incidents grow (or public sentiment turns against powerful AI), regulatory pressure or backlash could hit valuation and growth hard. This remains a non-trivial danger. (ResearchGate)
Why OpenAI might still make it
Real demand for AI services and infrastructure
Many companies and industries are already paying for AI services — enterprise clients, subscription users, cloud-AI integration. That gives OpenAI some real cash flow, not just speculative promise.
As long as large-scale compute, cloud services and scalable infrastructures remain central to digital transformation, the demand for what they build remains real. (GARP)
First-mover advantage and brand strength
OpenAI has become a reference name in AI. That market leadership makes it easier to attract investment, partnerships, and customers. If they manage burn and execution well, they are positioned to survive better than smaller rivals.
Long-term upside if core breakthroughs pay off
If OpenAI (or its successors) deliver dramatic productivity gains — in areas like automation, enterprise AI, scientific research, or general-purpose tools — then the current massive investment might look justified.
Possibility of business model pivot or diversification
Even critics note that some of OpenAI’s plans (e.g. data-centres, enterprise AI services) could — in theory — turn profitable, especially as hardware costs drop or as enterprise demand broadens. (The Guardian)
References [as generated by ChatGPT, unverified]
Sources and URLs used
Bank of England financial stability comments on AI valuations and investment risks
Wall Street Journal on AI exposure in bond markets
Business Insider interview with Ruchir Sharma on AI bubble signs
Bloomberg newsletter on AI bubble vs dot-com fever
Reuters report on US Federal Reserve view that AI gains are unlikely to fully repeat dot-com
Data-driven and qualitative comparisons of dot-com vs AI eras
Sectoral AI impact and use-case overviews
https://www.sciencedirect.com/science/article/pii/S2773207X24001386
https://codewave.com/insights/top-ai-applications-major-industries
https://www.thoughtful.ai/blog/examples-of-artificial-intelligence-ai-in-7-industries
https://www.algoworks.com/blog/ai-impact-on-healthcare-retail-finance-manufacturing-marketing
https://www.howdy.com/blog/14-industries-ai-will-impact-in-the-next-10-years
https://asianbusinessreview.com/information-technology/in-focus/which-sectors-will-benefit-most-ai
