For more than three years, the ascent of U.S. equities has rested on a deceptively simple proposition: artificial intelligence would act as a universal productivity multiplier. From enterprise software to legacy industrial firms, executives embraced AI not merely as a tool, but as a narrative—one that justified higher margins, leaner workforces, and, crucially, richer valuations.
That narrative is now fraying.
A single breakthrough in large language models, or the release of a new developer-facing AI product, can suddenly erase billions of dollars in market capitalization across entire cohorts of public companies. What was once framed as AI enhancing business models is increasingly perceived as AI substituting them. The distinction matters. Enhancement supports valuation; substitution destroys it.
This shift has already left its mark on the software sector, where stock price corrections have been sharp, persistent, and indiscriminate. The question confronting markets is no longer whether AI will raise productivity, but whether it will hollow out the very income streams that once sustained demand.
A Thought Experiment That Refuses to Stay Fictional
On February 23, Citrini Research, a U.S.-based equity research firm, released a report titled The 2028 Global Intelligence Crisis. Written as a retrospective narrative from the future, the report deliberately adopts a fictionalized tone. Its purpose is not to forecast precise outcomes, but to stress-test the global economy under conditions of repeated AI capability breakthroughs.
The central claim is stark: any business model dependent on information asymmetry, human mediation, or consumer inertia is structurally vulnerable. Consulting, market research, recruiting, advertising, insurance brokerage, and even parts of legal and financial services are not merely exposed—they are conceptually obsolete in a world where intelligence itself becomes cheap, abundant, and instantly deployable.
The report goes further. As AI displaces human cognition as a core input to production, white-collar employment absorbs the shock. Consumer spending, long the bedrock of developed economies, weakens accordingly. Productivity rises, but purchasing power does not.
While the report’s narrative is set in the future, its opening chapters feel uncomfortably familiar.
When Software Became Reproducible
The inflection point, in this telling, arrives in late 2025. AI-assisted coding tools cross a threshold where replicating the core functionality of a high-priced SaaS product no longer requires a large engineering team—or much time at all. With tools comparable to Claude Code and Codex, internal corporate teams can recreate software once sold for hundreds of thousands of dollars annually in a matter of weeks.
The first visible response appears in 2026 procurement budgets. Long-tail SaaS providers—companies serving fragmented, highly specific workflows—are hit earliest. Firms resembling Monday.com, Zapier, and Asana find themselves squeezed from both sides: customers question renewal costs, while internal teams quietly experiment with bespoke AI-driven alternatives.
Larger SaaS platforms initially assume insulation. They are wrong. AI collapses barriers to entry, erodes switching costs, and forces incumbents into direct competition with model providers such as OpenAI and Anthropic. At the same time, headcount reductions—celebrated as efficiency gains—undermine the per-seat licensing model on which enterprise software economics depend.
In October 2026, ServiceNow announces layoffs while reallocating capital toward AI development. The move is praised by markets. It also accelerates the cycle that makes the next round of layoffs inevitable.
Markets Celebrate, Then Misprice the Risk
For much of 2026, capital markets interpret these shifts benignly. Layoffs boost margins. Earnings beats multiply. AI-related compute spending is booked as capital expenditure, reinforcing the idea of a structural productivity revolution rather than a demand shock.
By the autumn, equity indices reflect that optimism. The S&P 500 approaches 8,000. The Nasdaq Composite breaches 30,000. Output per hour grows at its fastest pace in decades.
Yet beneath the surface, real wage growth stalls. High-paying managerial and specialist roles disappear, replaced not by equivalent opportunities but by lower-paying service and gig work. The economy produces more—but distributes less.
The Collapse of the Intermediary Layer
As AI capabilities expand beyond software development, the impact on intermediary industries becomes unavoidable. The core value proposition of consulting firms, research agencies, recruiters, and brokers—aggregating and interpreting information—loses its economic foundation.
From the second half of 2026 onward, corporations begin internalizing analysis once outsourced. Custom AI systems generate strategic reports in minutes that previously required weeks of billable human labor. Revenue at mid-sized consulting firms contracts sharply, triggering mass layoffs.
This is not incremental automation. It is margin compression at the industry level. When the cost of intelligence approaches zero, the intermediary profit layer collapses.
A parallel shock hits consumer-facing businesses reliant on inertia. Subscription models built on auto-renewals and behavioral friction are quietly dismantled as AI agents enter everyday use. By 2027, personal AI proxies routinely cancel unused services, renegotiate contracts, and switch providers without human intervention. “Sticky” revenue is no longer sticky.
Ghost GDP and the Demand Vacuum
The macroeconomic consequences emerge slowly, then all at once. Consumption growth decelerates. Credit card delinquencies rise. Early warnings are dismissed as cyclical noise.
Observers begin using a new term: Ghost GDP—output generated by AI systems that appears on balance sheets but never reaches household incomes. Productivity rises even as demand weakens.
Corporations respond predictably. Labor savings are reinvested into compute, reinforcing a loop in which AI replaces more workers, suppresses more income, and further undermines consumption. Unlike humans, AI does not spend.
By mid-2027, valuation assumptions unravel. Software firms dependent on recurring revenue face stalled renewals and rising defaults. Private equity-backed assets falter. After a decade without a true default cycle, markets are forced to reprice risk abruptly.
The strain extends into housing. A mortgage market built on stable white-collar incomes confronts rising credit risk. Asset prices stagnate. Deflationary expectations take hold.
A Global Shock, Not a Local One
The downturn that follows, in late 2027, is not sector-specific. It reflects a collective reassessment of the relationship between productivity, income, and consumption.
The shock propagates globally. For two decades, countries such as India anchored growth to the export of white-collar services. AI breaks the logic of labor arbitrage itself. Firms no longer shift work to cheaper geographies; they eliminate the work altogether.
By replacing offshore developers, support staff, and data processors with internal AI systems, multinational corporations dismantle the service outsourcing model. Employment contracts in both developed and emerging markets simultaneously. The expansion of the global middle class stalls.
The Misunderstood Constraint
The lesson embedded in this scenario is not that AI fails to deliver productivity. It is that productivity alone cannot sustain a consumption-driven economy.
Markets spent years assuming that efficiency gains would naturally translate into prosperity. AI exposes the flaw in that assumption. Intelligence, once a scarce and monetizable input, becomes abundant. When income growth decouples from output growth, the system destabilizes.
The uncomfortable implication is that the AI era does not merely demand new technologies or new regulations. It demands a rethinking of how economic value is distributed once intelligence itself is no longer the limiting factor.
That reckoning, fictionalized or not, is already underway.
I’ve spent years analyzing technology cycles through the lens of markets rather than engineering hype. What stands out in this AI wave is not the speed of innovation, but the silence around its second-order effects.
Productivity has always been celebrated as an unquestioned good. Yet history shows that economies do not fracture when they fail to produce—they fracture when production detaches from income. AI accelerates that detachment.
This is not a prediction of collapse, nor a rejection of technological progress. It is a reminder that intelligence, once commoditized, stops being a source of broad-based prosperity unless economic structures adapt. Markets may applaud efficiency today, but sustainability is decided elsewhere.