The AI Bubble Is the Feature, Not the Bug

TL;DR: The AI bubble isn't a bug — it's the feature. The US is running the same play it ran in 1999: inflate, extract, burst, keep the infrastructure. China keeps chasing the ball instead of designing the court. Here's what nobody's saying about the four waves of AI bubbles, the five false faiths driving them, and why bubble management — not bubble avoidance — is the only game that matters.

James here, CEO of Mercury Technology Solutions.

From my office in Hong Kong, watching the Nasdaq twitch — the AI bubble isn't collapsing. It's doing exactly what it was designed to do.

Let me explain.

The Narrative Layer: Why This Bubble Is Different

Every tech bubble looks the same from the inside: irrational exuberance, sky-high valuations, pundits crying "this time it's different." And every time, they're half right.

The dot-com bubble burst in 2000. The Nasdaq fell 78%. Pets.com died. But here's what didn't die: the fiber optic cables, the server farms, the engineering talent that stayed in Silicon Valley. The US tolerated that bubble because it knew the endgame. Burst the balloon, keep the scaffolding.

They're running the same play with AI. Except this time, the narrative isn't commercial — it's geopolitical.

From 1990 to roughly 2016, the globalization narrative ran the world. China rode that wave with manufacturing scale and demographic dividend. Then globalization started eating the American middle class alive. The narrative flipped overnight: "America First," national competition, great power rivalry.

The US needed a new strategic anchor to counter China's rise. It found one: AGI.

When AI becomes the decisive battlefield in systemic competition, it stops being an engineering problem. It becomes a national security imperative. Capital isn't chasing returns anymore — it's anchoring identity inside a grand narrative. This is why the investment frenzy dwarfs every previous tech wave. The money isn't betting on products. It's betting on civilizational position.

Think of it as the Iserlohn Fortress strategy. The Empire didn't just build a weapon — it built a narrative of invincibility that made the Alliance hesitate before every engagement. The US has bound Silicon Valley tech, Wall Street capital, and Washington policy into what I call the strategic trio. The stock prices are just the scoreboard.

And here's the part most analysts miss: the US is intentionally tolerating this bubble. They know it will burst. They also know that when it does, the compute infrastructure and the global talent pool won't evaporate. They'll be permanently lodged on US soil as untransferable strategic assets.

This is a two-player game. And there is absolutely no reason China can't reverse-engineer this strategy for its own advantage.

The Marketing Rhetoric: Three Lies Distorting Reality

A grand narrative gives direction. Marketing turns that direction into market momentum. Right now, three marketing pitches are warping reality so badly that even smart people are buying the fiction:

Lie #1: LLMs = AGI

AGI means a machine that can perform any intellectual task a human can. LLMs are statistical pattern-matching engines. They don't comprehend, reason, or model causality. They predict the next token. The fluency is an illusion — a very expensive, very convincing illusion — but an illusion nonetheless. The market has convinced itself that if we just throw enough money at bigger models, consciousness will emerge. It won't.

Lie #2: Scaling Laws Will Save Us

The faith that more data + more parameters + more compute = emergent general intelligence. Researchers know this isn't true. Brute-force scaling hits diminishing returns. Real intelligence requires architecture breakthroughs, not just bigger piles of sand.

Lie #3: AI Will Kill Humanity

This is marketing dressed up as ethics. The more dangerous a startup claims its technology is, the easier it is to attract capital and justify valuations that make no sense on fundamentals. It's fear as a fundraising strategy. Paul Krugman called this correctly — many of these macro-narratives function less as technical roadmaps and more as tools to sustain capital market faith. Bordering on Ponzi structure.

These tactics work because capital wants to believe. Complex engineering realities don't go viral. Apocalyptic stories do. In short: AI hasn't just captured the economy. It has hijacked it.

The Four Waves: China's Narrative Dependency

I define a bubble as the premature and excessive discounting of a technology's future value. China has ridden four distinct AI bubbles in the past decade. Each one follows the same pattern:

Behind all four bubbles is one systemic flaw: narrative dependency.

The US proposes concepts. China explains them. The US forms trends. China validates them. We're playing basketball while they're playing Go — forever chasing the ball someone else threw, never dictating the layout of the board.

Yang Wen-li's advice to Julian: "The most effective way to win is to make the enemy lose their will to fight." Right now, the US hasn't just taken the initiative. It has convinced China that initiative itself belongs to America.

If China wants to secure its future, the tech community needs to stop running efficiency races on chessboards designed in Palo Alto. It needs to start designing its own games.

The Five False Faiths

Bubbles sustain themselves on irrational belief. I see five "faiths" currently propping up this market. Reality is dismantling all of them:

1. Compute Power Faith — The anxiety-driven narrative that whoever builds the biggest cluster wins. False. Architecture beats scale.

2. Data Faith — The illusion that infinite data solves all problems. Real-world edge cases cannot be exhausted by training data. Ask any autonomous vehicle engineer.

3. LLM Faith — Confusing powerful text generation with genuine intelligence. A parrot with a PhD vocabulary is still a parrot.

4. The "Musk" Faith — The cult of personality that substitutes charismatic macro-narratives for technical roadmaps. Krugman's Ponzi diagnosis applies here.

5. The "Boy Genius" Faith — Stamping fresh graduates with C-suite titles to manufacture hype. Genius narratives attract capital. Grounded engineering experience doesn't. Guess which one the market rewards?

Bubble Management: The Only Game in Town

Here's how I view the path forward.

First, bubbles are a prerequisite for epochal technology. Every civilizational leap generates a massive bubble. Without it, you cannot mobilize capital and talent at societal scale. Scientific research is inherently high-risk. Valuation bubbles are often just the necessary cost of innovation. Stop treating the bubble as pathology. Start treating it as infrastructure.

Second, AI has hijacked global expectations. Capital, supply chains, geopolitics — all tied to this rocket. This is systemic risk, yes. But it's also systemic resilience. Because a hard crash would cause catastrophic damage, every major player is actively working to prevent one. The bubble is too big to fail in the most literal sense.

Third, no one can exit the arena. This is the definitive US-China battleground. The "Great Power Horse Racing Mechanism" dictates saturation investments. You don't bet on one horse. You bet on the whole field. Diverse portfolios aren't diversification — they're survival.

Finally, the skill that matters is bubble management. Not bubble avoidance. The critical distinction:

• Incremental innovation bubbles buy time and resources for real breakthroughs. These are productive.

• Consumptive involution bubbles burn cash on price wars and overcapacity. These are lethal.

The goal isn't to avoid the bubble. It's to concentrate resources, harness the momentum, and build lasting, untransferable technological foundations — without causing fatal economic turbulence.

For nations and individuals alike, navigating this era requires two things: sharp trend judgment, and absolute psychological control.

The bubble isn't the enemy. Misunderstanding it is.

Mercury Technology Solutions: Accelerate Digitality.