Every few years, the century bond returns — not exactly with a bang, but with a new, shiny calculation involved.
As reported by the FT and Bloomberg last month, Google’s parent company has been busy tapping the bond market for everything from 3-year to 50-year bonds in several major markets (US, Switzerland, UK). Now it’s trying for the ultimate prize: a century bond.
Century bonds are exactly what they sound like: a bond, often issued by a government or a long-lived institution like a university, that runs for a hundred years, often at interest rates somewhat above prevailing market rates or reference rates. For the issuer, they make a ton of sense, generationally and actuarially: They receive funds for investments right now, lock in financing costs for a long time, and face no financing rollover risk.
It’s more of a puzzle why buyers show up for these issuances, especially since the market participants have very recent examples of being seriously burned. As a bondholder of extremely long-dated bonds, you’re always living in financial terror, waiting for rates to rise and inflict multiplied damage on the market value of your investment. Since bond prices move inversely with interest rates, the effects are more pronounced the further into the future — the longer-dated — the bond is. The loss of market value on a hundred-year bond when interest rates increase is far greater than on a ten- or two-year bond. When it does, this “dangers of duration,” as FT journalist Robin Wigglesworth has called it, completely undermines your finances for decades on end.
The last two times century bonds popped out of academic obscurity, they got a well-deserved bad rap. In the 1990s, several large companies tried them — and locked in steep and expensive rates in the five-percent region while interest rates kept falling toward zero for decades. In the late 2010s, with ZIRP dominating the world’s financial markets and trillions of mostly government debt traded at negative yield, we started seeing new players dusting off this old idea, since money now was just so cheap to be had: Universities like UPenn, Virginia, Oxford, and Rutgers took in funds for a hundred years in the two to four percent range. Then the opportunistic governments (including Ireland, Belgium, Mexico, Argentina) also successfully placed century bonds at eye-poppingly low rates. The most extreme participant was Austria, whose perfectly timed century bonds — of 2.1 percent in 2017, and then 0.85 percent and even zero percent right on the cusp of the ‘rona inflation — saved its taxpayers a fortune.
Logic alone dictates that if you sell debt due in 2120 for zero percent or 0.85 percent a year, and then CPI (and thus your incomes and tax revenues) rise to upward of 10 percent for a few years, you’re doing great. Indeed, bond math logic made that exercise even more painful, with some of these placements trading at cents on the euro a few years after issuance, vaporizing bondholders’ money.
Enter Big Tech: The AI Investment Needs Meet Underwater Pension Systems
With Alphabet/Google’s placements in November, and now its Swiss franc and sterling placements, it’s the first time a big tech company has dared to come back to this perilous market since the 1990s.
What’s so odd about this hunger for long duration is that in the 2010s and during the pre-inflation pandemic years, at least bond investors collectively were starved for yield, ready to do anything to eke out a few extra basis points of return. In 2026, there are still positive yields to be had. The mystery, then, is not why Google issued but why investors rushed in.
The $20-billion USD placement finalized at treasury yields+95 bps, while the GBP bond of £5.5, about $7.5 billion, placed at 120 basis points above gilts; the £1 billion century bond was ten times oversubscribed, according to Reuters, and carries a coupon of 6.125 percent until 2126. With the fairly recent history of century bond investors burning obscene amounts of money on mistaken duration bets, the question remains: Why were these long-maturity debt placements massively oversubscribed, at rate spreads of only about a hundred basis points above safe assets?
Three candidate explanations.
First, defined-benefit pension funds are extremely hungry for long-dated debt. Reinsurers and life insurance providers, too. They have long-dated liabilities that rise and fall in (net present) value with interest rates; owning equally long-dated assets compensates somewhat for that. “Strong demand from UK pension funds and insurers has made the sterling market a go-to venue for issuers seeking longer-dated funding,” reported Tasos Vossos for Bloomberg.
Second, the large embedded rate-leverage is kind of capital efficient. Bond funds don’t buy individual issuances in isolation, but compose a portfolio with a combined desired outcome, constantly micromanaging exposure and duration vis-à-vis a benchmark index. Remarked Marcus Ashworth in an opinion piece,
Buyers of these types of security are looking to dynamically balance the risk of an entire portfolio rather than caring about annual coupon weights. Ultra-long debt allows portfolio managers to barbell their duration needs by buying more 10-year liquid debt rather than illiquid 20- to 50-year maturities.
Which is, perversely, why riding that zero-percent Austrian century bond all the way to the basement between 2020 and 2025 could have been beneficial to certain bond portfolios that paired it with other investments. When rates fall, these bonds soar, letting market participants ride the convexity game in a broad macro rate (and now also FX) game. Under functional monetary regimes with price predictability, long-dated debt is both efficient and common — think about the perpetual British Consol, the financial instrument that made the British Empire.
Plus, there’s a ton of capital-intensive bond market efficiency involved in going way out on maturity. Explains quant researcher Navnor Bawa discussing the news on his Substack:
A one percent decline in long-term rates generates approximately 40-50 percent capital appreciation on century bonds versus 15-20 percent on 30-year debt — making it the most capital-efficient duration exposure available for institutions positioning for lower long-term rates.
Third, this time is different and bond investors have inflation fatigue. Many might just believe that what happened during the pandemic was a once-in-a-generation one-off event and that central bankers will do better going forward. If inflation settles back near its more usual two to three percent, or indeed rates fall back toward zero in an easy/easier monetary policy regime, locking in long real duration today at slightly higher rates might look clever rather than reckless.
This might all work out great for the issuers, and bondholders will celebrate their contribution to the AI revolution… for a few years until we have the next bout of runaway inflation prints. Most likely, the investors in these oversubscribed bonds are setting themselves up for financial troubles, by the sudden jolt of interest rates jumping higher or the slow, gradual erosion of fiat purchasing power.
The spread over sovereign credit says something about the balance sheet strength and the optimism surrounding AI-related big tech investments, certainly with behemoths like Google. Funding operations at these rates show either incredible creditworthiness — with Google and the current wave of AI optimism, at least believable — or devastating mispricing.
Investors got the memo about AI capacity building needs, loud and clear. With collective bond market amnesia, that was about the only meaningful bit they took away.
