Every few decades, a new technology convinces the world's biggest companies to rebuild their foundations at the same time. Railroads did it in the 1800s. Electrification did it a century later. Fiber-optic networks did it in the late 1990s. Today, it is artificial intelligence.

Microsoft, Alphabet, Amazon and Meta are on track to spend somewhere between $700 billion and $750 billion on capital expenditures in 2026, much of it on AI chips, data centers and the power to run them. By one Financial Times compilation of first-quarter earnings, that is up roughly 77% from about $410 billion in 2025. Few corporate investment cycles in history have grown this fast.

The real question is not whether AI is useful. It clearly is. The question is whether the revenue and profits it generates will arrive soon enough, and in large enough amounts, to justify the bill. This guide breaks down where the money is going, who is spending it, what returns investors are counting on, and where the risks of overinvestment may hide.

Where the Money Is Going: The Real Cost of AI Infrastructure

AI infrastructure spending is often described as a single number, but it is really several large bills stacked on top of each other.

Chips and Accelerators

The most visible cost is the processors that train and run AI models. No company shows this better than Nvidia. In its second quarter of fiscal 2027, which ended July 26, 2026, Nvidia reported revenue of $96.2 billion, up 106% from a year earlier. Its Data Center segment alone brought in $89.0 billion, up 117%, with a gross margin of 75%, according to the company's earnings release. For the next quarter, Nvidia guided to about $108 billion in revenue, give or take 2%, and said its outlook assumes no data center compute revenue from China.

It helps to remember that one company's spending is another company's sales. When investors read headlines about record capex, they are often reading Nvidia's revenue line from the other side.

Prices matter, too. Microsoft's chief financial officer said about $25 billion of the company's 2026 capex budget was tied to rising memory chip prices, according to the Financial Times. In other words, part of the increase is not buying more capacity. It is buying the same capacity at a higher price.

Data Centers, Networking and Land

Chips are only part of the bill. They have to live somewhere: in enormous buildings filled with servers, high-speed networking equipment and cooling systems. Goldman Sachs strategists now estimate that Meta, Microsoft, Amazon and Alphabet will spend a combined $5.3 trillion on capital expenditures from fiscal 2025 through fiscal 2030, up from a $4.5 trillion estimate made before the latest round of earnings. Across compute, data centers and power, Goldman's baseline for 2026 through 2031 is $7.6 trillion.

Electricity: The Quiet Constraint

Power may be the most underappreciated cost. The International Energy Agency (IEA) reports that electricity demand from data centers grew 17% in 2025, more than five times faster than overall global electricity demand growth of 3%. The agency projects that data center demand will double by 2030, and its earlier special report put that level at roughly 945 terawatt-hours, about what Japan consumes today. A single 100-megawatt data center can use as much electricity as roughly 100,000 households, the IEA noted.

For investors, that means AI is not only a software story or a chip story. It is also a utilities, grid and construction story, and each of those has its own bottlenecks and costs.

The Players: How Big Tech Is Funding the Buildout

The Four Biggest Spenders

The companies below, often called hyperscalers, account for the bulk of AI infrastructure spending. Their latest public guidance is summarized here. Note that the companies measure capex differently (for example, Meta's figure includes principal payments on finance leases), so the numbers are not perfectly comparable.

Guidance changes often, so check each company's most recent filing before relying on any figure. The next round of quarterly reports will update these numbers.

Beyond the Big Four

Oracle is frequently counted as a fifth hyperscaler, and a growing group of AI-focused cloud providers and AI labs are signing their own large compute deals. OpenAI, for example, was reportedly targeting roughly $600 billion in total compute spending through 2030, according to The Information. Commitments like these are a reminder that the spending is not limited to the largest public companies.

The Cash-Flow Squeeze

Historically, the hyperscalers funded growth from the cash their core businesses generated. That is getting harder at the current pace. Alphabet's second-quarter capital spending reached $44.9 billion, double the year-earlier figure, while its free cash flow turned negative at $5.9 billion, according to reports based on its filings. Amazon's trailing free cash flow fell to $1.2 billion from $25.9 billion a year earlier as equipment purchases rose by $59.3 billion.

Markets noticed. Alphabet's shares fell about 4% after its July report despite an earnings beat, a sign that investors are scrutinizing spending as closely as profits.

The Expected Return: What Must Go Right

The Case for Optimism

There is real evidence of demand. Cloud businesses, the main channel through which AI infrastructure turns into revenue, have been growing quickly. Reported quarterly cloud growth rates include roughly 40% at Microsoft Azure, 82% at Google Cloud and 28% at Amazon Web Services. Microsoft said in the spring that its AI business had passed a $37 billion annualized revenue rate.

The leading AI labs are also reporting rapid growth. Based on media reports, Anthropic's annualized revenue run rate rose from about $9 billion at the end of 2025 to roughly $65 billion by the end of July 2026, and OpenAI's was reported near $40 billion. These are unaudited run-rate figures, which extrapolate a recent period into a full year, and the companies may calculate them differently. Audited numbers should become available if these companies complete their planned public listings.

Some analysts argue the growth justifies the outlay. Brent Thill of Jefferies told the Financial Times that the AI economy is healthy and that recent revenue growth supports the spending.

The Case for Caution

Skeptics focus on the gap between spending and returns. Dec Mullarkey of SLC Management told the Financial Times that investors worry Meta is shifting from a capital-light business into what he called a "capital-intensive incinerator." Meta's AI spending is largely for its own advertising and recommendation systems, so its payoff depends on how much those tools lift ad revenue, rather than on selling cloud capacity to outside customers.

There is also a timing problem. Servers and chips lose value as they age and as better hardware arrives. Statista noted that part of the depreciation tied to this spending has not yet reached income statements and will weigh on profits next year.

Here is a simple illustration, using an assumed number rather than company data. If a business buys $100 billion of equipment that lasts about five years, it must eventually earn back that $100 billion plus a return on it. That implies roughly $20 billion a year in depreciation alone, before power, staff and financing costs. Multiply that logic across hundreds of billions of dollars a year, and the revenue needed to make the math work becomes enormous.

The Risks of Overinvestment

Rising Reliance on Debt

As cash flows get stretched, borrowing is rising. Morgan Stanley estimated that AI-related global debt issuance reached $236 billion as of May 31, about four times the amount a year earlier, and projected it could reach $570 billion by the end of 2026. In a January report covered by Reuters, Bank of America analysts expected the five biggest hyperscalers to borrow roughly $140 billion a year over the next three years, with the possibility of exceeding $300 billion annually. Alphabet's decision this year to include a 100-year bond in a debt sale of more than $30 billion, a structure not seen from a major tech company since Motorola in 1997, drew particular attention, according to Capital Group.

Lenders have started to push back. Barclays research cited by Creditflux showed spreads on AI data center credits widening by more than 100 basis points since June. Wider spreads mean borrowers pay more to raise money, which is an early signal bond investors are demanding more compensation for risk.

Circular Financing and Concentration

Some analysts have also flagged so-called vendor financing, where a supplier helps fund a customer who then buys the supplier's products. Used sparingly, it is a normal business practice. Used heavily, it can make demand look stronger than it is. Investors should also watch concentration: a small number of buyers account for a large share of chip sales, so a slowdown at any one of them could ripple through the entire supply chain.

Lessons From the Dot-Com Era

Comparisons to the late 1990s are common. A Capital Group strategist noted that one feature of that bubble was accelerating revenue growth paired with promises of future profitability. But there are important differences. Today's biggest spenders generally have large, profitable core businesses and relatively low leverage, which gives them a cushion that many telecom companies lacked in 2000. A downturn in AI would likely hurt returns and share prices before it threatened the survival of the largest players. Smaller, more indebted companies in the ecosystem face greater risk.

What Investors Should Watch

Whether you are new to investing or have decades of experience, a few indicators can help you cut through the headlines:

• Capex versus operating cash flow. Is spending outpacing the cash the business generates? Negative free cash flow is not automatically bad, but it raises the stakes.

• Cloud growth and backlog. Sustained double-digit growth in cloud revenue and contracted backlog suggests demand is real.

• Depreciation trends. Rising depreciation will pressure margins unless revenue keeps pace.

• Credit conditions. Watch bond spreads and issuance terms for signs that lenders are growing wary.

• Revenue quality. Distinguish between audited revenue and run-rate claims, and between recurring contracts and one-time deals.

• Forward guidance. Alphabet's chief financial officer has said 2027 capex should increase significantly from 2026, and Nvidia has guided to continued growth. Slower guidance from either would be a notable signal.

For beginners, the practical takeaway is diversification. Owning a broad index fund already includes meaningful exposure to these companies, so adding concentrated bets on AI stocks raises risk. More experienced investors may prefer to think in scenarios, such as asking how each holding would fare if AI revenue arrives on schedule, arrives three years late, or never reaches the levels the market expects. Your time horizon and risk tolerance matter more than any single forecast, and a licensed financial professional can help apply these ideas to your situation.

Conclusion: Real Demand, Real Risks

The evidence points in two directions at once. On one side, demand for AI computing is strong: Nvidia's results, cloud growth and rapid revenue gains at leading AI labs all suggest customers are paying for it. On the other side, the scale of spending is extraordinary, free cash flow is under pressure at some of the largest companies, borrowing is rising, and lenders are asking for more compensation.

Neither the bulls nor the bears have been proven right yet. If AI becomes as embedded in the economy as electricity or the internet, today's spending may look like the early cost of a transformative platform. If adoption or pricing falls short of expectations, parts of the buildout could look like overbuilding, much like the fiber networks of the late 1990s, which eventually proved useful but not before many investors lost money.

The most reasonable stance for investors is neither euphoria nor fear. It is attention. Track the cash flows, the debt and the audited revenue, stay diversified, and let the data, not the hype, update your view.

Sources and Further Reading

• NVIDIA, Financial Results for Second Quarter Fiscal 2027 (Aug. 26, 2026), via SEC filing: https://www.sec.gov/Archives/edgar/data/0001045810/000104581026000073/q2fy27pr.htm

• Reuters, "AI hyperscalers will drive higher US corporate bond supply in 2026, analysts say" (Jan. 15, 2026): https://finance.yahoo.com/news/ai-hyperscalers-drive-higher-us-225314460.html

• Financial Times capex compilation as reported by Tom's Hardware, "Big Tech's AI spending plans reach $725 billion": https://tomshardware.com/tech-industry/big-tech/big-techs-ai-spending-plans-reach-725-billion

• Goldman Sachs estimates as reported by Yahoo Finance: https://finance.yahoo.com/sectors/technology/article/meta-microsoft-amazon-and-alphabet-are-about-to-spend-a-shocking-amount-of-money-to-dominate-the-ai-era-115359575.html

• Sherwood News, "Alphabet, Amazon, Microsoft, and Meta plan to spend more than $700 billion on capex this year": https://sherwood.news/tech/alphabet-amazon-microsoft-meta-plan-more-than-700-billion-on-capex-this-year/

• Statista, Big Tech Q2 2026 capex chart: https://www.statista.com/chart/35046/

• International Energy Agency, Energy and AI: https://www.iea.org/reports/energy-and-ai (updated findings as reported by W.Media, April 20, 2026: https://w.media/data-center-power-demand-will-double-by-2030-iea/)

• Capital Group, "How to spot an AI bubble": https://www.capitalgroup.com/pcs/insights/articles/how-to-spot-ai-bubble.html

• Creditflux, "Debt markets preach caution on AI as bubble worries grow" (Aug. 2026): https://newsletter.creditflux.com/debt-markets-preach-caution-on-ai-as-bubble-worries-grow

• Morgan Stanley estimates as reported by The Motley Fool: https://www.webull.com/news/15236076935980032

• Renaissance Capital, "9 numbers to look for in Anthropic's IPO filing" (Oct. 5, 2026): https://www.renaissancecapital.com/IPO-Center/News/122058/9-numbers-to-look-for-in-Anthropic’s-IPO-filing

• Bloomberg Businessweek reporting on 2026 capex, as relayed by FourWeekMBA: https://fourweekmba.com/ai-meta-google-amazon-microsoft-near-750b-of-2026-capex/

Disclaimer: This content is for informational and educational purposes only and does not constitute financial, investment or legal advice. Investing involves risk, including the possible loss of principal.

suggestions: