The Coming Winter
In the second half of 2026, the years-long AI hype bubble will reach its definitive peak.
This is not pessimism toward the industry, but a sober judgment based on fundamental weaknesses. The ongoing AI boom is artificial, fueled by capital infusion and massive computing power expansion rather than genuine market demand.
Once the bubble bursts, the industry will enter a phase of contraction and layoffs instead of new growth. Unlike previous industrial revolutions, the first and hardest-hit groups will no longer be manual workers, but ordinary white-collar employees and middle-class practitioners across all sectors.
The Computing Arms Race: Endless Investment, No Sustainable Returns
The global AI industry has been trapped in an involuntary and self-perpetuating computing arms race.
Tech giants continue to pour hundreds of billions into data center expansion, upgrading chips, power supply systems and server infrastructure. The OpenAI-led Stargate project alone has a planned investment of $500 billion. The industry rule is brutal and straightforward: stagnation means falling behind; hesitation means elimination. No enterprise dares to halt expansion.
This is no longer rational commercial investment. Profitability and return on investment are no longer priorities; the only goal is to outspend and outlast competitors.
The fatal flaw of this frenzy is that nearly all capital and resources flood exclusively into the supply side.
Computing capacity expands continuously, large models iterate rapidly, and AI's per-token cost plummets year over year. Current industrial supply has far exceeded market demand, resulting in severe structural overcapacity.
While production capacity grows explosively, commercially viable, paid demand remains largely stagnant.
Token Overcapacity: The Supply-Demand Gap That Forms the AI Bubble
The harshest truth of the current industry: AI computing and token supply are thoroughly oversupplied.
Most tech firms rely on the Jevons paradox for growth forecasts: lower operational costs drive higher user adoption and expanded market demand. This logic works in the early technological popularization stage but has firm growth ceilings and cannot be applied indefinitely.
Sequoia Capital delivered a sobering calculation: to digest the massive global computing capacity accumulated over the years, the AI industry needs an additional $600 billion in annual revenue. This enormous gap has never been filled, as paid demand growth vastly lags behind computing expansion.
Most viral AI applications are pseudo-demand sustained solely by venture capital subsidies. They rely on cash burning to acquire users, accumulate data and create market hype. Once financing tightens and subsidies recede, most products without viable business models will quickly collapse.
The core essence of the AI bubble lies in the severe mismatch between massive idle computing capacity and sluggish paid market demand.
When capital ultimately realizes that endless computing investment cannot generate matching commercial returns, the AI bubble will deflate rapidly. The consequences will be frozen funding, halved corporate valuations, mass startup failures, and industry-wide layoffs.
Nevertheless, corporate closures and layoffs are merely superficial shocks. The far deeper impact is the collapse of the century-old stable economic cycle disrupted by AI.
Foundational Economic Rule: Producers Are the Core Consumers
A century ago, Henry Ford doubled wages for assembly line workers, a decision widely regarded as uneconomical at that time.
Yet Ford's logic was simple and profound: workers who manufacture cars must be able to afford the cars they build.
This is the century-old foundational cycle of the market economy: enterprises pay wages, residents generate consumption, consumption sustains corporate revenue, and revenue stabilizes employment. Producers and consumers are inherently the same group. Slashing ordinary people's income fundamentally drains the market's overall purchasing power.
This time-tested economic loop is now being rapidly dismantled by AI.
AI Replacing White-Collar Workers Drains Social Consumption Power
For individual enterprises, replacing standardized white-collar roles with AI is a fully rational business decision: it drastically cuts labor costs while maintaining or even improving operational efficiency.
However, when the entire industry adopts the same strategy, individual rationality evolves into collective irrationality and systemic disaster.
The white-collar employees laid off by enterprises are customers of other industries; the saved labor costs are vital liquidity supporting social consumption. Layoffs optimize finances for single companies but erode the fundamental consumption foundation of the entire economy.
This is a classic economic fallacy of composition: optimal solutions for individual entities eventually lead to systemic collapse.
Cost savings from AI-driven layoffs are not newly generated profits, but merely a transfer and overdraft of overall social purchasing power. The wider and faster AI replaces human labor, the lower household income falls, and the weaker market consumption becomes.
Vicious Cycle Formed: More AI Adoption, More Layoffs, Weaker Market Vitality
Two negative chains covering industry and economy intertwine tightly, forming an irreversible downward loop.
Industrial chain: Excessive computing expansion → Severe overcapacity → Supply-demand imbalance → Bubble burst → Capital flight → Large-scale industry layoffs.
Economic chain: AI replaces cognitive jobs → Middle-class income decline → Consumption contraction → Corporate revenue slump → Forced cost reduction → Further AI adoption and layoffs.
The two chains reinforce and worsen each other. Unemployment suppresses consumption, and sluggish consumption forces further layoffs, weakening economic vitality in every cycle.
This downturn will not be a one-time crash, but a gradual step-by-step decline. Each adjustment appears mild and controllable, until the public realizes that employment and the overall economy have deteriorated substantially.
Why White-Collar Workers Bear the Brunt of the AI Winter
Automation revolutions over the past two centuries mainly targeted manual labor. Manual workers were not core consumers, and emerging industries gradually absorbed displaced labor, providing sufficient social buffer time.
The current AI revolution breaks all previous rules.
AI precisely targets and replaces standardized cognitive work: basic programming, routine data analysis, template-based copywriting, standardized customer service, junior legal work and entry-level consulting. White-collar workers in these fields constitute the backbone of the middle class and the core driving force of social consumption.
Middle-class job losses and income shrinkage directly undermine the core momentum of economic circulation.
The most brutal factor is iteration speed. AI updates far faster than new jobs can be created, leaving ordinary people no time to transition or adapt, forcing them to accept pay cuts and unemployment. The unprecedented speed of technological iteration is the deadliest force of this industrial reshuffle.
Pragmatic Survival Rules for Ordinary People in the AI Winter
I do not deny AI's technological progress and efficiency value, but technological advancement does not equal economic prosperity, nor does it guarantee job security for ordinary people.
AI is eliminating replicable, process-driven, standardized white-collar work in batches. The era of making a living relying on proficiency, fixed templates and repetitive execution is completely over. Those who only operate AI tools and complete standardized tasks will be the first to be eliminated.
The most risk-resistant core skills for the future are not trendy AI tool proficiencies, but uniquely human capabilities beyond AI's reach: trade-off judgment in complex scenarios, in-depth practical understanding of real businesses, accountability for outcomes and risks, and industry experience accumulated through long-term practice.
These capabilities cannot be trained by data, copied from templates, or mass-produced. They are forged only through real business practice, trial and error, and taking responsibility for tangible results.
This is why I no longer chase AI industry fads. Market trends fade rapidly, and superficial follower skills are easily phased out by technological iteration. The most reliable personal asset is independent cognition and practical capability that belongs exclusively to oneself — platform-independent, hype-independent, and irreplaceable. Growth is slow yet solid and unassailable.
The bursting of the AI bubble, industrial cooling and shrinking employment are irreversible cyclical trends. Ordinary people cannot reverse the general environment, stop technological progress, or prevent corporate cost-cutting and layoffs.
The only viable option is to adapt proactively: abandon the fantasy of instant success, discard blind trend-chasing, and deepen exclusive core capabilities. In a downward cycle, quick wealth is an unrealistic delusion. For ordinary people, avoiding technological replacement and industrial elimination, and steadily safeguarding one's career and life, is the wisest choice in this era.