Alphabet Issues 100-Year Bond Amid Surge in AI Infrastructure Spending

Article by Sleepy

On February 10, 2026, Google’s parent company Alphabet issued a bond in London with a maturity date one hundred years from now.

One hundred years.

The person who buys this bond is essentially betting that the company will still be alive and able to repay when their grandchildren retire.

Throughout history, century-long bonds have been extremely rare. Disney issued one in 1993, Coca-Cola issued one, and earlier still, Norfolk Southern Railway did as well. In fact, this maturity was standard for 19th-century railroad companies, as they needed to lay tracks, dig tunnels, and build bridges—investments with such long payback periods that time had to be measured in centuries.

But now, an internet company is borrowing money like a railroad company—why?

Over the past eighteen months, the answer to this question has gradually emerged. It is not found in any AI product launch’s PowerPoint, not on benchmark leaderboards, and not in the debates over “when AGI will arrive.” It is hidden in the capital expenditures section of earnings reports, in the changing spreads of bond issuances, and in the steep plunge of free cash flow.

To understand this answer, first look at how something disappears.

What we first need to understand is that over the past two decades, Wall Street’s greatest belief was not in any single technology company itself, but in the financial structure that all these companies adopted.

These tech companies generate revenue from advertising, cloud services, and platform commissions—all digitally delivered with marginal costs approaching zero. They require no factories, no inventory, no mines or oil wells. The more users they have, the lower the cost per user and the higher the profit margin.

The direct result of this structure is free cash flow. Unlike net income on the income statement, which can be influenced by accounting standards, free cash flow represents actual money flowing into the bank account—money that can be used to repurchase shares, pay dividends, or invest in the future. This is why U.S. tech stocks command a premium valuation.

There was a joke once about Apple sitting on over $200 billion in cash, not knowing how to spend it; Google generating tens of billions in free cash flow year after year, as if its search box were directly connected to a gold mine; Amazon, masking itself as a low-margin e-commerce company, actually operating as a cloud computing money printer; and Meta making massive profits by having billions of people view ads every day.

Investors aren't just buying growth—they're buying the narrative of "asset-light, high-cash-flow" businesses, because it promises that these companies will never be weighed down by factories like General Motors, burdened by infrastructure debt like AT&T, or tormented by the capital expenditure cycle like oil companies. They can simply ignore the gravity of industrial capitalism.

Then AI arrived. And the results it brought were unexpectedly surprising.

At the end of April this year, Amazon released its Q1 financial report, with revenue, profit, and AWS growth all performing reasonably well. Over the past twelve months, Amazon’s operating cash flow reached $148.5 billion, a 30% year-over-year increase, which looks promising. However, free cash flow during the same period dropped from $25.9 billion to $1.2 billion, a 95% decline.

Where did the money go? Amazon’s Q1 capital expenditures reached $44.2 billion, a 76.7% year-over-year increase, with full-year guidance around $200 billion. Almost all of this funding is being directed toward AI infrastructure, such as data centers, GPUs, networking equipment, and power contracts.

Amazon isn't losing money—in fact, it's making more than ever. It's just pouring nearly all of that money into the bottomless pit of AI. Operating cash flow is the Yangtze River, capital expenditures are the Three Gorges Dam, and free cash flow has become a trickle below the dam.

The other几家 haven't fared much better.

In 2026, the combined capital expenditure guidance for the four giants totals $700 billion to $725 billion: Amazon approximately $200 billion, Microsoft approximately $190 billion, Alphabet approximately $185 billion, and Meta $125 billion to $145 billion. In 2022, the four combined spent $162 billion, a 4.5-fold increase over four years. Just in Q1 2026 alone, the four collectively spent over $130 billion—more than the annual GDP of many countries.

On the surface, these companies still look impressive. Revenue is rising, profit margins remain healthy, and AI product launches continue to generate excitement. But "free cash flow" thinks otherwise.

Profit is ultimately a matter of perspective—how you set depreciation periods, capitalize R&D, or recognize revenue leaves room for maneuver; but cash flow is a fact—exactly how much money comes in and how much goes out is clear and unambiguous. Profit tells a story; cash flow tells the truth.

So the truth is that the core financial advantage these companies spent two decades building—light asset ownership and high cash returns—is being gradually eroded by AI’s capital expenditures.

The next question is: if free cash flow has hit rock bottom but their investments are still increasing, where is this money coming from?

Borrow. And the way they are borrowing now is something we have never seen before.

Alphabet borrowed $32 billion in February.

A month later, in March, Amazon completed another bond issuance of $36.9 billion across 11 tranches, ranging from two-year to fifty-year maturities. Investor demand totaled $126 billion, more than 3.4 times oversubscribed. After this issuance, Amazon’s total debt nearly doubled in one year. A month after that, on April 30, Meta issued $25 billion in bonds.

One month later, on May 11, Alphabet announced it was preparing its first yen-denominated bond. This is interesting, as Alphabet’s bond issuance in February already included not only U.S. dollars but also 3.1 billion Swiss francs.

This is a company based in California, USA, with nearly all its revenue denominated in U.S. dollars, yet it went to Switzerland to borrow money. In May, it also turned its attention to the Japanese yen. Amazon’s March transactions also included euro-denominated batches.

This isn't a token diversification done by these giants' finance departments for appearances—it was forced upon them.

Look at Meta: its $25 billion bond issued in April, with the longest tranche maturing in 2066, had a spread of 147 basis points—this is the risk premium investors demanded over U.S. Treasuries. Six months earlier, in October 2025, when Meta issued a similar 40-year tranche, the spread was only 110 basis points. In six months, it widened by 37 basis points—and not just for the longest tranche; nearly all six tranches issued carried higher premiums than the previous round.

So, these giants need to seek out lower interest rates. The Swiss National Bank’s policy rate is the lowest among major economies, and Swiss franc bond yields are significantly lower than those of the U.S. dollar. Although Japan has ended its negative interest rate era, the cost of funding in yen still offers substantial advantages. More importantly, investors in Zurich and Tokyo haven’t been saturated with Silicon Valley’s tech debt—they still have fresh appetite, unlike the more discerning investors in New York. For borrowers with top-tier credit like Alphabet, borrowing elsewhere is both cheaper and avoids the queue.

AI's capital expenditures are concentrated in the United States (data centers) and Taiwan (chips), but the money to pay for them comes from Switzerland, Japan, and Europe. Silicon Valley has been technologically globalized for two decades; now it is also globally indebted.

The buyers of these bonds are not hedge funds or venture capital. Those who can absorb century- and fifty-year bonds are pension funds, insurance companies, and sovereign wealth funds—the most risk-averse capital in the global financial system. Their mission is to preserve capital, remain stable, and outpace inflation, not to take risks.

But now, the pension of a retired teacher in Zurich, the reserves set aside by a life insurance company in Tokyo, are flowing through the bond market’s transmission chain into data centers in Oregon or Virginia, becoming racks of GPUs and cooling towers on the roof. Most of these holders have no idea what the underlying assets of their bonds actually are; their fund managers buy “Alphabet Aa2 credit” or “Amazon A1 credit,” and the rating agencies’ letters provide a sense of security. As for what buildings were constructed with this money, what equipment was installed, what models are running, and whether those models can generate enough revenue to repay the debt—it’s obscured by too many intermediaries, invisible from Zurich and Tokyo.

The world's most conservative money is funding bets on the world's most radical technology.

But these funds did not go toward advertising, user subsidies, or stock buybacks—all of the usual paths tech companies have taken to spend money over the past two decades. This time, none of them were followed.

This money turned into concrete, steel, copper wire, transformers, and cooling pipes.

Amazon's $200 billion capital expenditure guidance for 2026 means it will spend $550 million per day, $23 million per hour, and $380,000 per minute. Microsoft has announced it will invest $10 billion in AI infrastructure in Japan alone between 2026 and 2029.

This is not the expansion pace of a software company; this is infrastructure.

But the essence of infrastructure is to make a company more asset-heavy.

The construction timeline, investment scale, and operational complexity of a large data center are on the same order of magnitude as those of an automobile assembly plant or a semiconductor wafer fab—requiring a full suite of processes including site selection, environmental impact assessment, power connection agreements, water supply assurance, and physical security.

GPUs play a role in AI similar to high-end machine tools in manufacturing—expensive, capacity-constrained, and rapidly depreciating; chips purchased at great cost today may become obsolete in two to three years, but you can’t wait, because your competitors won’t.

Electricity has become a strategic resource; a large AI data center consumes as much power as a medium-sized city, prompting giants to sign long-term power purchase agreements, invest in nuclear power, and negotiate dedicated power lines with utility companies.

Cooling water has begun competing with residents for water rights, and many communities in drought-stricken areas have found an unexpected newcomer on their water usage rankings.

These scenarios would have been impossible for a tech company twenty years ago. Site selection negotiations, grid connection, water rights disputes, local tax incentives—these are things railway companies, power companies, and oil refineries used to handle. And the last time financial instruments like century bonds, fifty-year bonds, and cross-currency issuances were used so intensively was during the great era of railway and telecommunications construction.

Looking at the balance sheets and cash flow statements for 2026, these companies' numbers are now much closer to those of TSMC, Duke Energy, or Union Pacific Railroad than to their own from a decade ago.

This brings us to valuation. In the past, investors priced tech giants based on the core assumption that marginal costs decline—adding one more user or one more ad incurs almost zero incremental cost, allowing profit margins to expand continuously. But this is not the case for AI’s infrastructure layer. Every additional model trained, every additional inference cluster deployed, and every new data center built requires real capital investment. Whether that investment pays off depends on whether customers are willing to pay, how model efficiency evolves, and how the competitive landscape shifts.

But all of these are uncertain.

It’s more like semiconductors, where each new process node requires larger fabs, and returns depend on yield and market demand. It’s also like electricity, where capacity is invested upfront, and returns depend on electricity prices and consumption. It’s even like railways, where tracks are laid first, and returns depend on whether the economies along the route can develop.

Therefore, as tech giants' financial structures increasingly resemble those of capital-intensive companies, the market's valuation multiples for them will eventually converge toward those of capital-intensive companies.

Some might say that once the infrastructure is built, we’ll return to a lightweight asset model. That’s naive. Railroads have been under construction for over a hundred years; power grids have never stopped expanding; semiconductor wafer fabs require upgrades every few years. Infrastructure for general-purpose technologies has never been, and never will be, “finished.”

AI may not be the continuation of the internet, but rather a resurgence of industrial capitalism, dressed in code and built on concrete foundations. The internet took twenty years to free tech companies from gravity; AI has pulled them back down in just two.

In 1840s Britain, railways were the AI of that era—freight jumped from a few miles per hour by horse-drawn carts to dozens of miles per hour by train, with an extraordinary leap in efficiency.

Capital flooded in. In 1846, the total authorized investment in British railways amounted to about £600 million, while the UK’s entire annual GDP at the time was only £500 million—a nation betting more than a full year’s national income on a single new technology. Today, that would be equivalent to the United States investing over $25 trillion into AI.

Early railways were primarily funded through stock sales, with buyers investing based on their vision of the future. As construction scaled up and returns failed to materialize, the quality of later projects declined, making equity financing insufficient—debt financing then stepped in. Railway companies began issuing bonds, using the future revenues of unfinished lines as collateral. Financing became increasingly aggressive, expanding from domestic borrowing to international markets.

It wasn't railway technology that killed the boom—it was interest rates. In 1846, the Bank of England tightened monetary policy due to grain imports and gold outflows caused by the Irish Famine, which had nothing to do with railways. But interest rates don’t care about causes—they only kill borrowers with the most fragile cash flows. Railway stocks collapsed, and countless railway companies went bankrupt.

Fortunately, the railways themselves remained. The tracks, stations, tunnels, and bridges did not disappear just because investors lost money. They were acquired at a discount by subsequent operators, integrated into a cohesive system, and ultimately became the lifelines of the British Industrial Revolution. The rise and fall of cities, the layout of industries, and the movement of populations were all realigned along the railway lines.

Twenty years later, the same drama played out across the Atlantic. After the American Civil War, the federal government encouraged western railroad construction through land grants and loan guarantees. Over 35,000 miles of track were built during the boom, with railroad bond yields ranging from 6.4% to 6.7%, making them the most attractive fixed-income instruments of the time. Capital flowed from the East Coast and from Europe toward the wilds of the American West.

In 1873, Jay Cooke & Company, once a major financier of the Northern Pacific Railway and one of the largest investment banks in the United States, declared bankruptcy. The ripple effect ultimately led to the collapse of 18,000 businesses within two years and the bankruptcy of 89 railroads over the next six years.

But the American railway network was eventually built. It served as the physical foundation for the United States becoming a super industrial power in the 20th century. However, the people who built the railroads were not the same as those who ultimately profited from them.

Similarly, there is fiber optics.

In the late 1990s, the rise of the internet fueled immense optimism about bandwidth. Telecom companies went on a frenzy to lay fiber optic cables, connecting not just cities, but continents and crossing oceans. Between 1996 and 2001, U.S. telecom companies issued over $500 billion in new bonds to finance this build-out, burying tens of millions of miles of cable underground and sinking them beneath the sea.

The pace of deployment far outstripped demand. When the bubble burst, only about 5% of the fiber optic cable laid across the U.S. had been activated with equipment and was transmitting data. The remaining 95% were "dark fiber," lying underground, waiting for a future that never came.

WorldCom, the second-largest long-distance telephone operator in the United States, with $107 billion in assets, filed for bankruptcy in 2002—the largest bankruptcy in U.S. history at the time. Global Crossing, which built one of the world’s largest fiber-optic networks, also collapsed that same year. Winstar, 360networks, McLeodUSA—a string of names fell victim to excess dark fiber.

But the fiber optic cables ultimately remained. Those undersea cables and metropolitan networks, ridiculed in the 1990s as overbuilt, became the backbone of the entire internet economy over the next two decades. Netflix streaming, Google searches, and Amazon’s cloud all run on those fiber optic cables—or their upgraded versions.

The same chain of logic repeatedly appears throughout these three historical periods.

First, the technology itself is real. Railroads are indeed faster than horse-drawn carriages, fiber optics are indeed faster than copper wires, and AI can indeed do things that were previously impossible. No one denies the value of the technology itself after the fact.

But the construction speed far exceeds short-term demand, because competition among peers doesn’t allow anyone to stop and wait for demand to catch up. You believe this is a winner-takes-all game, where the first to build locks in customers and ecosystems, so you have to keep running.

Everyone was running, leading to collective overbuilding. To support the pace of overbuilding, financing became increasingly aggressive—equity wasn’t enough, so debt was added; short-term wasn’t enough, so long-term was used; local currency wasn’t enough, so foreign currency was adopted. This applied to railways, fiber optics, and similarly to Swiss franc bonds, yen bonds, and century bonds.

What ultimately triggers an adjustment is rarely a technical failure, but rather a change in financial conditions. In 1846, it was rising interest rates; in 1873, it was the collapse of investment banks triggering a credit chain breakdown; in 2001, it was the dot-com bubble bursting alongside a recession. Technology continued to advance, but companies couldn’t hold on.

In the end, the infrastructure remained, but a significant portion of the builders did not. The beneficiaries of the railroads were the cities and factories along the routes, not necessarily the original shareholders of the railroad companies. The beneficiaries of fiber optics were Google, Netflix, and Amazon, not the bondholders of WorldCom.

Of course, today’s tech giants cannot be directly equated to 19th-century railroad barons or 1990s telecom entrepreneurs. The difference lies in the fact that today’s companies have massive, still-growing core business cash flows: Amazon has AWS and e-commerce, Alphabet has search and YouTube, Meta has the world’s largest social advertising network, and Microsoft has Office and Azure.

They are not startups that started from scratch and burned through investors' money to build data centers; they are giants squandering their own future while generating real profits.

So the question is whether the return period of capital expenditures can outpace the debt repayment cycle. Railroads are great, but borrowing money for six years to build a line that won’t break even for twenty years can still be deadly. Fiber optics are great, but borrowing for five years to lay cables that are only 5% utilized won’t save your balance sheet either.

AI data centers are certainly a good thing. But how much AI revenue would be needed to break even on $200 billion in annual capital expenditures? How many years would it take to recoup the total $700 billion investment? If model efficiency improves faster than expected—for instance, a new architecture requires only one-tenth the computing power for the same tasks—could today’s expensive computing infrastructure become the new dark fiber?

Go back to that century bond.

The institutional investor that bought it—perhaps a Swiss pension fund, perhaps a British insurance company—made a decision that day to lend money to Alphabet, with repayment due a century later.

This decision is rooted in a set of beliefs: that AI will be widely adopted, that Alphabet will outlast this competition, that its search and advertising businesses will continue to generate revenue, that its data centers will be fully utilized, and that no global catastrophe over the next century will destroy the company.

Holders of Amazon’s fifty-year bonds have roughly the same length of belief chain in their minds. Holders of Meta’s bonds accepted a record CDS premium, but their chain is shorter, as the market’s credit window for Meta is clearly narrower than for the other companies.

The chains vary in length, but what they’re buying is the same thing. It’s not GPUs, not data centers, not fiber optics or transformers—those are intermediaries. What they’re truly buying is time.

AI models are becoming increasingly homogeneous. Open-source models are catching up to proprietary ones, and smaller models are closing the performance gap with larger ones. Before this window closes and everyone can run nearly identical models, whoever gets ahead in deploying computing power and locks enterprise customers into their cloud platform will turn their temporary technological lead into a lasting business moat.

So the giants aren’t betting on “whose model is the smartest,” but on a more fundamental question: Before AI capabilities become widely accessible, can I build infrastructure and customer relationships so large that others can’t catch up?

This is time arbitrage: using today’s low-cost capital to purchase tomorrow’s market position.

Arbitrage over time has a harsh premise: the future must arrive on schedule.

The four companies face different time pressures.

Amazon is the most urgent case—its free cash flow has been consumed by capital expenditures, leaving only $1.2 billion. AWS’s AI service revenue must scale up within the next two to three years, or debt pressure will seep from the balance sheet into the income statement.

Meta is the most vulnerable: while social advertising is highly profitable, there’s a missing bridge between its AI infrastructure and its commercialization. Azure and AWS can directly sell computing power to enterprise customers, but after spending over $100 billion building its infrastructure, Meta still hasn’t clearly defined what product it’s creating, who it’s selling to, or how it will charge—this story remains unfinished, and market impatience is already reflected in its stock price and CDS.

Alphabet is the most composed, as its search and YouTube businesses generate consistent revenue with minimal maintenance—even if AI yields no short-term returns, its core operations provide a safety net. The market has granted it a century-long credit line, giving it the longest time horizon among the four. But its $185 billion in capital expenditures is 2.5 times last year’s amount; the very acceleration is eroding patience. Composure does not equal safety.

Microsoft is the clearest case, with its deep integration with OpenAI making Azure the direct beneficiary of AI commercialization. Copilot is already monetized, and GitHub Copilot is one of the most widely paid-for AI products among developers, offering the shortest path from infrastructure to revenue. However, a $190 billion capital expenditure means that even with a clear path, the scale of the bet is so large that everything must go exactly as planned to break even.

All four are betting on the same thing: borrowing money for the future to build something that isn’t yet fully understood today, hoping that use cases will explode before the debt comes due.

This path has been traveled by railways and by fiber optics. Each time, technology ultimately proved its value, and the infrastructure remained. But each time, there were also those—sometimes many—who paid for its construction but never lived to see the payoff. The technology was right; the timing was wrong. And financial markets don’t give wrong timing a second chance.

No one knows if AI’s “future” will arrive on schedule. The only certainty is that some of the world’s most conservative capital has signed a contract with Silicon Valley by purchasing century-long, fifty-year, and forty-year bonds.

The terms of the contract are simple: we lend you our time, and you repay us with the future.

No one can say for sure whether the contract will be honored in the future.

Click to learn about the open positions at BlockBeats

Welcome to the official BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group: https://t.me/BlockBeats_App

Official Twitter account: https://twitter.com/BlockBeatsAsia