The DeepSeek Shock: Where the Cracks First Appeared

The first real tremor hit in January 2025. A relatively unknown Chinese AI lab called DeepSeek released its R1 model — and claimed it could match the performance of leading Western AI systems at a fraction of the training cost. The market reaction was immediate and violent. Nvidia shares plunged 17% in a single session, erasing close to $600 billion in market capitalisation — the largest one-day loss for any single company in stock market history. Broadcom fell 17%. Oracle dropped 14%. The entire AI infrastructure thesis was suddenly in question.

The DeepSeek moment challenged two core assumptions the market had been pricing in. First, that US export controls on advanced chips gave American firms an unassailable lead in AI. Second, that building frontier AI models required billions of dollars in compute spending — which meant massive, ongoing demand for hardware. If a Chinese lab could deliver competitive results with older chips and a reported training budget of just $6 million, both of those assumptions looked fragile.

Many observers called it AI's "Sputnik moment." Whether that label sticks or not, it forced a global reassessment of who the winners and losers of the AI buildout would actually be — and whether the enormous capital expenditure commitments from Big Tech were truly as defensible as the market believed.

2026: The Year the Market Started Asking "Show Me the Money"

The DeepSeek shock of January 2025 ultimately proved temporary. Nvidia recovered. Broadcom and ASML finished 2025 significantly higher. AI spending didn't slow down — it actually accelerated through the year. Markets recovered their confidence, powered by genuine revenue growth from AI hyperscalers. Anthropic went from roughly $1 billion in revenue run rate to $9 billion. OpenAI expected $20 billion for the year.

But something fundamental shifted as 2026 began. The market moved from a phase you might call "AI euphoria" into what many are now calling "AI investigation." In 2024 and 2025, almost any mention of AI in a company's earnings call was enough to drive share prices higher. In 2026, the market is demanding proof. Can these companies actually turn massive AI investment into sustainable profit? That's the question now dominating investor psychology.

On February 13, 2026, the VIX — the market's "fear gauge" — surged nearly 18% in a single session, breaking through the psychologically significant 20-point level. The catalyst wasn't a single event but a convergence: growing "capex fatigue" among investors weary of the Big Tech "spend now, earn later" narrative, mounting concerns that AI may not deliver explosive returns on those multi-billion-dollar infrastructure commitments, and new regulatory headwinds from the EU AI Act starting to affect profit margins.

By mid-February, what started as sector rotation out of tech-heavy positions began to feel like something broader. Leveraged tech positions unwound rapidly, creating feedback loops that pushed volatility higher. The VIX calmed slightly to around 20.29 by February 18, but the message was clear: the "buy every dip" mentality that defined 2024–2025 was no longer the default setting.

A Viral Doomsday Report and an 800-Point Dow Drop

Then came the weekend that crystallised the fear.

On February 22, 2026, a small research firm called Citrini Research published a 7,000-word hypothetical scenario on its Substack. Titled "The 2028 Global Intelligence Crisis," it read like fiction set in the near future: unemployment above 10%, the S&P 500 down 38% from its highs, and an economy unravelling because AI had worked too well. In this scenario, AI agents replaced white-collar workers at scale — accountants, lawyers, marketers, software engineers — and the resulting collapse in consumer spending triggered a feedback loop the economy couldn't escape.

The report went viral. And when markets opened on Monday, February 24, the reaction was swift. The Dow dropped roughly 800 points. Software companies named in the report — Datadog, CrowdStrike, Zscaler — each fell more than 9%. IBM dropped 13%, its worst single-day performance since 2000. Financial firms including American Express, KKR, and Blackstone, also cited in the report, tumbled alongside them.

To be clear: the Citrini report is a thought experiment, not a forecast. Its own author, 33-year-old James van Geelen, said he never expected it to move markets. A top White House economist publicly dismissed it as "science fiction." Investors and economists from Citadel Securities, Deutsche Bank, Fidelity, and others have pushed back on the thesis. But the fact that a hypothetical scenario — published on a Substack — could trigger that kind of market response tells you something important about where investor sentiment sits right now.

Is This a Bubble? The Dot-Com Comparison

The word "bubble" is being used more frequently now than at any point since the AI rally began. The Case-Shiller price-to-earnings ratio for the US stock market has exceeded 40 for the first time since the dot-com era. Ray Dalio, co-chief investment officer at Bridgewater Associates, said in early 2025 that AI investment levels look "very similar" to the dot-com bubble. One researcher went further, calling the AI buildup "the biggest and most dangerous bubble the world has ever seen."

But there are important counterpoints. Major institutions like Goldman Sachs and JPMorgan argue this cycle is fundamentally different from the dot-com era. Today's market-leading AI companies are generating real revenue, have significant cash reserves, and maintain profit margins that the dot-com darlings never achieved. JPMorgan applied a five-factor diagnostic framework to the AI rally and concluded the sector shows structural utility rather than pure speculation.

The truth, as usual, probably sits somewhere in between. AI is real. The technology works. Revenue is growing. But the market is now reckoning with whether current valuations accurately reflect the pace at which returns will materialise — and that's a very different question from whether AI itself has value.

There's a useful historical lesson here. The internet absolutely transformed the global economy. But the market overshot massively on the way up, and the correction was painful. The S&P 500 eventually recovered and has returned over 2,500% since 1995. The internet didn't fail. But plenty of investors who were right about the technology still lost money because they were wrong about the timing and the valuation.

What This Means for Traders: 5 Things to Think About

1. AI fear is now a macro driver, not just a tech sector story. When AI panic pushes the VIX above 20, triggers rotation into defensive stocks, and starts affecting credit markets, it's no longer something you can ignore just because you don't trade tech. The spillover effects touch indices like the Dow (US30), safe havens like Gold, the US dollar, and even oil through risk appetite channels.

2. Narrative-driven volatility is the new normal. A 7,000-word thought experiment on Substack moved the Dow by 800 points. That tells you how sensitive the market is to narrative right now. Headlines and reports about AI disruption — whether based on data or speculation — can trigger sharp moves. Understanding the narrative is part of understanding the price action.

3. Watch Q1 2026 earnings season closely. The next major test for AI sentiment will come when companies report Q1 results and are forced to demonstrate tangible AI-driven revenue growth. If the numbers disappoint, expect more of the "capex fatigue" selling. If they deliver, it could stabilise sentiment — at least temporarily.

4. The "broadening" trade deserves attention. As the "Magnificent Seven" tech stocks face repricing pressure, the rest of the market — companies less dependent on GPU supply chains and AI infrastructure — has started to find support. The median S&P 500 stock is outperforming the tech-heavy benchmarks. If you trade indices, this broadening dynamic matters for where leadership sits.

5. Gold and the dollar are reading the same room. When AI fear triggers risk-off sentiment, the traditional macro playbook reasserts itself. Gold tends to benefit from uncertainty. The US dollar reacts to shifting rate expectations. Understanding how AI panic feeds through to these asset classes helps you stay ahead of moves in the markets you actually trade.

The Big Picture: Disruption Isn't Linear

What we're watching unfold isn't simply "AI is good" or "AI is bad" for markets. It's the messy, volatile process of price discovery — the market trying to figure out what AI is actually worth, how fast its impact will be felt, and who wins and who loses along the way.

There's a concept in economics called the Jevons Paradox: when a technology becomes more efficient, total demand for the resource it uses often increases rather than decreases. We've seen signs of this playing out with AI already. DeepSeek proved AI could be built more cheaply — and instead of reducing overall AI spending, the market expanded. More companies could afford to build. More use cases became viable. Efficiency created more demand, not less.

The key challenge for markets right now is that the promise of AI and the pricing of AI are two different things. The technology can be genuinely transformative and still be temporarily overvalued. The internet proved this exact lesson 25 years ago. The traders who understood that distinction — who respected the technology but also respected valuation and timing — were the ones who navigated that era successfully.

That's the mindset worth adopting now. Not predicting a crash. Not dismissing the fear. But understanding what the market is processing and why — so you're prepared for whatever happens next.