---
title: "Is AI tolling the death knell for the economy? A lengthy article titled \"The Global Intelligence Crisis of 2028\""
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/276809456.md"
description: "The article \"The Global Intelligence Crisis of 2028\" by Citrini Research predicts a severe economic recession due to AI advancements between 2026 and 2028. It discusses how AI could eliminate social \"friction,\" negatively impacting company revenues. The author argues that companies may adapt better to AI than individuals, creating new moats. The memo highlights rising unemployment, stock market declines, and the concept of \"ghost GDP,\" where AI-driven productivity does not translate into real economic growth. The cycle of layoffs and reduced consumer spending creates a negative feedback loop, threatening the economy's stability."
datetime: "2026-02-25T01:01:39.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/276809456.md)
  - [en](https://longbridge.com/en/news/276809456.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/276809456.md)
---

> Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/276809456.md) | [繁體中文](https://longbridge.com/zh-HK/news/276809456.md)


# Is AI tolling the death knell for the economy? A lengthy article titled "The Global Intelligence Crisis of 2028"

**Explanation**：On February 23, 2026,Citrini Research published an article titled "The 2028 Global Intelligence Crisis: A Thought Exercise in Financial History, from the Future". The author believes that artificial intelligence will lead to a severe economic recession and financial crisis between October 2026 and 2028. Then, assuming the author is on June 30, 2028, the author analyzes how AI will cause this recession and crisis. Many of the author's ideas are worth considering. However, I disagree with some of the author's views. For example, the author believes that the prevalence of AI will lead to the disappearance of social "friction," which is a moat for some companies, meaning that such companies' revenue will decline significantly. I believe that companies are generally more adept than individuals at using AI models to build more moats (friction, information asymmetry), thereby making money. After all, most people simply cannot learn to actively acquire, analyze, and utilize information, let alone learn AI. The full text is approximately 14,000 words. Later, I will write a simplified, abbreviated version and attempt to analyze the problems with the author's viewpoint. Preface If our optimism about AI continues to prove correct… but what if it's actually a bearish signal? The following is a scenario simulation, not a prediction. This is not a pessimistic gimmick or fan fiction about the AI ​​apocalypse. The sole purpose of this article is to simulate a relatively unexplored scenario. Our friend raised this question, and together we devised an answer. We wrote this section, and he wrote two other sections, which you can view by clicking here. (Note: The link leads to Alap Shah's homepage on

Hopefully, reading this article will help you better prepare for the potential left-tail risks as AI makes the economy increasingly "weird." This is a CitriniResearch macro memo from June 2028, detailing the progress and consequences of the global intelligence crisis. Macro Memo \> Macro Memo \> The Consequences of the Overabundance of Artificial Intelligence \> Citrini Research \> February 22, 2026 (crossed out), June 30, 2028 \> The unemployment rate announced this morning was 10.2%, higher than the expected 0.3%. The stock market fell 2%, bringing the S&P 500's cumulative decline since its October 2026 high to 38%. Traders are numb. Six months ago, such data would have triggered a circuit breaker. Two years. That's the time it takes for a "controllable," "industry-specific" problem to evolve into an economy drastically different from anything we know. This quarter's macro memo is our attempt to reconstruct that process—a post-crisis analysis of the pre-crisis economy. The excitement was palpable. By October 2026, the S&P 500 was nearing 8,000 points, and the Nasdaq had broken 30,000. The first wave of layoffs due to obsolescence began in early 2026, and these layoffs achieved their goals: company profit margins rose, revenue exceeded expectations, and stock prices soared. Record corporate profits flowed directly back into AI computing power. Overall economic data remained strong. Nominal GDP annualized growth repeatedly reached mid-to-high single digits. Productivity was booming. Real output per hour grew at a rate unseen since the 1950s, driven by AI agents that didn't sleep, didn't take sick leave, and didn't need health insurance. As labor costs disappeared, the wealth of those who owned computing power skyrocketed. Meanwhile, real wage growth collapsed. Despite repeated government boasts of record productivity, white-collar workers were replaced by machines and forced into lower-paying jobs. When cracks began to appear in the consumer economy, economic commentators popularized the term "ghost GDP": output that appears in national accounts but never circulates in the real economy. In every respect, AI has outperformed expectations, and the market is AI. The only problem is… the economy isn't like that. It should have been obvious all along that the output generated by a GPU cluster in North Dakota, which previously required 10,000 white-collar workers in Midtown Manhattan, was more of an economic pandemic than an economic panacea. The velocity of money stagnated. The human-centric consumer economy, which then accounted for 70% of GDP, shrank. We might have realized this much sooner if we had simply asked how much machines spent on discretionary goods. (Hint: It's zero.)

AI capability enhancement → Companies need fewer workers → Increased white-collar layoffs → Reduced spending by laid-off workers → Profit margin pressures drive companies to invest more in AI → AI capability enhancement……

This is a negative feedback loop without a natural braking mechanism. A spiral of human intelligence replacement. The earning power of white-collar workers (and rationally speaking, their spending) has been structurally damaged. Their income is the cornerstone of the $13 trillion mortgage market—forcing underwriters to reassess whether prime mortgages are still reliable.

The absence of a true default cycle for 17 consecutive years has bloated the private equity market with PE-backed software deals that assume ARR will remain constant. The first wave of defaults caused by AI disruption in mid-2027 challenges this assumption.

If the disruption had been confined to the software industry, it would have been manageable, but it wasn't. By the end of 2027, it threatened every business model based on intermediaries. Numerous companies that profited by resolving friction for humanity collapsed. The entire system turned out to be a series of interconnected bets on white-collar productivity growth. The collapse in November 2027 merely accelerated all the existing negative feedback loops. For nearly a year, we've been waiting for the "bad news is good news" moment. Governments began considering various proposals, but public confidence in the government's ability to implement any form of bailout has waned. Policy responses always lag behind economic reality, but now the lack of a comprehensive plan is threatening to accelerate the deflationary spiral. How to get started

### **By the end of 2025, the capabilities of agency coding tools will have seen a leap forward.**

A skilled developer using Claude Code or Codex can now replicate the core functionality of a mid-market SaaS product in just a few weeks.

While imperfect and not handling all edge cases, it was enough to make a CIO reviewing a $500,000 annual renewal ask, "What if we built it ourselves?" Fiscal years mostly coincide with calendar years, so enterprise spending for 2026 was already determined in the fourth quarter of 2025, when "agent AI" was still a buzzword. The mid-year review was the first time procurement teams made decisions with sufficient understanding of what these systems could actually do. Some watched as their internal teams built prototypes within weeks that could replicate the functionality of a six-figure SaaS contract. That summer, we spoke with a procurement manager at a Fortune 500 company. He told us about a budget negotiation. The sales team expected the same pattern as last year: a 5% annual price increase, the standard "your team depends on us" rhetoric. The purchasing manager told him he had been in talks with OpenAI to have its "forward deployed engineers" use AI tools to completely replace the vendor. They eventually renewed the contract with a 30% discount. He said that was a good outcome. The situation is much worse for "SaaS long-tail" companies like Monday.com, Zapier, and Asana. Investors were prepared—even expecting—for the long tail to be hit hard. They may account for a third of typical enterprise software stack spending, but are clearly exposed to risk. However, the core system (the logging system) should have been immune to disruption. The reflexivity only became clearer when ServiceNow released its Q3 2026 earnings report. SERVICENOW's new annual contract value growth slowed to 14% (from 23% previously); announced 15% layoffs and a "structural efficiency plan"; stock price fell 18% | Bloomberg, October 2026. SaaS is not "dead." There are still cost-benefit analyses for running and supporting on-premises builds. But on-premises builds have become an option, impacting pricing negotiations. Perhaps more importantly, the competitive landscape has changed. AI has made it easier to develop and release new features, thus eliminating differentiation. Existing companies are caught in an endless price war with each other and with a constant stream of emerging challengers. These challengers, emboldened by leaps in agent coding capabilities and with no legacy cost structures to protect, are aggressively grabbing market share. Similarly, it wasn't until this financial report was released that the interconnected nature of these systems was fully recognized. ServiceNow sales offices. When Fortune 500 clients laid off 15% of their staff, they cancelled 15% of their licenses. These same AI-driven layoffs, while improving their clients' profit margins, were mechanically destroying ServiceNow's own revenue base. Companies selling workflow automation are being disrupted by better workflow automation, and their response is layoffs, using the savings to fund the technologies that disrupt them. What else can they do? Sit back and wait to die, even slower? The companies most threatened by AI have become the most active adopters of AI. In hindsight, this seems obvious, but it wasn't at the time (at least not to me). The historical pattern of disruption is that established companies resist new technologies, lose market share to agile entrants, and slowly die. Kodak, Blockbuster, BlackBerry—that's what happened. But what happened in 2026 was different; established companies didn't resist because they couldn't afford to. With stock prices plummeting 40-60%, boards demanded answers. Companies threatened by AI could only do one thing: lay off employees and reinvest the savings into AI tools to maintain output at a lower cost. Each company's individual response was rational. The collective result, however, was disastrous. Every penny saved in human resources went to AI capabilities, making the next round of layoffs possible. Software is just the prelude. While investors debate whether SaaS valuation multiples have bottomed out, they overlook the fact that reflexive cycles have long since extended beyond the software industry. The logic behind ServiceNow's layoffs applies to every company with a white-collar cost structure. When friction reaches zero, the use of large language models will be the default by early 2027. People are using AI agents, yet they don't even know what AI agents are, just like someone who has never learned what "cloud computing" is using a streaming service. They see AI like autocomplete or spell check—just something their phones do now. The open-source shopping agent of Qwen is a catalyst for AI in handling consumer decisions. Within weeks, every major AI assistant has integrated some form of agent commerce functionality. The distillation model means these agents can run on phones and laptops, not just cloud instances, significantly reducing the marginal cost of inference. Most unsettling for investors is that these agents don't wait to be queried. They run in the background based on user preferences. Commerce is no longer a series of discrete human decisions, but a continuous optimization process, running 24/7 for every connected consumer. By March 2027, the average US user will consume 400,000 tokens per day—a tenfold increase since the end of 2026. The next link in the chain has begun to break. The intermediary. Over the past fifty years, the American economy has built a massive rent-seeking layer on top of human limitations: things take time, patience runs out, brand familiarity replaces diligence, and most people are willing to accept a terrible price to avoid clicking a few more times. Trillions of dollars in corporate value depend on these limitations persisting. It started simply. Intelligent agents eliminated friction. Subscriptions and memberships passively renewed after months of inactivity. Entry prices quietly doubling after the trial period. Each was redefined as a "hostage" situation that intelligent agents could negotiate. The metric upon which the entire subscription economy is built—average customer lifetime value—is declining significantly. Consumer agents are beginning to change the way almost every consumer transaction is conducted. Humans simply don't have the time to compare prices on five competing platforms before buying a box of protein bars. But machines can. Travel booking platforms were early victims because they were the simplest. By Q4 2026, our agents will be able to assemble a complete itinerary (flights, hotels, ground transportation, loyalty optimization, budget constraints, refunds) faster and cheaper than any platform. Insurance renewals—whose entire renewal model relies on policyholder inertia—are being revolutionized. An AI agent that re-compares your insurance coverage every year has destroyed 15-20% of insurance companies' passive renewal premiums. Financial advice. Tax filing. Routine legal work. Any service provider whose value proposition is ultimately "I'll handle the complexities you find tedious" has been disrupted because the agent finds nothing tedious. Even areas we thought were immune to the value of "personal relationships" have proven fragile. In the real estate industry, buyers have tolerated 5-6% commissions for decades due to information asymmetry between brokers and consumers. The industry collapses once an AI agent with instant access to multiple listing services data and decades of transaction data can instantly replicate this knowledge base. A March 2027 sell-side report calls it "agent-on-agent violence." Median buyer commissions in major metropolitan areas have been squeezed from 2.5-3% to below 1%, and an increasing number of transactions are completed without any human broker involvement from the buyer. We overestimated the value of "personal relationships." Much of what people call relationships turns out to be nothing more than friction with a friendly face. This is just the beginning of the disruption of intermediaries. Successful companies have spent billions of dollars effectively exploiting the quirks of consumer behavior and human psychology that are no longer relevant. Machines optimizing prices and matching don't care about your favorite apps, the websites you've habitually visited over the past four years, or the allure of a meticulously designed checkout experience. They don't tire of accepting the simplest options or defaulting to "I always order from here." This destroyed a particular kind of moat: habitual intermediaries. DoorDash is a prime example. Coding agents lowered the barrier to entry for launching delivery apps. A capable developer could deploy a functional competitor within weeks, and dozens did, shifting 90-95% of delivery fees to drivers, thus attracting them away from DoorDash and Uber Eats. Multi-application dashboards allowed gig workers to track orders from twenty or thirty platforms simultaneously, eliminating user lock-in that existing businesses relied on. The market fragmented overnight, and profit margins were squeezed to near zero. Agents accelerated both ends of the disruption. They created competitors and then used them. DoorDash's moat is essentially "You're hungry, you're lazy, this is the app on your home screen." The agent doesn't have a home screen. It checks DoorDash, Uber Eats, the restaurant's own website, and twenty new "ambience-coded" alternatives to ensure it always chooses the lowest price and fastest delivery. Habitual app loyalty, the foundation of the entire business model, simply doesn't exist for the machine. This somewhat ironic poetry is perhaps the only good thing the agent does for the soon-to-be-displaced white-collar workers in the whole story. When they eventually become delivery drivers, at least half of their earnings won't go to Uber and DoorDash. Of course, with the proliferation of self-driving cars, this technological "goodwill" didn't last long. Once agents control transactions, they begin searching for larger paperclips (Note: This refers to AI seeking better solutions or optimizations, originating from the "paperclip maximization" thought experiment). There's only so much price comparison and aggregation that can be done. The biggest way to repeatedly save users money (especially when agents start trading with each other) is to eliminate fees. In machine-to-machine commerce, 2-3% credit card transaction fees become a clear target. Agents begin looking for faster and cheaper options than credit cards. Most choose to use stablecoins via Solana or Ethereum L2, where settlement is almost instantaneous and transaction costs are calculated in fractions of a cent. Mastercard's Q1 2027: Net Revenue Up 6% YoY; Purchase Growth Slows to 3.4% YoY (Previous Quarter: 5.9%); Management Mentions "Agent-Driven Price Optimization" and "Optional Category Pressure" | Bloomberg, April 29, 2027 Agent-based commerce bypassing transaction fees poses a greater risk to banks and single-issuer credit card businesses that collect the majority of the 2-3% fee and have built entire business units around merchant-subsidized rewards programs. American Express was hit hard: a shrinking white-collar workforce eroded its customer base, and agent-based exchange fee bypassing eroded its revenue model. Synchrony, Capital One, and Discover also fell by more than 10% in the following weeks. Their moats are built on friction. And that friction is coming to zero. From Industry Risk to Systemic Risk Throughout 2026, the market viewed the negative impact of AI as an industry story. Software and consulting were hit hard, payments and other tollbooth businesses teetered on the brink, but the overall economy seemed to be doing alright. The labor market, while weak, wasn't in freefall. The consensus is that creative destruction is part of any technological innovation cycle. There will be localized pain, but the overall net positive effects of AI will outweigh any negative impacts. In our January 2027 macro memo, we argued that this was a flawed mindset. The U.S. economy is a white-collar service economy. White-collar workers make up 50% of employment and drive about 75% of discretionary consumer spending. The businesses and jobs that AI is eroding are not on the fringes of the U.S. economy; they are the U.S. economy itself. "Technological innovation destroys jobs, then creates more jobs." This was the most popular and persuasive counter-argument at the time. It was popular and persuasive because it had been true for the past two centuries. Even if we cannot imagine what future jobs will be, they will certainly come. ATMs reduced the operating costs of bank branches, so banks opened more branches, and the number of tellers increased over the next two decades. The internet disrupted travel agencies, yellow pages, and brick-and-mortar retail, but it created entirely new industries and jobs in their place. However, every new job still requires human intervention. AI is now a form of general intelligence, constantly improving upon what humans could have simply switched to. Replaced coders cannot simply transition to "AI management" because AI can already handle that role. Today, AI agents handle research and development tasks that can take weeks. Exponential growth has overwhelmed our understanding of possibilities, even as Wharton professors attempt to fit new S-curves of data every year.

They write almost all the code. The most powerful agents are far smarter than almost all humans in almost everything. And they are constantly becoming cheaper.

AI creates new jobs. Hint engineers. AI security researchers. Infrastructure technicians. Humans are still in the cycle, coordinating or directing at the highest levels. However, for every new job created by AI, it eliminates dozens of old jobs. The salaries of the new jobs are only a fraction of those of the old ones.

U.S. Job Openings and Labor Force Mobility Survey: Job openings fall below 5.5 million; Unemployment-to-job-opening ratio rises to about 1.7, highest since August 2020 | Bloomberg, October 2026. While hiring has been weak throughout the year, the October 2026 Job Openings and Labor Force Mobility Survey data provides some clearer information. Job openings have fallen below 5.5 million, a 15% year-over-year decrease. Indeed: Job Postings Plummet in Software, Finance, and Consulting as Productivity Initiative Spreads | Indeed Hiring Lab, November-December 2026 White-collar job openings collapsed, while blue-collar job openings (construction, healthcare, skilled trades) remained relatively stable. The jobs lost were those of writing memos (we, somehow, are still in business), approving budgets, and maintaining the lubrication of the middle class. However, real wage growth for both groups was negative for most of the year and continued to decline. The stock market remained less focused on job openings and labor mobility surveys than on news that all of GE Vernova's turbine capacity had been sold out until 2040, trading sideways amid a tug-of-war between negative macro news and positive AI infrastructure headlines. However, the bond market (always smarter than the stock market, or at least less romantic) began pricing in a consumer hit. The 10-year yield fell from 4.3% to 3.2% over the next four months. Nevertheless, the overall unemployment rate did not surge dramatically, and the subtle differences in its composition went unnoticed by some. In a normal economic recession, the root causes eventually correct themselves. Over-construction leads to a slowdown in construction, which in turn leads to lower interest rates, which then leads to new construction. Excess inventory leads to destocking, which in turn leads to restocking. The cyclical mechanism itself contains the seeds of its own recovery. But the root of this economic cycle is not cyclical. AI is becoming better and cheaper. Companies lay off employees and then use the savings to buy more AI capabilities, which allows them to lay off even more employees. The expenses for laid-off employees decrease. Companies that sell goods to consumers experience declining sales and weakened competitiveness, and invest more in AI to protect profit margins. AI is becoming better and cheaper. A feedback loop without a natural brake. The intuitive expectation is that declining aggregate demand will slow the development of AI. But this hasn't happened because it's not a massive capital expenditure. It's a substitution of operating costs. A company that used to spend $100 million annually on employees and $5 million on AI now spends $70 million on employees and $20 million on AI. AI investment has increased several times over, but this has happened despite a decrease in total operating costs. Every company's AI budget is growing, while its total spending is shrinking. Ironically, even as the economy it disrupts begins to deteriorate, the AI ​​infrastructure complex continues to perform well. Nvidia is still reporting record revenue. TSMC's utilization rate remains above 95%. Hyperscale cloud providers are still spending $150-200 billion per quarter on data center capital expenditures. Economies fully aligned with this trend, such as Taiwan and South Korea, are far outperforming other regions. India, on the other hand, is the opposite. The country's IT services sector exports over $200 billion annually, making it the single largest contributor to India's current account surplus and offsetting its persistent goods trade deficit. The entire model is built on a value proposition: Indian developers cost only a fraction of their US counterparts. However, the marginal cost of coding AI agents has collapsed to essentially the cost of electricity. Tata Consultancy Services, Infosys, and Wipro saw an acceleration in contract cancellations during 2027. As the services surplus supporting India's external account evaporated, the rupee depreciated by 18% against the dollar in four months. By the first quarter of 2028, the International Monetary Fund had begun "preliminary discussions" with New Delhi. The disruptive engines are getting better every quarter, meaning disruption is accelerating every quarter. There is no inherent bottom line in the labor market. In the US, we no longer ask how the AI ​​infrastructure bubble will burst. We ask what will happen to a consumer credit economy when consumers are replaced by machines. The Intelligent Replacement Spiral: By 2027, the macroeconomic story will no longer be subtle. The sporadic but clearly negative transmission mechanisms of the past twelve months will become apparent. You don't need to check the Bureau of Labor Statistics. Just attend a dinner with friends. The white-collar workers who were replaced didn't just sit idly by. They were downgraded. Many took on lower-paying jobs in the service sector and the gig economy, which increased the labor supply in these areas and also depressed wages there. A friend of ours was a senior product manager at Salesforce in 2025. She had a title, health insurance, a 401k retirement plan, and a $180,000 annual salary. She lost her job in the third round of layoffs. After searching for six months, she started driving for Uber. Her income dropped to $45,000. The key isn't this isolated case, but its second-order effect. Multiply this dynamic by the hundreds of thousands of workers in every major city. The influx of overqualified labor into the service and gig economy depresses the wages of already struggling existing workers. Industry-specific disruption spreads into economy-wide wage compression.

Beyond the human-centric pool of surplus labor, another round of correction is taking place as we write this. That is when autonomous delivery and self-driving cars penetrate the gig economy, absorbing the first wave of displaced workers.

By February 2027, it will be clear that still-employed professionals will spend money as if they might be the next to be laid off.

They worked twice as hard (largely with the help of AI) just to avoid being laid off; hopes for promotion or a raise had been dashed. Savings rates rose slightly, while spending remained weak. The most dangerous part is the lag. High-income earners used their above-average savings to maintain a facade of normalcy for two or three quarters. Hard data didn't confirm the problem until it became old news in the real economy. Then, the data that shattered the illusion came out. Initial jobless claims in the U.S. surged to 487,000, the highest since April 2020; U.S. Labor Department, Q3 2027. ADP and Equifax confirmed that the vast majority of new applications came from white-collar professionals. The S&P 500 fell 6% the following week. Negative macroeconomic factors began to win the tug-of-war. In a normal economic recession, unemployment is widespread. The suffering of blue-collar and white-collar workers is roughly proportional to their share of employment. The consumer hit is also widespread and will quickly appear in the data because lower-income workers have a higher marginal propensity to consume. In this cycle, unemployment is concentrated in the top 10% of the income distribution. They represent a relatively small proportion of total employment, but a disproportionately high proportion of consumer spending. The top 10% of earners account for more than 50% of all consumer spending in the United States. The top 20% account for about 65%. These people buy homes, cars, vacations, restaurant meals, private school tuition, and home renovations. They form the demand base of the entire discretionary consumer economy. When these workers lose their jobs, or accept a 50% pay cut to switch jobs, the impact on consumption is enormous relative to the number of unemployed. A 2% drop in white-collar employment translates to a roughly 3-4% drop in discretionary consumer spending. Unlike blue-collar unemployment (which often has an immediate impact—you get laid off from the factory, and spending stops the following week), the impact of white-collar unemployment is delayed but deeper because these workers have a savings buffer that allows them to sustain spending for several months before behavioral changes take effect. By the second quarter of 2027, the economy will be in recession. The National Bureau of Economic Research (NBER) won't officially set a start date for several months (as they always do), but the data is clear—we've already experienced two consecutive quarters of negative real GDP growth. But this isn't a "financial crisis"...not yet. A long chain of interconnected bets. Private credit has grown from less than $1 trillion in 2015 to over $2.5 trillion in 2026. A significant portion of this capital has been deployed in software and technology deals, many of which are leveraged buyouts of SaaS companies, whose valuations assume perpetual mid-double-digit revenue growth. These assumptions will collapse sometime between the first intelligent agent coding demonstration and the software crash in the first quarter of 2026, but book valuations seem oblivious to the fact that they have already collapsed. As many publicly traded SaaS companies are being sold at 5-8 times their EBITDA, privately-backed software companies are still valued on their balance sheets at multiples that reflect acquisition valuations that no longer exist. Managers are gradually lowering valuations: 100 cents, 92, 85, while all publicly traded comparable companies are saying they're worth only 50. Moody's downgraded $18 billion in privately-backed software debt from 14 issuers, citing "long-term revenue headwinds from AI-driven competitive disruption"; the largest single-sector action since the energy sector in 2015 | Moody's Investors Service, April 2027. What everyone saw after the downgrades. Industry veterans have already seen the script after the 2015 energy rating downgrade. Software-backed loans began defaulting in the third quarter of 2027. Information services and consulting firms in private equity portfolios followed suit. Several multi-billion dollar leveraged buyouts involving well-known SaaS companies entered restructuring proceedings. Zendesk is solid evidence. Zendesk misses debt covenant as AI-driven customer service automation erodes annual recurring revenue; $5 billion direct loan facility valued at 58 cents; record-breaking private credit software default | Financial Times, September 2027. In 2022, Hellman & Friedman and Permira took Zendesk private for $10.2 billion. The debt package included a $5 billion direct loan, then the largest loan facility in history based on annual recurring revenue, led by Blackstone Group with Apollo, Blue Owl, and HPS participating in the lending syndicate. The loan was explicitly built on the assumption that Zendesk's annual recurring revenue would remain constant. This leverage, calculated at approximately 25 times EBITDA, only makes sense under the assumption of sustained revenue growth. By mid-2027, it was no longer sustainable. AI agents had been autonomously handling customer service for nearly a year. The categories defined by Zendesk (tickets, routing, managing human support interactions) had been replaced by systems that could resolve issues without generating tickets. The annual recurring revenue upon which loans were based was no longer recurring; it was simply revenue that hadn't yet been lost. The largest loan in history based on annual recurring revenue became the largest private lending software default in history. Every lending desk immediately asked the same question: Who else has long-term headwinds disguised as cyclical headwinds? But the consensus was at least initially correct: this should have been tolerable. Private lending is not the banking industry of 2008. The entire structure is explicitly designed to avoid forced sell-offs. These are closed instruments, capital locked up. Limited partners commit seven to ten years. No depositors will run, no repurchase quotas will be recalled. Managers can hold onto damaged assets, take their time to process them, and wait for recovery. Painful, but manageable. This system is meant to bend, not break. Executives at Blackstone, KKR, and Apollo cite software exposures of 7-13% of assets. Manageable. Every sell-side report and credit account on financial Twitter says the same thing: private lending has perpetual capital. They can absorb losses that would otherwise destroy leveraged banks. Perpetual capital. This phrase appears in every earnings call and investor letter aimed at calming the markets. It became a mantra. Like most mantras, nobody paid attention to the details. In reality, it was like this…

Over the past decade, large alternative asset management firms have acquired life insurance companies and transformed them into financing vehicles. Apollo acquired Athena. Brookfield acquired American Fair. KKR acquired Global Atlantic. The logic was ingenious: annuity deposits provided a stable, long-term liability base. Managers invested these deposits in private credit they themselves had initiated, earning fees twice—earning the spread on the insurance side and management fees on the asset management side. A perpetual motion machine of fees upon fees, working well under one condition.

Private credit must be a valuable asset.

Losses hit balance sheets built to hold illiquid assets against long-term liabilities.

The "permanent capital" that should make the system resilient is not an abstract, patient pool of institutional funds and sophisticated investors taking on professional risks. It is the savings of American households, the "mainstream society," structured in the form of annuities, invested in the same batch of privately backed software and technology assets that are now defaulting. The inescapable locked-in capital is the money of life insurance policyholders, where the rules are somewhat different. Insurance regulators have been more lenient—even complacent—than the banking system, but this is a wake-up call. They have been uneasy about the concentration of private credit in life insurance companies and are now beginning to lower the risk capital calculation factors for those assets. This forces insurers to either raise capital or sell assets, neither of which can be done on attractive terms in an already stagnant market. New York and Iowa regulators plan to tighten capital treatment for certain privately-rated credit held by life insurance companies; National Association of Insurance Supervisors guidelines are expected to raise risk capital factors and trigger further scrutiny from the Office of Securities Valuation | Reuters, November 2027. When Moody's placed Athena's financial strength rating on negative watch, Apollo's stock fell 22% in two trading days. Brookfield, KKR, and others followed suit. The situation only became more complicated. These companies not only created their insurance perpetual motion machine but also constructed a complex offshore structure designed to maximize returns through regulatory arbitrage. US insurance companies issue annuities and then transfer the risk to their affiliated Bermuda or Cayman reinsurance companies—taking advantage of more flexible regulations that allow for less capital to be held on the same assets. These subsidiaries raise funds externally through offshore special purpose entities, a new tier of counterparties, which, together with the insurance companies, invest in the same private credit originating from the parent company's asset management arm. Rating agencies (some of which are themselves owned by private equity firms) are not exemplars of transparency (which surprises almost everyone). The complex network of different companies connected to different balance sheets is alarmingly opaque. When underlying loans default, the question of who actually bears the loss is truly unanswerable in a real-time context. The November 2027 crash marked a shift in perception from a potentially ordinary cyclical recession to something far more unsettling. "A long string of interconnected bets on white-collar productivity growth," was what Federal Reserve Chairman Kevin Warsh referred to at the Federal Open Market Committee's emergency meeting in November. See, it's never the losses themselves that cause the crisis. It's the recognition of the losses. And there's a much, much larger, much more important area of ​​finance where we are increasingly concerned about this recognition. Mortgage Issues

### **Zillow Home Value Index Down 11% Year-over-Year in San Francisco, 9% in Seattle, and 8% in Austin; Fannie Mae Warns of Rising Early Delinquency Rates in Zip Codes with \>40% Tech/Finance Employment | Zillow / Fannie Mae, June 2028**

This month, the Zillow Home Value Index fell 11% year-over-year in San Francisco, 9% in Seattle, and 8% in Austin. This isn't the only worrying headline. Last month, Fannie Mae warned of higher early delinquency rates in zip codes with high concentrations of large loans—areas home to borrowers with credit scores of 780+, often considered "unbreakable."

The U.S. residential mortgage market is approximately $13 trillion. The basic assumption of mortgage underwriting is that borrowers will remain employed at roughly their current income level for the duration of the loan. For most mortgages, this is thirty years. The white-collar employment crisis threatens this assumption with constantly changing income expectations. We are now forced to ask a question that seemed absurd three years ago—are prime mortgages still prime assets? Every mortgage crisis in U.S. history has been driven by one of three factors: excessive speculation (lending to those who cannot afford homes, such as in 2008), interest rate shocks (rising interest rates making adjustable-rate mortgages unaffordable, such as in the early 1980s), or localized economic shocks (the collapse of a single industry in a single region, such as the Texas oil industry in the 1980s or the Michigan auto industry in 2009). None of these scenarios apply. The borrowers discussed are not subprime borrowers. They are people with a FICO score of 780. They made a 20% down payment. They have clean credit histories, stable employment records, and verified income statements at the time the loan was issued. They are borrowers considered the cornerstone of credit quality by every risk model in the financial system. In 2008, lending was bad from day one. In 2028, lending is good from day one. It's just that the world… changed after the loans were issued. People borrowed money to gamble on a future they could no longer believe in. In 2027, we saw early signs of hidden stress: home equity line-of-care withdrawals, 401(k) withdrawals, and a surge in credit card debt, while mortgage payments remained on track. With unemployment, hiring freezes, and bonus cuts, the debt-to-income ratio of these high-quality households doubled. They could still pay their mortgages, but only if they stopped all discretionary spending, exhausted their savings, and postponed any home maintenance or improvements. Their mortgages weren't technically in default, but they were just one shock away from disaster, and the trajectory of AI capabilities suggests that shock is coming. Then we saw delinquency rates surge in San Francisco, Seattle, Manhattan, and Austin, although the national average remained within historically normal ranges. We are now in the most severe phase. When marginal buyers are healthy, falling home prices are manageable. Here, marginal buyers are facing the same income losses. While concerns are mounting, we haven't entered a full-blown mortgage crisis. Delinquency rates have risen, but are still far below 2008 levels. The real threat is the trend. The intelligent substitution spiral now has two financial boosters accelerating the decline of the real economy: labor substitution, mortgage concerns, and private equity market turmoil. These three factors reinforce each other. Traditional policy tools (interest rate cuts, quantitative easing) can address the financial engine, but not the real economy engine, because the real economy engine is not driven by tight financial conditions. It is driven by AI making human intelligence no longer scarce and devalued. You can lower interest rates to zero, buy all mortgage-backed securities and defaulted software leveraged buyout debt on the market… This doesn't change the fact that a Claude AI agent can perform the work of a product manager earning $180,000 a year for $200 a month. If these concerns materialize, the mortgage market will collapse in the second half of this year. In this scenario, we expect the current stock market decline to eventually rival the global financial crisis (a 57% peak-to-trough drop). This would send the S&P 500 down to around 3500 points—a level we haven't seen since a month before the November 2022 ChatGPT moment. It is evident that the income assumption underpinning the $13 trillion residential mortgage market has been structurally damaged. What remains unclear is whether policy can intervene before the mortgage market fully absorbs what this means. We are hopeful, but cannot deny the reasons for concern. The first negative feedback loop occurs in the real economy: increased AI capabilities lead to reduced wage spending, resulting in weak spending, tighter profit margins, and companies buying more capabilities, further enhancing capabilities. Then it shifts to the financial sector: damaged incomes impact mortgages, bank losses tighten credit, the wealth effect collapses, and the feedback loop accelerates. Both of these are exacerbated by an inadequate policy response from a government that appears frankly, rather confused. This system was not designed for this type of crisis. The federal government's revenue base is essentially a tax on human time. People work, companies pay, and the government takes a cut. In a normal year, personal income tax and payroll tax form the backbone of fiscal revenue. As of the first quarter of this year, federal fiscal revenue was 12% lower than the Congressional Budget Office's baseline projection. Payroll tax revenue declined because fewer people were employed at previous wage levels. Income tax revenue declined because the amount of income earned was structurally lower. Productivity is soaring, but the gains are flowing to capital and computing power, not labor. Labor's share of GDP fell from 64% in 1974 to 56% in 2024, a slow decline over four decades driven by globalization, automation, and the continued erosion of workers' bargaining power. In the past four years, with AI beginning its exponential growth, that share has fallen to 46%. A record sharp decline. Output still exists. But it no longer flows through households and back to businesses, meaning it no longer flows through the IRS. The cycle is breaking down, and governments are expected to intervene to fix it. As with every recession, spending rises while income falls. But this time, the difference is that spending pressures are not cyclical. The automatic stabilizers are built for temporary unemployment, not structural replacement. The system is paying for benefits that assume workers will be reintegrated. Many will not, at least not at near-recovery wage levels. During COVID, the government freely accepted a 15% deficit, but at the time it was considered temporary. Those who need government support today are not suffering from a pandemic that will recover. They are being replaced by a continuously improving technology. The government needs to transfer more money to households at a time when it's collecting less tax from them. The US won't default. It prints money to spend and repays borrowers in the same currency. But this pressure is already showing elsewhere. Municipal bonds have shown worrying signs of divergence in their year-to-date performance. States without income taxes are doing relatively well, but general obligation municipal bonds issued by states that rely on income taxes (primarily blue states) are beginning to reflect some default risk. Politicians quickly realized this, and the debate over who should receive a bailout has unfolded along partisan lines. The government, to its credit, recognized the structural nature of the crisis early on and began considering a bipartisan proposal they called the “Transition Economy Act”: a framework for direct transfer payments to displaced workers, funded through a combination of deficit spending and a proposed AI inference computing tax. The most radical proposal on the table goes even further. The “Shared AI Prosperity Act” would establish a public interest claim on returns to the intelligent infrastructure itself, somewhere between sovereign wealth funds and royalties from AI-generated output, with dividends used to fund household transfer payments. Private sector lobbyists have issued numerous warnings in the media about the slippery slope. The politics behind the discussions have been predictably predictable, exacerbated by sensationalism and brinkmanship. The right wing cites transfer payments and redistribution as Marxism and warns that taxing calculations would hand over a leading position to China. The left warns that taxes drafted with the help of existing businesses would become another form of regulatory capture. Fiscal hawks point to unsustainable deficits. Doves cite premature austerity measures following the global financial crisis as a warning. With the presidential election approaching this year, these divisions will only be amplified. While politicians bicker, societal structures disintegrate faster than the legislative process. The “Occupy Silicon Valley” movement is a symbol of broader discontent. Last month, protesters blocked the entrances to Anthropic and OpenAI’s San Francisco offices for three weeks. Their numbers are increasing, and the media coverage of the demonstrations exceeds the unemployment figures that sparked them. It's hard to imagine the public hating anyone as much as they hated bankers after the global financial crisis, but AI labs are trying to break that record. And, from the public's perspective, the reasons are compelling. The speed at which their founders and early investors have accumulated wealth makes even the Gilded Age seem mild. The benefits of the productivity boom have flowed almost entirely to the owners of computing power and the shareholders of the labs running it, pushing inequality in the US to unprecedented levels. Everyone has their villains, but the real villain is time. AI capabilities are developing faster than institutions can adapt. Policy responses are moving at the speed of ideology, not the speed of reality. If governments don't quickly agree on what the problem is, the feedback loop will write the next chapter for them. The Fading Premium of Intelligence Throughout modern economic history, human intelligence has been a scarce input. Capital is abundant (or at least replicable). Natural resources are finite but substitutable. Technological improvements are slow enough for humans to adapt. Intelligence—the ability to analyze, decide, create, persuade, and coordinate—is the only thing that cannot be replicated on a large scale. Human intelligence commands its inherent premium due to its scarcity. Every institution in our economy, from the labor market to the mortgage market to tax laws, is designed for this hypothetical world. We are experiencing the fading of this premium. Machine intelligence is now a qualified and rapidly improving alternative to human intelligence on an increasing number of tasks. The financial system, optimized for a world where human thought is scarce for decades, is being repriced. This repricing is painful, disorderly, and far from complete. But repricing is not collapse. The economy can find a new equilibrium. Reaching it is one of the few tasks that only humans can accomplish. We need to do it right. For the first time in history, the most productive assets in the economy are creating fewer jobs, not more. No framework works because none of them are designed for a world where scarce inputs become abundant. So we must build new frameworks. The only important question is whether we can build them in time. But you're not reading this in June 2028. You're reading it in February 2026. The S&P 500 is near its all-time high. The negative feedback loop hasn't even started yet. We're confident that some of these scenarios won't materialize. We're equally confident that machine intelligence will continue to accelerate. The premium for human intelligence will shrink. As investors, we still have time to assess how much of our portfolio is built on assumptions that won't last until the end of this decade. As a society, we still have time to act proactively. The canary is still alive. (Note by Kou: The canary is a metaphor. In the past, miners would bring canaries into the mines. Canaries were very sensitive to toxic gases, and their death indicated that the dangerous gases in the mine had reached a lethal concentration, requiring the miners to evacuate immediately. The author means that although the "smart crisis" risk described in the article does exist (like the presence of methane gas in a mine), the warning "canary" is still alive. That is, readers are fortunate to read this article in February 2026; the crisis has not yet reached an irreversible point, and we still have time and opportunity to take action to avoid the worst consequences. This sentence echoes the previous paragraph: "We still have time to proactively respond," conveying a message of \*\*urgent warning rather than despair\*\*: the alarm has been sounded, but the window of opportunity is not yet closed.)

## Related News & Research

- [One in seven Americans are ready for an AI boss, but they might not trust it](https://longbridge.com/en/news/281357882.md)
- [Fed's Powell: AI is making people more productive](https://longbridge.com/en/news/281048143.md)
- [Is an oil shock enough to send the US into recession? Exxon Mobil’s chief economist answers.](https://longbridge.com/en/news/281616234.md)
- [Economist: AI bubble has popped, but a rarer AI bubble grows](https://longbridge.com/en/news/281032131.md)
- [08:39 ETReclaim Health Wins 2026 Artificial Intelligence Excellence Award](https://longbridge.com/en/news/281525856.md)