Data Centers Above, Discount Aisles Below
The same AI boom can reward infrastructure owners, affluent consumers, and the retailers helping pressured households trade down.
Picture two households opening their phones on the same Friday night.
In the first, a professional watches an AI assistant finish the first draft of a task that once consumed an afternoon. Her company is spending heavily on automation, but her own role is becoming more valuable because she knows how to direct the system, catch its mistakes, and turn its output into a decision. Her retirement account is rising with the same semiconductor and cloud stocks powering the tool. A summer cruise that looked indulgent a year ago now feels affordable.
Across town, another household opens a different set of apps. A recent graduate has sent out dozens of applications for jobs that used to train people into a career: junior analyst, entry-level developer, marketing assistant, support specialist. Interviews are scarce. Rent is due. The grocery order moves from a familiar brand to a private label. A planned clothing purchase becomes a trip to an off-price store. The family is still spending, but every dollar has been given a harder job.
Both households appear inside the same consumer-spending number. Both live under the same unemployment rate. Both contribute to an economy that may still be growing.
Their private economies are moving in opposite directions.
That is the K-shaped economy. The upper arm compounds through asset ownership, scarce skills, business scale, and access to the technologies raising productivity. The lower arm bends under weaker bargaining power, fragile entry-level work, expensive necessities, and debt. Artificial intelligence did not create America’s fault lines in wealth, housing, education, or market access. It is beginning to press on all of them at once.
The investment story follows from that split. The companies selling the machinery of AI can thrive. So can the platforms serving households enriched by the boom. At the same time, merchants that help pressured consumers stretch a paycheck can gain traffic and market share.
One economy can produce record demand for data-center cooling and stronger demand for closeout merchandise. One stock market can support premium cruises while the same country produces more private-label grocery baskets.
The K is already becoming visible. Investors can own both arms.
The first rung of the career ladder is shaking
A K-shaped economy rarely announces itself with a recession siren. It appears first as a contradiction.
The June 2026 employment report looked subdued rather than disastrous. The United States added 57,000 payroll jobs. Unemployment held at 4.2%. Average monthly payroll growth over the preceding year was only 36,000, labor-force participation slipped to 61.5%, and April and May gains were revised down by a combined 74,000 jobs.[1]
Those numbers describe a labor market that is cooling. They do not describe what it feels like to be trying to enter one of the occupations AI can already touch.
The New York Fed reported that unemployment among recent college graduates aged 22 to 27 reached 5.6% in March 2026, up from 3.6% in March 2019.[2] Stanford Digital Economy Lab researchers then found something more specific in ADP payroll data: workers aged 22 to 25 in the most AI-exposed occupations experienced a 16% relative employment decline after the researchers controlled for firm-level shocks. Experienced workers in the same occupations remained comparatively stable. The damage was concentrated where AI looked automative; in work where AI augmented people, the same decline did not appear.[3]
This distinction may become one of the defining economic lines of the next decade.
An experienced lawyer can ask a model for a first-pass analysis and recognize the missing precedent. A senior engineer can use generated code and see the architectural flaw. A seasoned marketer can turn a synthetic draft into a campaign. Judgment makes the machine useful.
A junior employee was often paid to produce that first pass while learning how experts think. If software now produces the draft, summary, research memo, basic code, or scripted support response, the company can demand experience before it has created enough places for people to acquire it.
The ladder does not disappear all at once. The first few rungs become farther apart.
The evidence still calls for restraint. New York Fed research using Lightcast job postings found little indication that AI exposure alone caused a distinct fall in vacancies after ChatGPT’s release.[4] High interest rates, the post-pandemic technology correction, remote work, and slower white-collar hiring also shaped the market. AI is an accelerant, not a complete explanation.
Yet the direction matters. A technology can transform bargaining power before it produces mass unemployment. Companies need only discover that one experienced worker with AI can supervise work previously distributed across several junior seats.
That is how a productivity story becomes a distribution story.
The machine rewards the people who can afford to build it
AI adoption has moved beyond demos. Census Bureau survey data from late 2025 and early 2026 found that 18% of U.S. firms used AI in at least one business function. Larger companies adopted it more aggressively, so the employment-weighted adoption rate reached 32%.[5]
The gap between those two figures reveals the corporate version of the K.
A large enterprise can pay for models, cloud capacity, security, data preparation, consultants, workflow redesign, and employee training. It can spread those fixed costs over millions of transactions. It may even own proprietary data that makes a general model more useful inside its business.
A smaller company often buys the finished capability as a subscription. It receives some productivity gain, but the infrastructure margin flows elsewhere—to the chip designer, the networking vendor, the cloud platform, the power-and-cooling supplier, or the software company controlling distribution.
Aggregate productivity data fit this capital-deepening story, even though they cannot isolate AI as the cause. Nonfarm business labor productivity rose 2.8% year over year in the first quarter of 2026, while unit labor costs increased only 0.5% over four quarters. For 2025 as a whole, labor productivity rose 2.2%, including a 0.9-percentage-point contribution from capital intensity.[6]
Output can grow faster than labor hours when companies invest in computing, software, automation, and redesigned workflows. The first claim on those gains usually belongs to whoever owns the scarce capital.
This is the industrial logic explored in The Next Architecture of Intelligence. The first AI trade was scale: more accelerators, more networking, more data centers, more electricity, more cooling. Even if the next architecture eventually makes intelligence cheaper, today’s deployment wave remains a resource race.
That race pulls the upper arm of the K higher.
The stock market turns technology gains into household divergence
The ownership of the AI boom is even more concentrated than its use.
Federal Reserve distributional data show that in the first quarter of 2026, the wealthiest 1% owned 50.2% of corporate equities and mutual-fund shares. The next 9% owned another 37.2%. Together, the top 10% by wealth controlled 87.4% of the equity pool. The bottom half owned 1.1%.[7]
When the market assigns hundreds of billions of dollars of value to AI infrastructure and platforms, the gains do not arrive evenly. They land first in portfolios that already hold the most financial assets.
That creates a loop. AI expectations raise the value of the companies expected to build or monetize the technology. Rising portfolios make affluent households feel more secure. Those households keep investing and spend more freely on travel, convenience, services, and experiences. Their spending supports corporate earnings, which can reinforce asset prices.
New York Fed researchers found that since 2023, real net worth for the top income percentile grew more than 25%, while growth for the middle 40% remained below 10%. They also found that lower-income households consistently faced higher inflation than middle- and higher-income households beginning in late 2022.[8]
The same economy was therefore delivering faster wealth growth to people who owned financial assets and a harsher inflation mix to people whose budgets were dominated by necessities.
This extends the argument in The Rich-Person Password. Access determines who compounds with innovation. In private markets, wealth gates can reserve early exposure for accredited capital. In public markets, access is formally broader, but ownership remains so concentrated that a technology-led rally still produces a sharply unequal wealth effect.
AI becomes more than a software story at that point. It becomes a mechanism for distributing claims on future income.
Two consumers now walk through the same economy
By early 2026, the divergence had begun to show up at the cash register.
Bank of America’s aggregated card and deposit data found that January card spending rose 2.5% year over year for higher-income households, compared with 1.0% for middle-income households and 0.3% for lower-income households. After-tax wage growth showed an even wider split: 3.7% for the higher-income cohort, just under 1.6% for the middle, and 0.9% for the lower.[9]
A Moody’s Analytics estimate put the top 10% of earners’ share of consumption at 49.2% in the second quarter of 2025.[10] That estimate is debated, and other methods produce a lower share. The precise number matters less than the direction supported across several datasets: affluent households have supplied a disproportionate share of incremental spending, while lower-income households have become increasingly price-sensitive.
Debt narrows the choices available to the lower arm. New York Fed data show $18.8 trillion in household debt in the first quarter of 2026, including $1.25 trillion in credit-card balances and $1.69 trillion in auto loans. Some 4.8% of outstanding debt was in delinquency, while the annualized flow into early credit-card delinquency remained elevated at 8.6%.[11]
This does not mean the pressured household stops consuming. Life rarely permits that. Food, school supplies, soap, work clothes, and transportation still have to be purchased.
The household changes stores. It changes brands. It waits for a deal. It buys a smaller pack, accepts yesterday’s fashion at a discount, consolidates shopping trips, or pays for a membership that promises lower unit costs.
That is why a K-shaped economy can create two profit pools instead of one winner and one wasteland.
A previous Signal Before Consensus piece, Why Investors Should Reconsider the “Avocado Toast” Generation, argued that Millennials should be understood through economic pathways rather than as one average consumer. The K-shaped framework pushes that idea further. Even within the same generation, one household may be an AI-using manager with rising assets while another is a renter juggling debt and a more fragile career ladder. Age no longer tells investors enough. Ownership, income resilience, and bargaining power do.
Follow the money up the K
The upper arm begins in a data center.
Before an AI assistant can save an employee an hour, a chain of companies has to turn electricity into computation. Nvidia (NVDA) supplies the leading accelerated-computing platform for training and inference. Fiscal 2026 revenue reached $215.9 billion, up 65%.[12] The risk is equally large: hyperscaler spending can slow, custom silicon can improve, export controls can tighten, and a valuation built around scarcity can compress when supply catches up.
The clusters also need alternatives and connective tissue. Broadcom (AVGO) combines custom AI accelerators with networking that helps the largest cloud customers optimize their own systems. Its Q2 fiscal 2026 AI semiconductor revenue reached $10.8 billion, up 143% year over year.[13] That growth comes with customer concentration and lumpy program timing.
Arista Networks (ANET) supplies the high-speed Ethernet switching and software that allow thousands of machines to operate as one cluster. First-quarter 2026 revenue rose 35.1% year over year.[14] Its opportunity sits inside the traffic explosion; its risk sits in dependence on a small number of very large customers and competition from integrated vendors.
Then the computation becomes a heat and power problem. Vertiv (VRT) supplies power conversion, thermal management, and service for dense data-center infrastructure. First-quarter 2026 organic sales in the Americas rose 44% on data-center demand.[15] Vertiv is the physical reminder that “the cloud” is a building full of electrical constraints. Project timing, capacity expansion, cyclicality, and valuation can still punish investors if the buildout outruns demand.
Amazon (AMZN) sits above this machinery and inside it. AWS sells compute, custom chips, models, and AI services; Amazon’s retail and advertising operations can use automation at enormous scale. AWS first-quarter 2026 sales grew 28% to $37.6 billion.[16] The company’s planned 2026 capital spending of roughly $200 billion raises the central question of the upper arm: how much future return is already embedded in today’s investment?
The wealth created around this buildout produces a second set of beneficiaries.
Interactive Brokers (IBKR) is a technology-led toll collector on global market participation. Client equity reached $789.4 billion in the first quarter of 2026, up 38% year over year.[17] Rising asset values and activity can support the platform, while market declines, lower interest income, or subdued trading can reverse the effect.
Royal Caribbean (RCL) captures the experiential side of affluent resilience. First-quarter 2026 adjusted EBITDA rose 21%, bookings carried record prices, and onboard spending exceeded the prior year.[18] A cruise ship may look far removed from an AI rack, yet the financial connection runs through the household balance sheet. When portfolios rise and high-income wages remain strong, premium experiences retain pricing power. Fuel, leverage, geopolitics, travel disruption, and a market correction remain the obvious breaks in that chain.
These companies occupy different points along one river of money: capital spending flows into NVDA, AVGO, ANET, and VRT; cloud and platform economics accrue to AMZN; financial wealth and affluent spending can reach IBKR and RCL.
Follow the paycheck down the K
The lower arm begins with a different question: who can make a constrained dollar feel larger?
Walmart (WMT) has the broadest answer. Its purchasing scale supports price leadership, while delivery, marketplace, membership, and advertising improve the profit mix. Fiscal 2026 revenue rose 4.7% to $713.2 billion, and global advertising grew 46%.[19] Walmart can attract a household trading down on groceries and another household paying for fast delivery. Tariffs, wages, food disinflation, and the difficulty of protecting margins during a price war remain real risks.
Off-price retail turns somebody else’s forecasting error into the customer’s bargain.
TJX Companies (TJX) buys branded excess inventory and sells it at discounts commonly advertised at 20% to 60%. Fiscal 2026 comparable sales rose 5%, with every division growing at least 4%.[20] Ross Stores (ROST) offers a similar treasure-hunt model with a lower-income customer skew; first-quarter 2026 comparable-store sales rose 17%.[21]
Their appeal grows when the shopper still wants a recognizable brand but refuses the full-price channel. Their vulnerability appears when attractive inventory becomes scarce, freight costs rise, fashion bets miss, or pressure on the customer becomes severe enough to reduce traffic altogether.
Ollie’s Bargain Outlet (OLLI) pushes the same logic into closeouts and “extreme value.” First-quarter fiscal 2026 net sales rose 14%.[22] Its store expansion adds a growth engine, while inconsistent deal flow and execution risk can make the model uneven.
Dollar General (DG) is the more fragile expression of the lower arm. Its small rural stores and heavy consumables mix matter when transportation, time, and budgets are tight. First-quarter 2026 traffic rose 1.4%, operating profit rose 10.8%, and same-store sales rose 2.0%.[23] Yet the company serves a customer with very little room left to trade down. Shrink, labor costs, tariffs, and a shrinking basket can overwhelm thin margins. Consumer stress can create traffic and destroy profit in the same quarter.
That distinction is essential. A weak consumer does not automatically produce a strong discount retailer. The durable winners have purchasing power, inventory discipline, useful locations, and enough margin flexibility to share savings with the customer without surrendering the economics of the business.
A few companies can stand in the middle
Some businesses do not need to choose one household.
Costco (COST) sells thrift in a format affluent consumers enjoy. The membership fee converts loyalty into recurring, high-margin income. Bulk economics appeal to value seekers; product quality, convenience, and the treasure-hunt assortment keep wealthier households engaged. Fiscal third-quarter 2026 net sales rose 11.6%, adjusted U.S. comparable sales rose 6.8%, and digitally enabled comparable sales rose 20.8%.[24]
Walmart increasingly spans the K for similar reasons. Food and low prices anchor the lower arm. Delivery, Walmart+, marketplace breadth, advertising, and Sam’s Club reach households higher up the income ladder.
Amazon also crosses the split. AWS and advertising sit firmly on the upper arm, while retail selection, logistics, and price comparison serve cost-conscious consumers. Its enormous capital commitment makes the stock a more direct wager on the upper arm, but the consumer platform still sees both households.
These bridge companies may be especially valuable when the macro story is right but the timing is unclear. They can gain from divergence without requiring an investor to predict which side accelerates first.
Build the barbell before the slogan becomes consensus
The cleanest portfolio expression is a barbell rather than a heroic bet on one macro outcome.
On one end sit the scarce inputs and scalable platforms of AI: NVDA, AVGO, ANET, VRT, and AMZN. They benefit while companies continue buying the capacity required to automate and augment work.
Alongside them sit businesses exposed to rising assets and affluent resilience: IBKR and RCL.
On the other end sit the merchants of trade-down: WMT, TJX, ROST, and OLLI, with DG reserved for investors willing to accept greater operational and customer stress. COST and WMT can act as bridge holdings because they serve both the household protecting a budget and the household paying for convenience.
The construction still requires valuation discipline. A correct social observation can become a terrible stock purchase when the price assumes flawless execution.
AI infrastructure companies can suffer if hyperscalers pause capital spending, customers move toward custom chips, inference becomes dramatically more efficient, or a shortage turns into excess capacity. Premium-consumption companies can weaken quickly after a market correction because the wealth effect runs in both directions. Value retailers face tariffs, wages, shrink, and customers whose budgets may become too strained even for the cheapest discretionary purchase.
The thesis should therefore be monitored through behavior rather than repeated as a slogan. Watch hyperscaler capital-expenditure guidance and AI revenue conversion. Follow networking, power, and cooling backlogs. Track employment outcomes for young workers in exposed occupations, spending and wage growth by income cohort, credit delinquencies, value-retail traffic and margins, and premium-travel pricing.
The shape of the K will show up in those numbers before it becomes obvious in GDP.
The K can bend
The strongest objection is that the consumer split remains real enough to investigate and too contested to declare permanent.
Bank of America and New York Fed data show meaningful divergence by income. Moody’s estimates an extraordinary concentration of spending. Other transaction-based and government-derived measures suggest lower-income consumption held up better through 2025 and that the top decile’s share is well below 49%. Stripe Economics argues that the clearest K appears in corporate profits and equity returns, while the consumer split remains less conclusive.[25]
AI causation deserves the same caution. Stanford’s payroll evidence is striking, but Federal Reserve job-posting research has yet to show a broad AI-specific collapse. Interest rates, remote work, immigration, demographics, and post-pandemic normalization also shape hiring.
Diffusion could narrow the K. Cheaper AI may help small businesses compete. Productivity gains may eventually lift real wages. New occupations may absorb displaced workers. A generation that learns to use AI early may turn apparent vulnerability into an advantage.
Those possibilities do not erase the investable pattern already in front of us. The high-confidence claim is that AI currently rewards scarce capital, scale, and experienced judgment. The high-confidence consumer claim is that affluent spending remains resilient while value-seeking behavior has strengthened. The uncertain part is how quickly those forces harden into a lasting social structure.
Owning both arms is a way to invest through that uncertainty rather than pretending it does not exist.
The cash flows tell the story
Return to the two phones on Friday night.
One displays a rising brokerage balance, an AI assistant, and a travel booking. The other displays unanswered job applications, a credit-card balance, and a search for the lowest price.
The screens look unrelated. The cash flows connect them.
The data center behind the assistant sends revenue toward Nvidia, Broadcom, Arista, Vertiv, and Amazon. The portfolio gains around the boom can send assets and activity toward Interactive Brokers and discretionary spending toward Royal Caribbean. The tighter paycheck sends traffic toward Walmart, TJX, Ross, Ollie’s, Dollar General, and Costco.
This is the uncomfortable elegance of the K-shaped investment thesis: the same technological wave can strengthen the companies building the future and the merchants helping people survive its uneven arrival.
The upper arm favors NVDA, AVGO, ANET, VRT, AMZN, IBKR, and RCL. The lower arm favors WMT, TJX, ROST, OLLI, and, at higher risk, DG. COST and WMT possess the rare ability to serve both.
Investing in this split does not require celebrating it. It requires seeing where capital, wages, and household spending are moving before the average statistics make the divergence impossible to ignore.
The K is more than a letter laid over a chart. It is two American nights unfolding at once.
Sources and research notes
U.S. Bureau of Labor Statistics, The Employment Situation—June 2026.
Federal Reserve Bank of New York, New York Fed to Release Research on the Role of Remote Work in Youth Unemployment.
Stanford Digital Economy Lab, Canaries in the Coal Mine?.
Federal Reserve Bank of New York, Do Job Postings Show Early Labor-Market Effects of AI?.
U.S. Census Bureau, AI Use at U.S. Businesses.
U.S. Bureau of Labor Statistics, Productivity and Costs—First Quarter 2026 and Total Factor Productivity—2025.
Federal Reserve Distributional Financial Accounts via FRED, Q1 2026 equity ownership table.
Federal Reserve Bank of New York, Explaining the K-Shaped Economy: What’s Behind the Divide?.
Bank of America Institute, Consumer Checkpoint: Weathering the Storm.
Federal Reserve Bank of Dallas, Consumption concentration may be up, adding slightly to economic fragility.
Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit—Q1 2026.
Nvidia Investor Relations, Fiscal 2026 results.
Broadcom Investor Relations, Q2 FY2026 results.
Arista Networks Investor Relations, Q1 2026 results.
Vertiv Investor Relations, Q1 2026 results.
Amazon Investor Relations, Q1 2026 results.
Interactive Brokers, IBKR Fact Sheet.
Royal Caribbean Group, Q1 2026 earnings release.
Walmart Investor Relations, FY2026 Q4 earnings release.
TJX Investor Relations, Q4 and FY2026 results.
Ross Stores Investor Relations, Q1 2026 Form 8-K.
Ollie’s Bargain Outlet, Investor home and Q1 FY2026 results.
Dollar General Investor Relations, Q1 2026 results.
Costco Investor Relations, Q3 FY2026 results.
Stripe Economics, K-shaped economy?.
Disclosure: This article is for research and educational purposes and is not individualized investment advice. The author may hold securities discussed. Public-company fundamentals, prices, and risks can change quickly; readers should review current filings and valuation before investing.