One rule changed. Borrowing went up fivefold.

The bill came. My power bill, and I read the number twice before I paid it.

It went up again. Not by much. Enough that I noticed.

I never used to think about who pays for the wires. The poles on my street, the substation down the road, the cable under the pavement. Somebody buys all of that first. Then they charge me a little every month, forever.

Those companies are borrowing more than they ever have. They need new lines and new generation, and most of the new demand comes from data centres built for artificial intelligence.

But a company can only borrow so much. Past a point the rating agencies call it risky, and then every future loan costs more. So there is a ceiling.

Unless you borrow something that doesn't get counted.

In February 2024, Moody's changed one rule. It doubled the equity credit it gives American hybrid bonds, from 25% to 50%.

Nothing about the bond changed. Only the counting changed.

A hybrid is a bond. It pays a coupon. The tax authorities treat that coupon as interest, so the company deducts it from its taxable income. Investors hold it in bond funds and bond indices.

After that one rule change, half of it stopped counting as debt.

Interest to the tax man. Equity to the rating agency. The same piece of paper, counted twice.

Treasurers did the arithmetic fast. American issuance of these bonds went from about $6.7 billion in 2023 to $32.4 billion in 2024, according to ICE's fixed income data team.

Utilities and energy companies did most of it. They are the ones with the capital plans.

So why does a bond get to be half equity? Because of two features written into the document.

The first is that it lasts a very long time, often thirty years, and the company alone decides when to repay it. The second is that the company can stop paying the coupon for years without triggering a default. TC Energy's filings this year describe a deferral right of up to ten years on notes it sold in April.

In other words, the equity part is an option the company holds and the investor wrote.

Nobody uses it. Nomura Asset Management's own paper on the sector says that as far as they know, no investment grade issuer has ever missed a hybrid coupon. That record is the whole argument for owning them.

It is also the same record every safety valve has, right up until the first one opens.

There is a second option in there too. Investors price these bonds to a first call date five to ten years out, and assume the company repays them then. The company decides that as well. American hybrids, unlike the European ones, mostly carry no coupon step-up to punish a company that leaves the bond outstanding.

So the pressure to repay is an accounting pressure, not a contractual one. Leave it outstanding and the rating agency takes the equity treatment away. That is the discipline. That is all of it.

I have no idea when any of this matters. Probably not this quarter.

But the Federal Reserve raised its policy rate yesterday to a 3.75% to 4% range, the first increase since July 2023, and the ten-year Treasury sat around 5%. Higher rates change what it costs to replace one of these bonds. So the option gets more valuable to the company and more expensive for whoever sold it.

And I keep thinking about my bill on the kitchen table.

This connects directly to the borrowing. A utility does not raise capital for fun. It raises it against a plan, and the plan right now is data centres.

CreditSights did the simple version of the arithmetic. On the supply side they added up the firm grid connection commitments that utilities themselves report. That came to 112 gigawatts by 2030.

On the demand side they took the outside forecasts, from Bloomberg, S&P Global and the investment banks, and landed near 94 gigawatts for the same year.

But about 35 gigawatts of that demand already exists and is running today. So the growth everyone forecasts is closer to 59 gigawatts, and the utilities have committed to roughly twice it.

Not everyone agrees. Janus Henderson went through the projects one by one and argued that permits, equipment and labour mean only about 85 of the 157 gigawatts announced will ever energise. Their worry is under-delivery, not oversupply.

Both sides can't be right. And the spending happens either way.

And here's where it spreads.

US utility capital spending was $104 billion in 2015. It reached $208 billion in 2025, on Edison Electric Institute figures, and CreditSights expects around $230 billion this year.

Across 2026 to 2030 the industry's own plans point to something near $1.3 trillion.

Here is the part most people miss. A regulated utility earns a set return on every dollar it puts into the ground. Spending is not a cost to it. Spending is the business model.

So the incentive runs one way. Build it, add it to the rate base, collect a return on it from everybody connected to the wires.

Regulators have noticed. More than a dozen states have approved special data centre tariffs with up-front collateral, minimum volume commitments and exit fees, so the tenant pays if it walks away. California and Texas have more legislation coming.

Those protections vary state by state. Mine are whatever my state decided. So are yours.

Meanwhile, across the Atlantic.

Europe invented this market and still dominates it. Issuance reached nearly €85 billion in 2025, the busiest year in a decade, and by mid-July 2026 the year had already passed €51 billion. Outstanding paper globally is above €350 billion.

France and Germany lead the issuer list. Utilities lead the sectors, for the same reason they lead everywhere: heavy spending and a rating they don't want to lose.

Then there is the March deal that shows how far the counting can stretch.

The satellite operator SES sold €650 million of subordinated perpetual notes at a 7.375% coupon. The notes carry a junk rating of Ba3. And on the company's own announcement, they were expected to receive 100% equity credit from Moody's.

A bond, paying 7.375%, sold to bond investors, counted as entirely not debt. The order book was five times oversubscribed.

Everyone involved knows exactly what it is. That's what interests me.

Three more things worth keeping track of this week.

Gold and oil have split. Gold sat near $4,283 an ounce this month and drifted lower while Brent ran hard. That pairing is odd. An oil-driven inflation scare is usually the best week gold gets, and instead a stronger dollar and rising real yields have been pushing it the other way. Saxo's desk flagged the same split. So one of those two markets is telling a story the other one doesn't believe.

New Zealand raised rates again. The Reserve Bank lifted its cash rate by a quarter point to 2.75% in September, the second consecutive increase, with energy prices named as the reason. A small open economy hiking twice in a row is the sort of thing nobody outside Wellington reads about. But small open economies feel imported energy inflation first, because they import all of it.

Europe is paying more for gas in a weaker currency. Dutch wholesale gas traded at €83.77 per megawatt hour on 14 September, its highest since 2023, and the euro slipped to about $1.15 the same day. Europe buys that gas in dollars. So the bill goes up twice, once on the commodity and once on the exchange rate, and a higher energy bill is exactly what weakens the growth case for the rates a currency needs. We'll see.

Small banks had a problem. They wanted to grow, and growing meant raising capital, and nobody wanted shares in a four-branch bank in Ohio.

So in the mid-1990s Wall Street built them something. It was called a trust preferred security.

The bank set up a trust. The trust sold securities to investors. The bank paid the trust, and the trust passed the money on.

The tax authorities looked at those payments and called them interest. So the bank deducted them. The bank regulators looked at the same instrument and called it Tier 1 capital, the cushion that stands between a bank and failure.

One piece of paper. Deductible at one desk, capital at the other.

The market grew past $60 billion before 2008. And there was one more feature in the documents. A bank could stop paying for up to five years without it counting as a default. Everyone described that as a safety valve nobody would ever pull.

Small banks couldn't sell these one at a time, so the investment banks pooled hundreds of them together and sold slices of the pool.

By February 2010, more than a quarter of the banks inside those pools had stopped paying.

Fitch ran an index on it. The combined default and deferral rate passed 27%.

The FDIC studied what had happened. Banks that used these securities as regulatory capital turned out to be weaker, to take more risk, and to fail more often than the banks that never touched them.

Congress ended it. The Collins Amendment, Section 171 of the Dodd-Frank Act signed in July 2010, stopped the securities counting as Tier 1 capital, and they were phased out over the years that followed.

Nothing about them was fraudulent. Every term was printed in the prospectus. The deferral clause did exactly what it said it would do, and the pooling was disclosed, and the ratings were issued in good faith.

The paperwork was fine. The capital was not.

Because a thing is only capital if it absorbs a loss. If the company still owes it, and still deducts it, and only stops paying when it is already in trouble, then it was borrowing all along and the ledger was the only place it looked otherwise.

We can't know when a rule like that gets tested again. Nobody can.

Counting is not cushion. We'll see.