Series: AI Singularity — Part 14 — February 2026

The Losers: Obsolescence Trades

Capitalism is creative destruction. AI is about to destroy a lot of capital. If your business is "middleman arbitrage", your stock is going to zero.

Business Model Extinction Call Centers & BPO Education & Media Short Selling
AI Singularity14/15

Section 1: The Innovator's Dilemma

Framework: Clayton Christensen's Innovator's Dilemma

In 1997, Harvard professor Clayton Christensen published The Innovator's Dilemma, explaining why well-managed companies fail. His central thesis: incumbents are rational to ignore disruptive technologies because those technologies initially serve a smaller, less profitable market. By the time the new technology improves enough to threaten the core business, it is too late. The incumbent's organizational structure, incentive systems, and customer relationships are optimized for the old world.

The classic pattern: a disruptor enters the low end of the market with an inferior but cheaper product. Incumbents dismiss it as a toy. The disruptor improves rapidly. One day, the "toy" is good enough for 80% of use cases at 10% of the cost. The incumbent's premium product becomes irrelevant.

Why AI disruption is faster than any before it: Previous disruptions were hardware-based (film to digital, physical stores to e-commerce). Hardware disruptions take 10-15 years because atoms are slow. AI disruption is software-based. Software scales at the speed of deployment. Once an AI model is trained, the marginal cost of serving one more customer is approximately zero. There are no factories to build, no supply chains to establish. The disruption cycle compresses from a decade to 2-3 years.

Every great disruption follows the same arc: denial, negotiation, capitulation. Here is how history repeated itself.

Incumbent Disruptor Peak Valuation Time to Obsolescence Key Mistake
Kodak Digital cameras / Smartphones $31B (1997) ~15 years Invented the digital camera in 1975 but shelved it to protect film revenue
Blockbuster Netflix (DVD-by-mail, then streaming) $5B (2004) ~10 years Turned down acquisition of Netflix for $50M in 2000; protected late-fee revenue
Nokia Apple iPhone / Android $250B (2007) ~6 years Symbian OS was "good enough"; dismissed capacitive touch as a gimmick
Blackberry Apple iPhone / Android $83B (2008) ~5 years Believed enterprise security moat was unbreachable; keyboard loyalty
Yellow Pages Google Search $17B (2006) ~8 years Print advertising model had no digital equivalent; structural decline
Encyclopedia Britannica Wikipedia $650M (1990) ~12 years "Free" seemed impossible; underestimated crowd-sourced quality
BPO / IT Services AI Agents (2024+) $250B sector ~3-5 years (est.) Software disruption is 3x faster than hardware disruption

Notice the pattern: the time to obsolescence has been compressing. Kodak had 15 years. Nokia had 6. BPO companies may have 3-5. When the disruptor is software and the incumbent sells labor hours, the collapse is near-instantaneous once the technology crosses the "good enough" threshold.

Section 2: The BPO / Outsourcing Collapse

The $250 billion IT services and business process outsourcing industry is built on a simple premise: it is cheaper to hire engineers in Bangalore than in Boston. For three decades, this labor arbitrage generated enormous value for companies like Accenture, Infosys, Wipro, and Cognizant. But AI does not live in Bangalore or Boston. AI lives everywhere, instantly, for a fraction of a cent per transaction.

Case Study: Klarna
In February 2024, Klarna announced that its AI assistant, powered by OpenAI, was handling two-thirds of all customer service chats within its first month of deployment. That is the equivalent of 700 full-time customer service agents. Resolution time dropped from 11 minutes to 2 minutes. Customer satisfaction scores remained identical. Klarna estimated $40 million in annualized savings. One fintech company. One deployment. 700 jobs eliminated overnight.

The BPO revenue model works like this: charge clients $25-50/hour for offshore labor that costs $8-15/hour. The margin is in the spread. But when an AI agent handles the same task for $0.01-0.10 per interaction, the entire spread evaporates. There is no hourly rate to arbitrage when the marginal cost approaches zero.

Revenue Growth Deceleration

The leading IT services companies have seen their organic revenue growth collapse from double digits to low single digits, and the trend is accelerating.

The "Body Shop" Business Model

IT services companies are sometimes called "body shops" because their primary product is human labor, measured in billable hours or "FTEs" (full-time equivalents). Accenture employs over 700,000 people. Infosys employs 315,000. Their revenue is approximately (number of employees) x (average billing rate) x (utilization rate). When AI can perform the same tasks, the formula breaks: the "number of employees" variable collapses, and with it, revenue. These companies cannot pivot to selling AI because their margins depend on the labor spread, which AI eliminates. Accenture charging $200/hour for work that GPT-5 does for $0.05 is not a sustainable business model.

The Timeline of Collapse

Phase 1 (2024-2025): Hiring freezes. Attrition not replaced. "AI-augmented delivery" marketing. Revenue growth stalls to low single digits.

Phase 2 (2026-2027): Clients begin insourcing with AI tools. Contract renewals come at 30-50% lower rates. Headcount declines 15-25%. Margin compression forces restructuring charges.

Phase 3 (2028-2030): Structural decline. Only companies with proprietary data or deep domain expertise survive. Pure-play BPO companies face existential risk. The $250B industry contracts to $100-150B.

Section 3: The Staffing Agencies

The temporary staffing industry generates $500 billion globally by placing workers in roles their clients cannot or will not fill permanently. The model: recruit candidates, match them to job openings, charge the client a markup (typically 25-75% over the worker's pay rate). For decades, this was a stable, counter-cyclical business. Economic downturns increased demand for flexible labor. Growth periods increased demand for specialized skills.

AI breaks this model in two ways. First, white-collar temporary placements are the most vulnerable. Data entry, bookkeeping, administrative support, junior financial analysis, paralegal work, and basic software testing can all be handled by AI agents at near-zero marginal cost. Second, AI is also disrupting the matching function itself. The core value-add of a staffing company (knowing which candidate fits which role) is a pattern-matching problem that AI solves better than any human recruiter.

Company Ticker Peak Rev ($B) LTM Rev ($B) Decline % White-Collar Exposure AI Risk
Robert Half RHI $7.2 (2022) $5.1 -29% ~85% (finance, tech, legal) Critical
ManpowerGroup MAN $20.7 (2022) $18.2 -12% ~45% (mixed with industrial) High
Hays plc HAS.L £7.1 (2023) £5.7 -20% ~70% (tech, finance, HR) High
Adecco Group ADEN.SW CHF 23.7 (2022) CHF 21.8 -8% ~40% (broad mix) Medium-High
Kforce KFRC $1.7 (2022) $1.3 -24% ~90% (tech, finance) Critical

Robert Half is the most exposed. Nearly 85% of their placements are white-collar: accountants, financial analysts, IT professionals, and legal staff. These are precisely the roles where AI is already demonstrating competence. The company's revenue has declined from $7.2B peak to $5.1B, and the decline is accelerating. Their Protiviti consulting arm provides some cushion, but the core staffing business is in structural decline.

The paradox: staffing companies are trying to sell "AI staffing solutions" to their clients, but this is like Blockbuster trying to sell streaming. The product they are selling (human labor) is the very thing being displaced. You cannot sell the cure when you are the disease.

Section 4: Education Disruption Victims

Education technology was supposed to be the future. Instead, AI has turned several "edtech" companies into cautionary tales. The core problem: if students can get personalized tutoring, instant homework help, and comprehensive study materials from an AI for free, why would they pay $15-20/month for Chegg?

Chegg: The Canary in the Coal Mine

Chegg (CHGG) is the single most dramatic example of AI disruption in public markets. The stock peaked at $115 in February 2021 with a market cap of $14.5 billion. When ChatGPT launched in November 2022, Chegg's management initially dismissed it. By May 2023, they acknowledged AI was hurting subscriber growth. The stock cratered. Today, Chegg trades below $2, a decline of over 98% from its peak.

What happened: Chegg's core product was a homework answer database. Students paid monthly subscriptions to access step-by-step solutions. ChatGPT provided the same service (and often better, with explanations) for free. The business model disintegrated in real time. Subscribers fell from 8.2 million at peak to under 4 million. Revenue declined 20%+ year-over-year. The company announced layoffs of 23% of its workforce.

The Broader Casualties

Section 5: Media & Content Losers

The creative economy is experiencing a tectonic shift. AI can now generate images (Midjourney, DALL-E, Stable Diffusion), write copy (Claude, GPT), compose music (Suno, Udio), and produce video (Sora, Runway). The immediate casualties are companies whose business model is aggregating and selling human-created content at scale.

Company Ticker Business Peak Price Current Decline Core Threat
Getty Images GETY Stock photography $6.30 (2023) ~$2.50 -60% AI image generation eliminates need for stock photos
Shutterstock SSTK Stock photography $175 (2021) ~$25 -86% Partnered with OpenAI but core library usage declining
Fiverr FVRR Freelance marketplace $336 (2021) ~$25 -93% Logo design, copywriting, translation gigs automated
Upwork UPWK Freelance marketplace $64 (2021) ~$12 -81% Entry-level freelance work displaced by AI tools
Gannett GCI Print/digital media $18 (2019) ~$4 -78% Local news summarized by AI; ad revenue migrating

Stock photography is the clearest casualty. Getty Images and Shutterstock sell licenses to photographs taken by humans. But when a designer can type "professional team meeting in modern office, diverse, natural light, 8K" into Midjourney and get a perfect image in 30 seconds for $0.05, there is no reason to pay $500 for a Getty license. Getty's pivot to offering its own AI generator is a Hail Mary: they are trying to sell the weapon that is killing them.

Freelance platforms are bifurcating. Fiverr and Upwork are experiencing a collapse in demand for low-end creative services (logo design, basic copywriting, data entry, simple web development). However, demand for high-end, specialized freelance work (senior software architecture, complex data science, UX research) is holding steady or growing. The platforms survive only if they can successfully pivot upmarket, which requires a fundamentally different value proposition.

Print media acceleration. Newspapers were already in structural decline. AI accelerates the timeline by (a) making it trivial to summarize and aggregate news, reducing the value of individual articles, and (b) absorbing advertising budgets through AI-powered programmatic optimization that bypasses traditional publishers entirely.

Section 6: The "Zero" List

These are stocks where the terminal value is zero or near-zero. Not all will go bankrupt; some may be acquired at distressed valuations. But the equity is effectively worthless on a 3-5 year horizon. This is not hyperbole. It is math. When your revenue model depends on selling something that AI produces for free, the terminal value of that revenue stream is zero.

Ticker Company Market Cap AI Threat Level Timeline Short Thesis Conviction
CHGG Chegg ~$250M Terminal 12-24 months ChatGPT is a free, better Chegg. Subscribers in freefall. Cash burn accelerating. No pivot path. Very High
TTEC TTEC Holdings ~$200M Terminal 18-36 months 60,000 call center agents. Voice AI (ElevenLabs, Bland AI) does this 24/7 for 1/100th cost. Klarna precedent. Very High
LZ LegalZoom ~$1.3B Severe 24-48 months Legal document automation is trivial for GPT-class models. LLC formation, contracts, wills: all commoditized. High
GETY Getty Images ~$700M Severe 24-48 months AI image generation is better, faster, cheaper than licensing. Library value depreciates daily. High
WIT Wipro ~$28B High 36-60 months Pure-play BPO with lowest margin in IT services. Revenue already declining YoY. 230,000 employees at risk. Medium-High
TASK TaskUs ~$1.5B High 24-36 months Content moderation and customer support outsourcing. Both are AI-first domains now. Medium-High
FVRR Fiverr ~$900M High 24-48 months Low-end gig economy for creative work. AI handles logos, copy, translation, basic dev for pennies. Medium
Important Caveat: "Going to zero" does not mean these companies disappear tomorrow. Many have cash on their balance sheets, real estate, or other assets that provide a floor. But the equity value (market cap minus debt and obligations) trends toward zero as revenue declines and cash is consumed. Some may be acquired at distressed valuations by competitors or private equity. The point is: do not hold these names long.

Section 7: How to Short Disruption

Concept: The Mechanics of Short Selling

Short selling is the practice of borrowing shares from a broker, selling them immediately at the current price, and buying them back later (hopefully at a lower price) to return to the broker. Your profit is the difference between the sale price and the buyback price, minus borrowing costs.

Example: You borrow 100 shares of CHGG at $5.00 and sell them for $500. The stock drops to $1.00. You buy 100 shares for $100 and return them to the broker. Your profit: $400 (minus fees and borrow costs).

The danger: When you buy a stock (go long), your maximum loss is 100% (the stock goes to zero). When you short a stock, your maximum loss is theoretically infinite because the stock can rise without limit. If you short at $5 and it goes to $50, you've lost 900%. This asymmetry makes position sizing critical.

Alternatives to direct shorts: (1) Put options: buy the right to sell at a specific price. Maximum loss is limited to the premium paid. (2) Bear put spreads: buy a put and sell a lower-strike put to reduce cost. (3) Inverse ETFs: funds that go up when a sector goes down. Each has different risk/reward profiles.

Instrument Comparison

Direct Short
Unlimited Risk
Requires margin. Borrow costs on hard-to-borrow names (CHGG: 15-25% annualized). Best for high conviction with tight stops.
Put Options
Defined Risk
Max loss = premium paid. Time decay works against you. Best for 3-6 month thesis. Buy slightly ITM for higher delta.
Bear Put Spread
Defined Risk + Lower Cost
Buy a put, sell a lower-strike put. Caps upside but cuts premium cost by 40-60%. Best for gradual declines.
Pair Trade
Market Neutral
Long AI winner + Short AI loser. Hedges market risk. Example: Long MSFT / Short CHGG. Profit from relative performance.

Position Sizing Rules

Short positions must be smaller than long positions. The asymmetric risk profile demands discipline.

Hypothetical Pair Trade Basket Performance

Hypothetical illustration only. Past performance does not predict future results. Short selling involves unlimited risk.

Section 8: Risk Analysis

Concept: Why Shorts Are Harder Than Longs

Legendary investor Charlie Munger once said: "Being short and seeing a promoter take the stock up is very irritating. It's not worth it to have that much irritation in your life." Shorting is psychologically and structurally harder than going long for several reasons:

1. Asymmetric payoff: Longs can gain infinity, lose 100%. Shorts can gain 100% (stock to zero), lose infinity. The math is against you.

2. Timing is everything: You can be right about the thesis and still lose money if the stock rises before it falls. "The market can stay irrational longer than you can stay solvent." (Keynes)

3. Borrow costs: You pay interest to borrow shares. Hard-to-borrow stocks can cost 20-50% annualized. This creates a constant drag.

4. Short squeezes: If a heavily shorted stock rises rapidly, shorts are forced to buy (cover), which pushes the price higher, forcing more covering. GameStop (GME) in January 2021 is the canonical example: the stock went from $20 to $483 in two weeks, bankrupting several short sellers.

Key Risks to the Obsolescence Thesis

Short Squeeze Risk

Several names on the "Zero" list have short interest exceeding 20% of float (CHGG: ~35%, GETY: ~18%). A positive earnings surprise, acquisition rumor, or retail-trader rally can trigger violent squeezes. Mitigation: use puts instead of direct shorts for the most crowded names. Defined-risk structures limit damage.

Regulatory Protection

Governments may intervene to protect industries from AI disruption. France already mandates human customer service for certain financial products. The EU AI Act restricts autonomous decision-making. India may protect its $200B IT services export industry. If regulation slows AI adoption, the timeline extends, and shorts bleed on borrow costs.

Slower Adoption Than Expected

Enterprise AI adoption is slower than consumer adoption. Large companies have procurement cycles, security reviews, integration requirements, and change management processes that take 12-24 months. If adoption stalls at the "pilot phase" without scaling, BPO companies retain revenue longer than expected. The thesis is right but the timing is wrong, which in markets, is the same as being wrong.

Incumbent Pivot

Some incumbents successfully transform. Accenture has invested $3 billion in AI capabilities and repositioned as an "AI transformation partner." If incumbents become the AI delivery vehicle for their clients (rather than the body being displaced), the short thesis weakens. Watch for: rising AI revenue as % of total, successful case studies, and improving margins despite headcount reductions.

Short Interest Landscape

Names above 20% short interest carry elevated squeeze risk. Prefer puts or spreads over direct shorts for these names.

Section 9: Validation, Invalidation & Catalysts

Thesis Validation Signals

  • BPO industry revenue declining >15% YoY at sector level
  • Staffing agencies (RHI, MAN) reporting accelerating volume decline in white-collar placements
  • Major enterprises publicly announcing AI replacement of outsourced functions (more Klarna-type announcements)
  • Legacy SaaS vendors losing >20% of customers to AI-native alternatives
  • Stock photography download volumes declining >30% YoY
  • Freelance platform GMV (Fiverr, Upwork) declining in core categories

Thesis Invalidation Signals

  • Incumbents successfully pivoting to AI: rising AI revenue as % of total revenue, improving margins
  • Regulatory protection: EU, India, or US mandate human involvement for outsourced services
  • AI adoption plateaus: enterprise AI spending growth slows to <10% YoY
  • Short squeeze in multiple names simultaneously, forcing portfolio-wide covering
  • AI quality stalls: models fail to improve in complex, multi-step tasks (customer service, legal, financial analysis)
  • Acquisition wave: private equity or strategic buyers acquire targets at significant premiums to current prices

Upcoming Catalysts

Q1 2026
Earnings Season
CHGG, RHI, ACN, INFY report Q1 results. Watch for: subscriber churn (CHGG), placement volumes (RHI), organic growth (ACN/INFY). Guidance downgrades are the trigger.
H1 2026
OpenAI Enterprise Launch
GPT-5 enterprise rollout will unlock new waves of automation. Each Fortune 500 deployment = fewer outsourced contracts renewed.
2026-2027
AI Agent Proliferation
Anthropic, Google, Microsoft shipping autonomous agents that handle multi-step workflows. Customer support, data entry, document review: all agent-first by 2027.
Ongoing
Contract Renewals
BPO contracts typically run 3-5 years. The 2021-2022 vintage of contracts comes up for renewal in 2025-2027. This is when the revenue cliff materializes for IT services companies.

The Takeaway

Shorting disruption victims is one of the most reliable strategies in markets, but it requires patience, discipline, and risk management. The history of Kodak, Blockbuster, and Nokia proves that incumbents can decline for years before the final capitulation. The key is to be right on the direction and survive the volatility. Use defined-risk instruments (puts, spreads), size positions conservatively (max 2-3% per name), and pair shorts with long positions in AI winners to create a market-neutral portfolio with positive expected value. The disruption is inevitable. The timing is uncertain. Plan accordingly.

Disclaimer: This analysis is for educational and informational purposes only. It does not constitute investment advice, a recommendation, or a solicitation to buy or sell any securities. Short selling involves unlimited risk. Options involve risk and are not suitable for all investors. Past performance does not guarantee future results. Always consult a qualified financial advisor before making investment decisions. The author may hold positions in the securities discussed.

Part 13: The Geopolitical AI Race Series Index Part 15: Building the AI Disruption Portfolio

Back to Market Watch  ·  AI Singularity Series  ·  February 2026

AI Singularity14/15