Something shifted in Silicon Valley this year, and it is not about models or benchmarks. It is about money. Specifically, where Big Tech is getting it.

For decades, the biggest tech companies funded their investments from cash flow. Apple sits on $160 billion. Alphabet generates $100 billion a year. Amazon prints money from AWS. The whole identity of Big Tech was built on not needing Wall Street's permission to spend.

That era is over. The AI infrastructure race is so expensive that even the richest companies on Earth are borrowing to keep up.

On Monday, Alphabet disclosed plans to sell Japanese yen-denominated bonds for the first time in its history. The issuance is expected to total several hundred billion yen. This comes just days after Google's parent raised nearly $17 billion through two bond sales: a 9 billion euro issue and a $6.2 billion dollar issue. Last February, Alphabet sold a rare 100-year bond worth 1 billion pounds. A century bond. From a company that did not exist 30 years ago.

Amazon is doing the same thing in Swiss francs. The company is preparing a six-part Swiss franc bond offering, its first in that currency. It already has $122 billion in outstanding debt.

Oracle is raising $45 to $50 billion in a mix of debt and equity this year alone. And it is getting sued for it. Bondholders filed a lawsuit in January claiming Oracle failed to disclose how much additional debt it would need to build out AI infrastructure, causing losses on existing bonds.

Salesforce priced a $25 billion debt offering in March. The combined AI infrastructure spending of Alphabet, Amazon, Microsoft, and Meta is now projected to exceed $700 billion in 2026, up from roughly $600 billion previously expected. Bridgewater Associates, the world's largest hedge fund, warned in February that the AI boom has entered a "more dangerous phase" marked by "exponentially rising investments in physical infrastructure and growing reliance on outside capital."

Let that sink in. Bridgewater called it dangerous. Not exciting. Not promising. Dangerous.

The reason is straightforward. Cash flow funds reversible bets. Debt funds commitments. When you borrow billions to pour concrete for data centers, you cannot unpour the concrete if the revenue does not materialize. And the revenue question is real. We reported last cycle that the Fed flagged AI as a standalone financial stability risk for the first time. Fifty percent of market professionals surveyed said AI threatened financial stability. Cloud firms raised $100 billion in Q1 bonds alone.

The currency diversification tells its own story. Alphabet is not selling yen bonds because it loves Japan. It is arbitraging interest rate differentials. Japanese rates are near zero. US rates are not. Borrowing in yen is cheaper. Amazon's Swiss franc play follows the same logic. This is financial engineering layered on top of physical engineering, and the combined leverage is staggering.

The optimistic read: these companies have the revenue and credit ratings to borrow at near-sovereign rates, and AI infrastructure will generate returns that dwarf the interest costs. The pessimistic read: the last time tech companies borrowed this aggressively to fund infrastructure was the dot-com telecom buildout, and that ended with WorldCom and Global Crossing in bankruptcy.

The truth is probably somewhere in between. But the fact that Bridgewater is using the word "dangerous" and the Fed is flagging AI in its financial stability report suggests the adults in the room are paying attention.

First reported by Reuters and Bloomberg.