The S&P 500 - the index that tracks the 500 largest US companies and is widely used as a proxy for the health of the American economy - just hit an all-time high of 7,400 points. Oil is above $80 a barrel. The 10-year US Treasury yield, which reflects the interest rate the US government pays to borrow money for a decade, is sitting at around 4.5%. Normally, expensive oil and high borrowing costs drag stock prices down. They make goods cost more and they make debt harder to service. So why is the market celebrating?

The uncomfortable answer, according to Eraldo de Paola, CIO of LS Advisory and a Miami-based institutional investment strategist, is that 75% of US GDP growth in the first quarter of 2026 came from a single source: investment in data center infrastructure for artificial intelligence. Not consumer spending, which historically drives 70% of the US economy. Not exports. Not government programs. A wave of AI-related construction, hardware, and compute spending - funded almost entirely by seven companies.

That is a surreal number. And the people who understand what it means are quietly buying insurance against what happens next.

The background

To understand why this matters, you need to understand how economies are normally supposed to grow. GDP - gross domestic product, the total value of everything a country produces in a year - is typically powered by millions of decisions made by ordinary people: buying a car, going to dinner, paying rent. Consumer spending makes up about 70% of the US economy. When households are doing well, the economy is doing well.

That relationship has broken down. Consumer sentiment in April 2026 fell to 47.6, a reading similar to June 2022, near a historic low. Americans feel economically anxious. But the stock market - and, by extension, the headline GDP number - looks fine. The reason is that a small group of very large companies is spending enormous sums of money on infrastructure, and that spending is showing up in GDP before most people have had any chance to benefit from it.

Hyperscalers is the industry term for the companies that own and operate the massive cloud computing networks that power most of modern digital life: Amazon (via AWS), Microsoft (Azure), and Google (GCP) are the three biggest. Add Meta, Nvidia, Micron, and Broadcom, and you have a cluster of seven companies whose investment decisions are now, quite literally, holding up the global economy.

Capex - capital expenditure, the money companies spend on physical assets like buildings, servers, and equipment - from these seven firms has become, as de Paola puts it, "the driver of marginal growth," not just for corporate earnings, but for the GDP number itself.

What is actually happening

The scale of this spending has very few historical parallels. Google, Amazon, Microsoft, and Meta collectively plan to spend $725 billion on capital expenditures in 2026, up 77% from the previous year's record of $410 billion, according to first-quarter earnings compiled by the Financial Times. That is not a typo. Three quarters of a trillion dollars, in a single year, from four companies.

Amazon alone projects $200 billion in capex for 2026, Alphabet between $175 and $185 billion, Meta between $115 and $135 billion, and Microsoft tracking toward $120 billion or more. To put the combined number in perspective, de Paola notes in the discussion that JP Morgan calculates total global data center infrastructure investment - public and private, listed and unlisted - now rivals the entire annual global military budget. You could add up what the US, Russia, and China spend on defense combined and roughly match what the world is pouring into AI infrastructure this year and next.

Roughly 75% of that hyperscaler spend is directly tied to AI infrastructure - servers, GPUs, data centers, networking equipment - rather than traditional cloud. These companies are increasingly leaning on debt markets to bridge the gap between rapidly rising capex budgets and their internal free cash flow, transforming historically cash-funded business models into ones utilizing leverage. A telling detail from de Paola: the tech sector, which in past years represented around 3% of all investment-grade corporate bonds issued globally, is on track to account for 15.5% this year - and by year-end, according to JP Morgan, the sector could hold more total debt than the entire US financial industry. Google recently issued a 100-year bond. Amazon put $54 billion of new debt into the market alongside a free cash flow quarter of $1.2 trillion.

Meanwhile, Microsoft's CFO attributed $25 billion of the company's record capex budget to rising memory chip prices alone. The cost of RAM chips has jumped from $3.76 to $9.71.

The fiscal policy backdrop is adding fuel. The "One Big Beautiful Act" introduced under the Trump administration includes bonus depreciation - an accounting rule that lets businesses immediately write off the full cost of equipment purchases against their taxes, rather than spreading that deduction over many years. For a small landscaping company in Florida, that might mean replacing a pickup truck and zeroing out the tax bill. For Microsoft announcing $190 billion in infrastructure spending, it is an enormous subsidy to keep spending faster.

The money trail

The financial structure underpinning all of this is more circular than it first appears. De Paola describes what analysts at JP Morgan call a closed loop - or circularity: the hyperscalers, AI labs like OpenAI, and hardware suppliers like Nvidia all invest in each other, finance each other, and buy each other's products, in a system where revenue is recognized based on orders placed rather than transactions completed.

That circular relationship is why Oracle's debt-funded deal with OpenAI - a company whose ability to honor such a large contract remains publicly unverified - spooked credit markets. It is why JP Morgan estimates there are roughly $30 billion in receivables circulating through this ecosystem that are generating unflattering headlines in credit analysis circles.

The equity market has so far decided to look past this. But the stock market's comfort rests on a very narrow base. Just three AI hyperscalers - Alphabet, Amazon, and Meta - account for approximately 70% of S&P 500 earnings growth expectations for 2026 in dollar terms, according to analysis highlighted by Charles Schwab. De Paola's own figures, referencing Morgan Stanley consensus data, put it slightly differently: seven companies - Nvidia, Micron, Broadcom, Microsoft, Meta, Amazon, and Google - are responsible for 52% of expected earnings-per-share growth across all 500 companies in the index for 2026.

The 10 largest companies in the S&P 500 now control roughly 40% of the entire index by market capitalization, a level that has exceeded the peak concentration seen during the dot-com bubble of 2000, when the top 10 held around 27%.

The paradox is sharp. These are not fraudulent companies running on vaporware. Their balance sheets are strong, their cash flows are historically large, and their products are real. But those same companies are now spending 100% of their free cash flow on capex - up from a historical average of around 40%. Free cash flow at the biggest tech names is falling even as reported earnings grow 20%, because so much cash is going out the door on construction and hardware. When free cash flow falls, it has historically been an early warning sign of an eventual earnings decline.

In the first half of 2025, AI-related capex contributed more to US GDP growth than consumer spending, according to KKR analysis. That ratio held into Q1 2026. The consumer is still there - and the top 10% of earners, who now account for 50% of US consumption (up from a third before the pandemic), are still spending. But the rest of the consumer base is under pressure, and the macro indicators that have reliably predicted recessions in the past - from the inverted yield curve to traditional credit measures - are giving erratic, contradictory readings.

De Paola's firm uses a proprietary cycle-tracking model called the GCLI (Growth Coincidence and Leaders Index), which has operated for over 30 years. For most of the past 12 months, that model has flagged the US economy as being in "contraction." It has recently shifted toward a "slowdown and recovery" reading, which the model treats as a transitional regime - statistically unlikely to persist. The directional ambiguity is deliberate: the economy is genuinely hard to read when GDP is being propped up by a spending category that traditional macro indicators were not designed to capture.

What people are doing about it

The options market is pricing in fear that most headlines are not. De Paola describes pricing a put option - a financial contract that pays out if an asset falls in value - on the Nasdaq at two standard deviations below current implied volatility for a six-month horizon. In normal market conditions, that level typically lands around -20% to -25%. Today, that same contract is priced at -35%. The market for downside protection, in other words, has moved. A lot of people with serious money are spending real dollars to hedge against a crash, even as the index sets new highs.

The credit default swap market - where investors buy insurance against companies defaulting on their bonds - is also being watched. According to JP Morgan data cited by de Paola, protection against default by the major hyperscalers is currently trading at around 80 basis points above the rate on the underlying bonds. That is cheap, by historical standards. But the fact that a market for this protection exists at all, and is being actively traded, signals that institutional investors are not uniformly convinced the current trajectory holds.

Michael Burry - the investor made famous in The Big Short for correctly predicting the 2008 US housing collapse - has warned that the Nasdaq 100 is headed toward a dramatic reversal after a "parabolic" surge that has driven technology valuations to unsustainable heights. Writing on Substack on May 8, Burry said the market resembles the peak of the dot-com bubble just before it burst. His most pointed statistical comparison was to the Philadelphia Semiconductor Index - which tracks Nvidia, Broadcom, Intel, Micron, and TSMC - which rose more than 10% in a single week ending May 8, pushing its 2026 gains to approximately 65%.

Burry has backed his words with options positions: he has added to Nvidia puts for January 2027 at a strike of $115 and March 2027 at $125, and to QQQ puts - bets against the Nasdaq-100 tracking ETF - for January 2027 at $550, while loading up on short positions in the semiconductor ETF SOXX through January 2027.

Institutional allocators who are not betting on a crash are nonetheless quietly rebalancing. Advisors at firms like WWM Investments are trimming positions in names that have had exceptional runs and reallocating toward areas with more attractive valuations. De Paola's own firm is currently shifting client focus from equity growth positions toward high-quality bonds, particularly in three-to-five-year US Treasury maturities, where yields have risen sharply enough to offer real compensation. A five-year Treasury that paid 3.5% a month ago is now paying 4.1% - a meaningful pickup in return without taking on stock-market risk.

The bottom line

The US economy is growing, but the growth engine is small, concentrated, and running hot. Seven companies are spending at a pace that has no real historical precedent, on the bet that AI infrastructure will eventually generate returns large enough to justify the outlay. The stock market believes them. Consumer sentiment does not. Traditional economic models cannot fully account for what is happening, because nothing quite like this has happened before. Michael Burry thinks it ends badly and soon. JP Morgan's credit team says the CDS market is calm for now. Both things are true at the same time. The honest read is that the world is running a very large, very expensive experiment - and has not yet received the results.

Timeline

Summary

Who: Seven major technology companies - Amazon, Microsoft, Alphabet/Google, Meta, Nvidia, Micron, and Broadcom - along with institutional investors, allocators, and market analysts including Eraldo de Paola (CIO, LS Advisory) and Michael Burry (Scion Asset Management)

What: A historically unprecedented concentration of corporate investment in AI infrastructure is now responsible for the majority of US economic growth, while driving record stock market valuations. The same seven companies account for over half of expected S&P 500 earnings growth for 2026. Options markets and credit analysts are quietly pricing in elevated downside risk.

When: The buildup has accelerated through 2025 and into 2026, reaching its most acute point in Q1 2026, with Burry's crash warning issued on May 8, 2026.

Where: Primarily in the United States, where two-thirds of global AI infrastructure investment is concentrated, though supply chains - including helium for chip manufacturing sourced partly from the Persian Gulf - introduce global dependencies.

Why: Expectations that AI will generate transformative economic returns are driving a massive capital race among technology companies, amplified by US fiscal policy (bonus depreciation) and debt markets willing to fund these bets. Whether those returns materialize fast enough to justify the spending is the central unresolved question facing global markets.