There is an obvious question to ask when a technology company sells bonds in the middle of an AI build, and it is the wrong one. The obvious question is whether the borrowing means the spending is going up. It leads nowhere here: a debut bond in a new currency is a decision about where the money comes from, not how much is being spent, and this one is far too small to move a capital budget of this size either way. The better question is harder. Why did a company that told its own underwriters in July it was finished borrowing for the year come back in September, in a currency it had never used? And what does the order book say about how much more of this the market will absorb?
Amazon priced its first sterling bonds on 9 September 2026 — £4.25 billion in four tranches, at spreads of 53 to 93 basis points over gilts. Two months earlier, raising $25 billion in July, it had told the banks running that deal it would issue no further debt in 2026. The book was covered about 2.5 times. Alphabet's sterling debut in February was covered roughly twice as heavily.
The pre-open wires said "at least £3bn". That was the floor; the book closed 42% above it.
The points
- Four tranches, £4.25 billion, on the terms in the table above, trading 9 September and settling on the 14th. All of it is from the pricing term sheet Amazon filed with the SEC that day — the coupons left blank in the preliminary prospectus are on the record, not in lead-manager colour.
- £4.25 billion is $5.77 billion at the European Central Bank's reference rates on the pricing date — 2.6% of the capital expenditure Amazon has guided to for 2026.
- The order book was £10.65 billion, down from a peak near £12 billion as the price tightened: 2.51 times covered at the final book, 2.82 at the peak.
- The prospectus earmarks nothing. Net proceeds go to "general corporate purposes", a list running from "repayment of debt" through "capital expenditures" to "repurchases of outstanding shares of our common stock." Buybacks sit there on equal footing with data centres. Calling this AI funding is an analyst's characterisation, not Amazon's.
- Amazon stopped at nineteen years. Its longest tranche matures in 2045, a weighted-average life of 9.6 years. Alphabet's February book ran to 30.6 years and included a 100-year bond.
- The interest is £249.5 million a year — about $338 million — on a weighted-average coupon of 5.871%. Amazon's interest expense was already $1,314 million in the June quarter alone, against $516 million a year earlier.
- The closing filing has not landed. No Item 8.01 8-K yet; on each of Amazon's three previous 2026 deals it arrived within two days of settlement, pointing to 14–16 September.
What £4.25 billion actually buys
Nine and a half days.
Andy Jassy raised the 2026 plan to approximately $220 billion of cash capital expenditure on the 30 July earnings call, up from the roughly $200 billion that triggered February's selloff, and named the reason: memory costs. Against $220 billion, the entire debut sterling book funds 2.62% of the year — about 9.6 days of spending at the annual rate, or 10.8 days against what Amazon actually spent in the first half.
Those two figures sit on slightly different bases. Jassy's $220 billion is cash capex, net of proceeds from equipment sales and incentives; Amazon's first-half purchases of property and equipment were $98.4 billion gross, or $96.3 billion net of $2.1 billion of such proceeds, which is the basis he is quoting. Reaching $220 billion on that basis needs a second half 28% heavier than the first.
So this is not a capex signal. It is a funding decision, and that is where the numbers get interesting.
Amazon has opened five currencies in six months
Every SEC-registered deal below comes from Amazon's own pricing term sheets. The Swiss franc leg is the exception — documented in Switzerland, absent from the SEC's system, and the one line sourced to the press.
Each deal in its own currency, converted at ECB reference rates on its pricing date
| Priced | Raised | US$ equivalent | Covered |
|---|---|---|---|
| 10 Mar · dollars | $37.0bn | $37.00bn | 3.41x |
| 11 Mar · euros | €14.5bn | $16.79bn | 1.92x |
| 12 May · francs | CHF 2.82bn | $3.61bn | — |
| 8 Jun · Canadian | C$14.0bn | $10.05bn | — |
| 7 Jul · dollars | $25.0bn | $25.00bn | — |
| 9 Sep · sterling | £4.25bn | $5.77bn | 2.51x |
| 2026 to date | five currencies | $98.21bn | — |
The dollar, euro, Canadian dollar and sterling deals are taken tranche by tranche from Amazon's own pricing term sheets filed with the SEC under registration 333-293246 on 10 and 11 March, 8 June, 7 July and 9 September 2026. The Swiss franc deal was documented in Switzerland, is not on the SEC's system, and its CHF 2.82bn size is Bloomberg's reporting of 12 May. Order books, and so every coverage figure, are press-reported by Bloomberg and Reuters rather than filed; no book was reported for the franc, Canadian dollar or July dollar deals. Dollar equivalents and the $98.21bn total are our arithmetic at European Central Bank daily reference rates on each pricing date, so the total mixes rates across six months and is not a single-date figure.
Two things fall out of it. Amazon has borrowed $98.2 billion in 2026 against a $220 billion capital programme — new debt is funding close to half of this year's build. And the sterling deal, for all the attention a debut attracts, is 5.9% of that borrowing and the smallest of the five.
The stack behind it has roughly doubled in six months: long-term debt went from $68.4 billion at the end of 2025 to $132.2 billion at 30 June. This deal adds 4.4% to that.
The coverage ratio is the only forward-looking number here
| Deal | Raised | Order book | Covered |
|---|---|---|---|
| March, dollars | $37.0bn | $126bn | 3.41x |
| March, euros | €14.5bn | €27.9bn | 1.92x |
| September, sterling | £4.25bn | £10.65bn | 2.51x |
The sterling book beat Amazon's record euro debut and trailed its dollar deal. Against Alphabet, in the same market seven months earlier, it looks thin: £5.5 billion on roughly five times cover against £4.25 billion on two and a half.
The tenor says it more sharply. Alphabet sold a £1 billion 100-year bond in February at 6.125%, just under par, into demand reported at ten times the tranche. By 23 July it was quoted at 89.98 pence — a buyer at issue down more than a tenth of their principal in five months. Amazon stopped at nineteen years. Given what happened at the long end of sterling in between, that reads less like modesty than like pricing the lesson in.
What would change this reading
The counter-case survives the evidence. A book covered 2.5 times at 53 basis points over gilts is not a company struggling to raise money. And Alphabet's five times came with a century bond that gave duration buyers something they cannot get elsewhere; Amazon offered nothing past 2045, so a thinner book is partly the arithmetic of a shorter, plainer deal rather than a verdict on the credit.
A second sterling deal settles it: return in 2027 at similar cover and February's five times was the anomaly; return wider on thinner demand and the non-USD bid was already near its limit.
The July pledge deserves its exact weight, too. It was never a public commitment — it was reported guidance to underwriters, and Amazon's spokesperson said only that the company regularly evaluates its operating plan and makes financing decisions accordingly. What happened in between is that the capital plan grew by $20 billion on 30 July, three weeks after the pledge. This sterling book covers 29% of that increase.
What to watch
- The Item 8.01 closing 8-K, due around 14–16 September, confirming net proceeds and final terms.
- Whether a sixth currency follows. Five in six months is a pattern, and the pledge meant to stop at four binds nothing now.
- The September-quarter capex line, which shows whether the second half is running at the 28% step-up $220 billion requires.
- Where the 2045s trade — the honest gauge of the duration bid Alphabet's century bond has been failing since February.
Sources and provenance: every tranche size, coupon, maturity, spread, benchmark gilt, price and settlement date is from Amazon's pricing term sheet filed with the SEC on 9 September 2026 under registration 333-293246, and the use-of-proceeds language is quoted from the preliminary prospectus supplement filed the same day; the 2026 issuance table is built the same way from Amazon's own term sheets for its March, June and July deals. The Swiss franc size, every order book, the coverage on Alphabet's February sterling issue and the July guidance to underwriters are press reporting — Bloomberg, Reuters and CNBC — rather than filed figures. Alphabet's sterling tranches are from its deal counsel's announcement, and the 89.98 pence quote on its 2126 bond is a 23 July indication. The $220 billion capex plan is Andy Jassy's own on the 30 July earnings call. Capital expenditure, interest expense and debt balances are as Amazon reported them. Dollar equivalents use European Central Bank daily reference rates on each pricing date, and every ratio, coverage figure, weighted-average life and days-of-capex conversion here is our arithmetic on those numbers.
