The AI Bubble: Markets Are Pricing Dreams, Not Deliverables

Artificial intelligence has become the markets’ favourite story. Capital is chasing anything remotely branded “AI,” and the valuations reflect it. The problem is simple: the narrative has run far ahead of execution.

A Rally Fueled by Concentration

A handful of AI-linked stocks have been doing the heavy lifting for global equity indices this year. The data below shows how skewed market performance has become:

Name	Market Cap (as on 17th Nov 2025)	1 Year Returns	Price to Earnings	Weight in S&P 500
Nvidia	4.63 Trillion	35.7%	54.03	7.63%
Microsoft	3.79 Trillion	22.7%	36.23	6.25%
Alphabet Inc.	3.36 Trillion	56.18%	27.31	5.5%
Meta Platforms	1.53 Trillion	5.6%	26.96	2.53%
Oracle Corporation	635 Billion	19.14%	51.59	1.05%

Despite the S&P 500 containing 500 companies, nearly 23% of the index is now concentrated in just five tech names. Nvidia alone carries a larger weight than the entire energy sector. Alphabet’s rally in 2025 has been so sharp that its single-year returns exceed the combined returns of several major sectors.

This level of concentration is not normal market behaviour—it’s a bet. Investors are pricing AI-driven earnings far ahead of actual monetisation. When the top 1% of the index accounts for a disproportionate share of total gains, it’s not broad-based strength; it’s a narrative trade.

The Cost Problem Nobody Talks About

Developing cutting-edge AI models is incredibly expensive, with training costs reaching tens of millions of dollars; for instance, a model like GPT-4 is estimated to have cost over US$100 million. Since 2016, the cost of compute has been rising dramatically at a rate of approximately 2.4 times annually. While cloud GPU rates in 2025 range from US$2–15 per hour, this only accounts for part of the overall expenditure. 

Organizations must also budget for data preparation, storage, staff, compliance, and ongoing model maintenance. The financial outlook is further strained by the fact that most AI startups rent their infrastructure instead of owning it, which significantly squeezes their profit margins.

Adoption Lags the Noise

Corporates are experimenting aggressively with AI but deploying slowly. Across industries, thousands of proof-of-concept pilots are running, but few have crossed into full-scale rollout. The reasons are predictable:

• data quality problems

• security concerns

• regulation

• unclear return on investment

• integration difficulties with legacy systems

Enterprise tech cycles move in multi-year phases, not months. The market is pricing AI monetisation as if the entire Fortune 500 is ready to integrate AI across every workflow tomorrow. The reality is far more measured.

The Business Model Issue

A bubble doesn’t require bad technology. It only needs expectations to outrun fundamentals—exactly what’s happening now.

Stretched P/E multiples, mega-cap dominance, and speculative inflows into any stock even remotely attached to AI all point to the same issue: the market is extrapolating long-term potential into short-term valuations.

The most exposed areas are:

• startups dependent on subsidised cloud credits

• companies with no proprietary data advantage

• public names with valuations priced for perfection

• AI tools that can be replaced easily or replicated quickly

When liquidity tightens or earnings disappoint, these names correct first and hardest.

Where the Real Value Will Stay

The correction, when it comes, won’t kill AI—it will kill the noise around it. The companies that will hold their ground are the ones that treat AI as infrastructure, not branding.

The durable winners are likely to be:

• firms with proprietary datasets competitors cannot access

• chipmakers and cloud providers that power the entire ecosystem

• tech majors embedding AI deeply into profitable existing products

• enterprise players solving real workflow problems with measurable ROI

AI will create winners. It just won’t be everyone.

The Takeaway

Every transformational technology—computing, internet, smartphones—has gone through a bubble phase. AI is no different. The technology is genuine and powerful. But the valuations assigned to AI-linked companies have outpaced current monetisation and realistic adoption timelines.

Markets are betting on future earnings long before they materialise. When expectations correct to more reasonable levels, only companies with real defensibility will remain at the top. The hype will deflate, but the technology will continue its steady, structural climb.