Battle of the Bubbles

Every few decades, we meet a moment where optimism outpaces reality, and right now, AI might be standing on that edge. When the IMF, Bank of England, and leading fund managers begin warning of an “AI bubble,” it’s a sign the conversation has shifted from speculation to sober reflection.

Across 2025, the numbers have become almost surreal:

• Nvidia’s valuation hit an unprecedented $4.5 trillion in October, briefly making it the world’s most valuable public company.

• OpenAI, still private and unprofitable, was valued at $500 billion following a $6.6 billion share sale led by SoftBank.

• AI startups are absorbing 58% of global venture investment, more than $70 billion in just the first quarter of 2025.​

  
  

These valuations reflect investor confidence that artificial intelligence will define the next century. But history tells us every revolutionary technology has its overheated phase. The dot-com bubble of 2000 and even the housing boom in 2008 were also rooted in genuine innovation before excess took over. The market today mirrors those dynamics: genuine breakthroughs tangled up with speculative fever.

According to BofA’s October Fund Manager Survey, 54% of global investors already believe AI stocks are in a bubble. The IMF and Bank of England have echoed these warnings, with both institutions cautioning about a “sharp market correction” if demand falters or if spending continues outpacing real productivity gains.

Sam Altman, OpenAI’s CEO, seems to agree, albeit with nuance. In August, he told The Verge: “When bubbles happen, smart people get overexcited about a kernel of truth... Are we overexcited about AI? Probably. But the long-term impact will still be extraordinary.”

For industry leaders, much of the AI spending spree is flowing into foundational infrastructure — chips, data centers, and training compute. Nvidia alone commands over 70% of the global AI GPU market, and its chips power systems from OpenAI’s GPT-5 to Meta’s LLaMA-3 and Google’s Gemini. Microsoft and Amazon are also racing to double data center capacity by 2027, driving global AI spending to an estimated $1.5 trillion by year-end, according to Gartner.​

Another revealing signal comes from the private market. In venture capital, round sizes and valuations are rising faster than adoption curves. Bloomberg reports that more than 1,300 AI startups now hold valuations exceeding $100 million, including nearly 500 unicorns. Companies like Anthropic, Mistral, and xAI have all secured multi-billion-dollar rounds despite producing minimal revenue at present.

The upside, of course, is undeniable. Goldman Sachs estimates AI could add $7 trillion to global GDP by 2030, largely through automation, design optimization, and new product categories. This explains the frenzy; investors aren’t irrational, but they may be impatient. The breakthroughs in generative design, agentic AI, and multimodal intelligence promise structural shifts; the question is whether near-term returns can keep pace with those expectations.​

The immediate risk is not a total collapse but a capital misallocation crisis. Gartner calls this “the mispricing of innovation” — when money flows into buzzword-heavy projects rather than enduring R&D. Many early-stage AI firms funnel most of their capital into compute costs and branding rather than robust product-market fit.​

There’s also a sustainability angle.

According to Cisco’s 2025 AI report, every ChatGPT query already consumes 10× more power than a Google search, contributing to a looming 160% surge in global data center power demand by 2030. This raises uncomfortable questions: Can society sustain AI’s energy appetite? And who will bear the cost when scale meets scarcity?​

Moreover, geopolitical and regulatory dynamics could accelerate a correction. New U.S. and EU AI governance frameworks limit access to high-end GPUs and impose stricter reporting on model training data. For smaller players, compliance costs may become existential.

Personally, I don’t think the story ends with a crash, but a reset seems inevitable. Like the dot-com era, much will be destroyed, yet what remains could be stronger, leaner, and more focused. The AI leaders of the next decade won’t be those who shouted the loudest but those who translated intelligence into resilience, utility, and trust.

For founders and strategists, this is a time for discipline. The pressure to “AI-wash” products will be immense; the smarter move is to focus on creating defensible, outcome-driven systems — smaller, bespoke models that solve specific problems with measurable ROI.

The real test now is whether we can build AI that earns its valuations, not through hype cycles, but through lasting impact on how we live, work, and think.