Alphabet Borrowed for a Century. The Market Is Starting to Blink.

Alphabet issued a bond that doesn't mature until 2126. And when the company sold it in February, investors didn't just buy it — the sterling tranche drew nearly eight times the amount on offer.

The funny part isn't the maturity. Governments have been issuing century bonds since the 19th century. The funny part is what the bond doesn't have: standard investor protections. There is no change-in-control covenant, which means if Alphabet gets acquired, restructured, or fundamentally transformed, the bondholders can't demand early repayment. For an investment-grade corporate bond, that's unusual. For a 100-year one, it's an assumption that the company will not only survive but stay creditworthy through five presidential administrations and several technology cycles we can't yet imagine.

The basic point is that the Alphabet bond deal in February read less like a financing transaction and more like a coordination event. The kind where everyone looks at everyone else and decides the price has to go down because the asset is just that good.

Here's what happened in February, in order.

Alphabet announced a $20 billion U.S. dollar bond offering. It was quickly upsized because demand was strong, leading to the upsizing. The 40-year tranche saw its spread compress by 25 basis points during bookbuilding, meaning buyers competed so aggressively that Alphabet paid less than it initially planned to.

Then came the international tranches. A 100-year sterling bond, sized at £750 million ($1.03 billion at the time), drew £5.75 billion in orders. Swiss franc bonds joined in. Total raised across all currencies: roughly $32 billion.

Pension funds and insurance companies were the main buyers. That makes sense on the liability side — these institutions have obligations stretching 50 or 60 years into the future and need long-duration assets to match them. Alphabet's century bond was basically tailor-made for a UK pension fund that needs something to sit in its portfolio until 2080 and didn't want to buy another gilt.

But the covenant structure adds a wrinkle. Reuters highlighted what the bonds lacked, using the phrase "lack of guardrails". The change-in-control covenant is the protection investors normally get — if the company is sold or its ownership structure changes materially, bondholders can call the debt early. Alphabet left it out. Credit analysts told reporters this is standard for hyperscalers now, because they want flexibility and investors don't think they'll need the protection anyway.

The cleanest way to think about that is: the bondholders are betting Alphabet is too big and too entrenched to lose its investment-grade status over the next century. Which is a reasonable bet over five years. It's an interesting bet over 100.

The older comparison is railroad bonds from the 1870s or utility financing from the postwar buildout. Those were massive capital projects funded by long-dated debt, with investors buying into the idea that infrastructure spending today creates predictable cash flows tomorrow. The difference is that railroads and utilities were usually regulated monopolies with visible, stable revenue streams. Alphabet is betting that AI will create something analogous. The bondholders are betting that Alphabet's bet will work. Those are not the same thing.

Now here's the thing that the February headline story doesn't capture, because it's happening in real time.

Alphabet is back in the market today, August 6, seeking to raise up to $25 billion more in U.S. dollar bonds. This is expected to be its last offering of the year. And the reception looks different. Spreads on long-dated hyperscaler bonds have widened.

Several things shifted. The July selloff that preceded this offering was the bond market's way of asking a question equity investors have been ignoring: at what point does capex guidance outpace conviction?

The simplest model is this: in February, hyperscaler debt was a scarce asset. Pension funds needed duration, tech credit looked pristine, and the supply of high-quality long bonds was thin. The premium wasn't in the coupon — it was in the access. You had to bid aggressively just to get allocation.

Now the supply problem has become the demand problem. All four major hyperscalers have issued roughly $194 billion in bonds through late July of this year, already dwarfing the full-year total from 2025.

When a class of bonds goes from "everyone wants some" to "the book barely covers the deal," something in the calculation has changed. Either the price is wrong, the duration is wrong, or the credit is wrong. The market hasn't decided which one it is yet, which is why spreads are widening but haven't run away.

The covenant-light structure that investors accepted in February now looks more interesting in hindsight. If the hyperscaler debt trade goes south — if capex spending overwhelms cash flow and credit quality actually degrades — the bondholders won't have the change-in-control protections that would have given them an exit ramp. They're stuck with the duration, the widening spreads, and whatever recovery they can negotiate. The equity holders, meanwhile, get all the upside if AI monetization accelerates. It's a funding model where the borrowers keep the flexibility and the lenders keep the maturity.

Alphabet's stock is down less than 1% today, trading around $359. The equity market still seems to believe in the payoff. The bond market, which used to agree enthusiastically, is at least asking for more compensation to keep the faith. That divergence — between the people buying the dream and the people financing the buildout — is the real story the August offering tests.

The structural point is straightforward. The hyperscaler debt machine worked beautifully while supply was scarce and conviction was high. It's now running into the oldest constraint in fixed income: when the amount of debt being sold exceeds the amount of comfortable duration that buyers actually need, someone has to pay more, someone has to back away, or someone has to rethink what they thought they knew about the credit. Today's $25 billion test finds out which one it is.