Category Archives: #AI

AI is a bubble that is not going to burst (but it’s still a bubble)

Posted on by
Reply

AI: A Bubble, But Not a Bubble (written with help from ChatGPT-5)

Or: Why the “bubble” narrative around AI misses something deeper

The question keeps coming up: is artificial intelligence (AI) currently in a speculative bubble that’s destined to burst, or is this something more enduring — a transformational wave disguised in bubble clothing?

Let’s unpack the paradox: on one hand, many of the classic hallmarks of a bubble are present. On the other, there are structural, strategic and systemic factors that suggest this may be a bubble that isn’t just a bubble. Below is a needle-scan breakdown of key indicators.

1. Bubble-Warning Signals

These are the red flags: the parts of the story that say: yes, this does look like a bubble.

  • Hype vs returns: Global investment into AI startups exceeded $50 billion in 2023, yet very few of these companies are profitable or generating recurring revenue (BuiltIn).
  • Valuations stretching history: Nvidia’s market cap crossed $2 trillion in 2024 — the fastest growth in tech history — stoking concerns that AI-driven valuation multiples are disconnected from current fundamentals.
  • Warnings from the top: OpenAI CEO Sam Altman himself said in 2024, “Yes, it’s a bubble… and that’s OK.”
  • Concentration & fragility: As of mid-2025, the top five AI companies (Nvidia, Microsoft, Alphabet, Meta, and Amazon) control over 85% of the global AI compute infrastructure.
  • Speculative patterning: Startups with no product and mere “AI wrappers” around ChatGPT are raising millions in pre-seed funding, echoing dot-com era exuberance.
  • “AI to the rescue? Not so fast. SVB cleverly asks in the title of this third chart, “Chat, What’s Another Word for Bubble?”

2. But It’s Not Just a Bubble

Now the other side: what suggests AI is more than froth and fear.

  • Infrastructure build-out: $200B+ projected spending on AI data centers between 2024-2027 (McKinsey). These aren’t ephemeral assets; they’re physical, long-term capital investments.
  • Government policy shifts: The EU, U.S., China, and UAE have all declared national AI strategies. The UK launched “Frontier AI Taskforce” with a £100M fund. These are state-level stakes.
  • Societal adoption: ChatGPT reached 100 million users in two months — the fastest adoption of any consumer app in history. It’s now integrated into Office365, Shopify, Duolingo, and dozens of platforms.
  • Cross-system integration: AI is now used in logistics, drug discovery, customer service, legal contracts, climate modeling, and more. It’s not one vertical; it’s multi-sectoral.
  • Decentralised movement: Projects like SingularityNET, Bittensor, and Fetch.ai aim to provide counterweights to centralized AI monopolies. Though small in market cap, their ideology is sticky and increasingly resonant.
  • Talent pipeline: Top universities report record-breaking enrolment in machine learning & data science tracks. MIT saw a 73% increase in AI-related thesis topics from 2021 to 2024.
  • Powell says that, unlike the dotcom boom, AI spending isn’t a bubble: ‘I won’t go into particular names, but they actually have earnings’.

3. Why This Matters

Because how one interprets this moment drives strategy.

  • Treating AI as just a bubble? You risk ignoring long-term infrastructure and missing the strategic layer.
  • Treating AI as only hype-free? You risk capital misallocation and being blindsided by volatility.

Instead, the correct lens may be dual-layered: short-term froth, long-term wave.

4. Key Signals That It’s Not Just a Bubble

Here are seven indicators that suggest AI is here for the long haul:

  1. $200B+ in AI infra spend (2024-2027) — Source: McKinsey
  2. 40+ nations with national AI plans — Source: OECD AI Policy Observatory
  3. 100M+ users for ChatGPT within 60 days of launch
  4. AI cited in 60%+ of S&P 500 earnings calls in 2024 (Goldman Sachs)
  5. 5,000+ AI-related job listings on LinkedIn UK in July 2025 alone
  6. AI + Crypto projects growing: Over $4.3B market cap in AI-token sector (CoinGecko, Q2 2025)
  7. Cross-sector resilience: AI use cases now span healthcare, finance, media, law, education, and urban planning

5. Strategy for Navigators

If you’re an investor, policymaker, or DeSci founder:

  • Use caution: Recognize speculative behaviour where it exists.
  • Track fundamentals: Focus on infrastructure, partnerships, developer traction.
  • Scan for decentralisation: Keep eyes on AI x Web3 convergence.
  • Measure what matters: User adoption, SDK integrations, compute dependencies, data partnerships.
  • Diversify bets: Don’t just follow LLMs and chips — track edge AI, tokenised compute, AI governance tools.

6. Final Thought

Yes — there is a bubble vibe to AI right now. The hype is real, the valuations are stretched, and not all will survive.

But it’s not just a bubble.

It’s a complex, layered, evolving ecosystem with speculative peaks but deep structural roots. Infrastructure, strategy, adoption and decentralisation all suggest this is not a passing moment. The wave has momentum.

“It may wobble, it may correct, it may reshape — but the foundations are being laid for a long-term wave, not just a feast followed by famine.”

Further Reading:

Why 4 Minutes Matter: The Hidden Cost of Imprecise Data in AI

Featured

Posted on by
Reply

Most of us grow up believing that a day is exactly 24 hours long. It’s tidy, convenient, and feels close enough to reality. But strictly speaking, the Earth completes one rotation on its axis in 23 hours and 56 minutes — what astronomers call a sidereal day. The extra four minutes come from the Earth’s simultaneous orbit around the Sun. If we ignored this subtlety, our sense of time would slowly drift out of sync with the Sun itself. Noon would stop being “midday.”

Those four minutes are a small detail — but they matter.

The Data Analogy

This is exactly what happens when organisations feed “close enough” data into AI systems. At first, the model might seem fine. Predictions look reasonable. The dashboards tick over. But just like those four missing minutes, tiny inaccuracies and fuzzy definitions build up. Over weeks, months, or years, the system drifts further from reality.

Suddenly, your AI isn’t aligned with the world as it actually is. Recommendations miss the mark. Bias creeps in. Customers lose trust.

The lesson? Precision in data is not pedantry. It’s the difference between alignment and drift.

Why Precision Matters

  • Compounding effect: Small errors accumulate over time. Like four minutes a day becoming hours, days, and months of misalignment.
  • AI is literal: Models take inputs as ground truth. A vague definition or inconsistent label isn’t “good enough.” It’s an anchor point for bad predictions.
  • Trust is fragile: Once stakeholders see AI outputs wobble, confidence in the entire system erodes.

The Needle Framework: Finding the Signal

Getting data right is about finding the needle in the haystack: the clear, sharp definition hidden among the fuzz. When you sharpen the data — consistent labels, correct units, precise categories — you give AI a sidereal day to lock onto. A stable reference point. A system that stays in sync instead of drifting.

So What?

AI isn’t magic; it’s alignment. And alignment starts with data. Just as astronomers can’t afford to ignore the missing four minutes, companies can’t afford to wave away small inconsistencies. The cost of “close enough” is hidden drift.

The sharper your data, the sharper your AI. And that’s where the real value emerges.

Four minutes matter in astronomy. And they matter in AI. Get your data precise, and your systems won’t just work today — they’ll stay aligned tomorrow.