A360
Atlas AI
Yesterday’s GDP report did not prove that America has become a K-shaped economy. One quarterly release cannot do that. But it added an important piece of evidence.
2%. The headline was respectable. The composition was more revealing. Business investment surged, helped by spending on equipment, software and intellectual-property products linked to artificial intelligence. 6%, with goods spending weakening even as services held up. 9% in the prior quarter.
That is not a recession. It is not a broad boom either. It is an economy increasingly shaped by a split between capital-intensive growth at the top and household strain below.
Understanding the K-shaped Economy
A K-shaped economy is not merely an unequal economy. It is one in which the same macro environment produces opposite outcomes for different groups. Capital owners benefit from asset appreciation. Wage earners face higher prices. AI suppliers enjoy surging demand. Consumer businesses see shoppers trade down. Large firms finance a new investment cycle. Smaller firms and households remain exposed to expensive credit.
Yesterday’s numbers support that interpretation, but they do not complete the case. For that, one has to look at this week’s earnings, inflation dynamics, rates and the wealth effect from the AI trade.
The most striking feature of the first-quarter GDP report was the strength of business investment. Overall business investment rose at a 10.4% annualised rate, the fastest pace in nearly three years. Spending on categories closely tied to AI, including equipment and intellectual property products, was especially strong.
Oliver Allen of Pantheon Macroeconomics put a number on the phenomenon: “Even after accounting for the fact that most computer equipment is imported, AI investment seems like it accounted for about half of the overall GDP growth in the first quarter.”
That sentence deserves attention, but also caution. AI-related investment is not a perfectly clean national-accounts category. Some computer equipment is imported, which subtracts from GDP through net exports. Some software and equipment spending may not be purely AI. Some capex may represent defensive overbuilding rather than immediately productive investment.
Still, the broad point is hard to dismiss: AI-linked investment is now large enough to matter at the macro level. It is not just moving valuations. It is moving output.
AI's Impact on the Real Economy
That marks a shift. For much of the past two years, AI was primarily a market narrative. It justified higher multiples for chipmakers, cloud platforms and software firms. Now it is becoming a real-economy force: data centres, chips, cloud capacity, power infrastructure, cooling systems, networking equipment and specialised labour.
The question is whether this is merely an investment boom or the beginning of a productivity boom. Those are not the same thing.
Today’s AI capex boosts aggregate demand. Companies spend money on equipment, construction, software and infrastructure. That appears immediately in GDP. Tomorrow’s hoped-for payoff is a supply-side improvement: higher output per worker, better margins, faster innovation and lower unit costs.
The risk is that America gets the investment boom before it gets the productivity dividend.
Janet Yellen’s old warning about productivity growth is useful here: “Productivity growth, however it occurs, has a disruptive side to it. In the short term, most things that contribute to productivity growth are very painful.” In other words, even if AI eventually raises living standards, the transition may first concentrate gains among capital owners and firms with scale.
That is what the current data suggest. Capital deepening is arriving faster than broad household relief.
Earnings Confirm the Economic Split
The GDP report was not the only evidence. This week’s earnings told the same story in corporate form.
Big Tech continues to invest at a scale that looks less like a normal business cycle and more like the construction of a new industrial platform. Cloud providers, chipmakers and AI infrastructure firms are still benefiting from extraordinary demand. The upper branch of the K remains powerful: firms with capital, data, distribution and computing infrastructure are expanding aggressively.
But consumer-facing companies are sending a more cautious message.
The uploaded analysis notes that Domino’s revised down its U.S. same-store sales growth estimates after demand softened, with lower-income shoppers particularly pulling back. Procter & Gamble, a bellwether for household spending, reported higher quarterly sales but warned that geopolitical uncertainty and political tensions were weighing on shoppers. Its finance chief, Andre Schulten, put it simply: “The consumer is still a little bit muted.”
That is not a collapse. It is something subtler and more important: the marginal consumer is rationing.
This distinction matters. Consumer spending still grew. Services spending held up. The labour market has not broken. But lower-income households are becoming more selective, more price-sensitive and more vulnerable to fuel and financing costs.
That is how K-shaped slowdowns often begin. The aggregate data stay respectable because higher-income households, asset owners and large firms keep spending. But beneath the aggregate, the lower branch weakens.
The Wealth Effect and Uneven Gains
The K-shape is not just AI capex versus consumer spending. It is also asset owners versus wage earners.
A rising AI market supports high-income consumption through equity portfolios, stock compensation, retirement accounts, private-company valuations and corporate confidence. The households that own financial assets benefit first. The households that depend mainly on wages feel inflation first.
That matters because stock ownership and capital income are highly concentrated. When AI-linked equities rise, the gains do not flow evenly through the economy. They boost the balance sheets of upper-income households and the investment capacity of large companies. Meanwhile, households without meaningful financial assets encounter AI indirectly: through changing labour markets, higher electricity demand, higher service prices, and potentially greater job insecurity.
This is the strongest economic foundation for the K-shaped thesis. Not simply that business investment rose and consumption slowed, but that the mechanisms of growth are increasingly capital-heavy and asset-heavy.
Mohamed El-Erian has framed the optimistic version of this as a possible “non-inflationary boom” tied to AI spending and productivity. That is the bull case. If AI raises output per worker quickly enough, it can support growth without reigniting inflation. But yesterday’s GDP report was less clean than that. It showed the AI investment boom arriving alongside hotter inflation and slower consumer spending.
That is why the report felt so revealing. It contained both the dream and the problem.
Inflation Exacerbates the Divide
Inflation is not just a macro variable. It is distributional.
The first-quarter PCE price index rose at a 4.5% annualised rate, compared with 2.9% in the previous quarter. For higher-income households, inflation is an inconvenience. For lower-income households, it is a binding constraint.
Gasoline intensifies that effect. The uploaded analysis notes that after the U.S. and Israel launched attacks on Iran on the last day of February, the average price of regular gasoline jumped sharply. Regular fuel averaged about $4.11 a gallon in April, up from March’s $3.67 average, and hit a postwar high of $4.30 on Thursday.
Energy inflation is especially regressive. It hits lower-income households harder because fuel and utilities take a larger share of their budgets. It also spills into expectations. When gasoline prices rise, consumers notice quickly. They may not read GDP tables, but they see the pump.
Joseph Brusuelas of RSM captured the timing problem: “What you have here is an AI-inspired GDP increase, that in coming quarters will see a drag,” because of the Iran war.
That is a crucial point. The first-quarter GDP report captured the AI upside more fully than it captured the gasoline downside. The second quarter may look less comfortable.
AI itself has a complicated relationship with inflation. In theory, it should be disinflationary over time if it raises productivity and lowers unit labour costs. But the buildout phase can be inflationary. It competes for electricity, skilled labour, construction capacity, chips, land and capital. The long-run productivity promise may be real, but the short-run demand shock is real too.
The economy is therefore living through a transition in which AI may reduce inflation tomorrow while adding to pressure in selected sectors today.
The Fed is facing a structural dilemma
This is why the Federal Reserve’s job is so difficult.
The problem is not simply that the Fed is stuck between growth and inflation. The deeper issue is that the same interest rate can look appropriate for one part of the economy and punishing for another.
Large technology companies can finance AI infrastructure because they have enormous cash flows, strong balance sheets and access to capital markets. Many households and smaller firms cannot. They face mortgage rates, credit-card APRs, auto-loan payments and refinancing costs.
An AI capex boom may also raise investment demand and potentially the neutral rate of interest if expected returns on capital rise. At the same time, higher rates bite harder on households with floating-rate debt and limited savings. Monetary policy can therefore feel simultaneously too loose for the AI-investment economy and too tight for the consumer-credit economy.
That is a more precise version of the Fed’s dilemma.
The uploaded analysis notes that investors began the year expecting two rate cuts in 2026. By Thursday, they saw nearly an 80% probability that the Fed would keep rates on hold through December. The Fed held rates steady this week, citing elevated inflation and weak job gains.
Mary Daly of the San Francisco Fed has argued that policymakers should not assume AI productivity gains before they show up in the data. That caution is right. Markets are pricing the AI future. The Fed has to manage the inflation present.
The result is uncomfortable. The Fed may have to keep policy tight because inflation and investment demand remain firm, even as lower-income consumers weaken. That would not create the K-shaped economy by itself. But it would reinforce it.
Markets should love the boom and fear its narrowness
For equity markets, the message is constructive but concentrated.
AI infrastructure remains the cleanest growth story in the economy. The beneficiaries are familiar: semiconductors, cloud platforms, data-centre operators, power equipment, utilities, grid suppliers, cybersecurity firms, networking companies and software businesses that can monetise AI rather than merely advertise it.
But the same concentration that makes the trade attractive also makes it fragile.
If a small number of companies are carrying a large share of the investment cycle, the market becomes vulnerable to any disappointment in AI monetisation. Spending alone will not be enough forever. Investors will increasingly ask whether capex produces returns, whether AI revenue can justify depreciation, and whether customers will pay enough for AI-enabled products to sustain margins.
Not every server rack is a productivity miracle. Some may turn out to be the fibre-optic cable of this cycle: essential in the long run, but ruinous for some investors in the short run.
For bonds, the picture is less friendly. Consumer softness would normally support Treasuries and rate-cut expectations. But sticky inflation, energy shocks and strong investment demand complicate that story. If inflation remains elevated, the Fed has little room to rescue the lower branch of the economy.
For credit, the implication is dispersion. Strong balance-sheet companies tied to AI infrastructure should remain resilient. Leveraged consumer-facing businesses, lower-quality borrowers, restaurants, retailers and housing-linked firms face greater risk.
For the dollar, higher-for-longer rates and geopolitical uncertainty may provide support. But the broader “American exceptionalism” thesis becomes more vulnerable if U.S. growth is seen as increasingly dependent on a narrow AI capex boom.
The political economy is combustible
The political problem is simple: macro resilience does not feel like resilience when the gains are uneven.
Officials can point to 2% GDP growth. Households can point to gasoline above $4. Companies can report strong earnings. Consumers can report weak confidence. Markets can celebrate AI. Workers can worry about job security.
All of these can be true at the same time.
That is what makes the K-shaped economy so politically dangerous. It creates a gap between statistical reality and lived experience. It allows policymakers to say the economy is strong while voters feel squeezed. It allows investors to celebrate productivity while workers fear displacement. It allows large firms to invest aggressively while smaller firms struggle with credit costs.
A capital-led expansion can be productive. It can also feel exclusionary.
The social risk is not that AI fails. The social risk is that AI succeeds first for those who already own capital, while the adjustment costs arrive earlier for those who depend on wages.
What would prove or disprove the thesis
A serious K-shaped argument needs more than one GDP report. The data to watch are clear.
First, consumer credit: credit-card delinquencies, auto-loan stress and subprime performance. If these worsen while headline GDP remains positive, the K-shaped thesis strengthens.
Second, labour-market dispersion: wage growth by income percentile, unemployment by education level, hiring in AI-linked sectors versus job losses in clerical, administrative and lower-skill knowledge-work roles.
Third, wealth effects: equity ownership, stock-based compensation, household net worth by income group, and high-end versus low-end consumption.
Fourth, corporate dispersion: earnings revisions and margins for AI-linked companies versus consumer-discretionary firms.
Fifth, productivity: whether AI spending actually raises output per hour, rather than merely increasing capital expenditure.
If productivity accelerates and real wage gains broaden, the K-shape may narrow. If AI capex stays strong while consumer credit deteriorates and lower-income spending weakens, the K-shape will deepen.
The bottom line
Yesterday’s GDP report does not prove that America is a K-shaped economy. But it strongly supports the hypothesis.
The economy is growing, but the sources of growth are becoming narrower and more capital-intensive. AI investment is lifting headline output. Asset owners are benefiting from the market’s enthusiasm. Large companies with scale are spending aggressively. At the same time, consumers are slowing, inflation is sticky, gasoline prices are painful and the Fed has little room to cut.
That is not a normal expansion. It is a split expansion.
The optimistic interpretation is that this is the beginning of an AI-led productivity boom that will eventually lift wages, margins and living standards. The pessimistic interpretation is that America is building a growth model that looks strong in GDP and earnings, but weak in household experience.
For now, the evidence points to something in between: a powerful but uneven boom.
The upper branch is being built in data centres. The lower branch is waiting for relief.
