Investors placed €53 billion of orders for €9 billion of Meta Platforms bonds this week, a ratio of nearly six to one that made the company’s first euro-denominated sale the most heavily subscribed corporate bond in European history.
Bloomberg reported the details on November 4. The deal was split into six tranches with maturities running from four years to 38 years, a structure that lets the issuer price each slice against a different part of the yield curve and lets buyers choose how much duration they want to hold. The order book exceeded the record set by AB InBev in 2016, a sale that itself set a record during the years when borrowing costs were close to zero.
The euro sale followed a dollar deal the previous day. Meta raised $15 billion in the U.S. market and drew $58 billion of orders for it. Two days, two currencies, about $24 billion raised in total, and more than $110 billion of orders recorded across the pair. For a company that generated tens of billions in free cash flow in its most recent fiscal year, the sums are not a response to a funding emergency. They are a decision about how to pay for computing.
That decision is now common across the largest technology companies. Data centers, power contracts, and the accelerators that fill them have moved from capital expenditure lines that could be funded from operations to commitments that stretch years into the future. Debt is the instrument that matches those commitments, because a bond can be sized to the life of the asset in a way that operating cash flow cannot.
The pricing has changed. Credit spreads on the bonds of the largest cloud and platform companies have widened in recent months, according to market participants, as investors have begun to distinguish between the businesses that generate the cash and the projects that consume it. A wider spread means the same bond costs more to issue, and it means the marginal data center has to clear a higher hurdle before it earns its cost of capital.
The 38-year tranche is the part of the structure worth watching. A bond maturing in the 2060s asks investors to underwrite not the current generation of models but the assumption that demand for computing keeps compounding for four decades. Buyers who take that duration are expressing a view about the technology, not about the company’s balance sheet, and they are being paid a fixed coupon for a risk that has no precedent to measure against.
Issuers rarely choose a currency by accident, and the euro market offered specific advantages. European institutional buyers, particularly insurers and pension funds with long-dated liabilities, have a structural appetite for duration that the dollar market does not always match, and issuing in euros lets a U.S. borrower reach that base without competing for the same investors as every other hyperscaler in dollars. Currency swaps convert the proceeds, so the funding cost ends up compared against the dollar curve rather than the euro one.
There is a simpler reading too. Order books of this size are partly a measure of how much cash is looking for a home in investment-grade credit, and how few issuers of Meta’s quality are offering paper. The demand says less about Meta’s prospects than about the scarcity of borrowers with its ratings and its cash generation.
The comparison that comes to mind is the telecom buildout at the end of the 1990s, when carriers borrowed against projections of internet traffic and some of them were right about the traffic and wrong about the timing. The difference is that Meta, Microsoft, Amazon, and Alphabet are funding their spending from a position of profitability, and the debt is a slice of the capital structure rather than the whole of it. The similarity is that both waves were justified by a forecast of demand that had not yet arrived.
For the bond market, the practical effect is a new benchmark. Once a borrower of this size establishes a euro curve across six maturities, the next issue prices against it, and every smaller technology company that wants European duration now has a reference point for what the market will charge. That reference currently looks generous, which encourages more issuance, which is how a record order book becomes an ordinary one.
None of this is a warning by itself. The open question is what happens to spreads if the spending continues and the revenue that justifies it arrives later than the models assume. A single record-breaking sale is a sign of appetite. A year of them, priced at widening spreads, would be a sign that investors have started to price the risk they are taking, and the companies doing the borrowing would have to answer with something other than another tranche.
