Three quarters of Goldman's issuance upgrade is debt the market has to fund with new money

Gross supply went up by $200 billion in the revision and net supply by $150 billion, so three quarters of the extra paper is debt the market has to fund with money it does not already hold ^[16][17][18]. That distinction is the whole mechanism, because a maturing bond hands cash back to the same account that buys the next one, and $2.3 trillion of gross against $1 trillion of net implies roughly $1.3 trillion of the calendar is that sort of recycling ^[19]. The other trillion needs a new bid.

Hold the 24% share flat across the full gross number and AI-related issuers alone would print about $552 billion ^[20]. That is the figure the European Central Bank's blog is arguing about when it says a large enough influx of tech paper can overwhelm finite investor capital and push borrowing costs up for companies with nothing to do with data centres, and for governments and agencies too ^[14][13].

Spreads have barely moved this year even as the quantity of AI-related debt has grown, which means the market is charging roughly the same price for a much larger pool of risk. The damage to total returns came instead from the government curve, 61 basis points in the US and 52 in Germany, a repricing of the risk-free leg rather than of the credit leg ^[8][9]. Both currencies are negative year to date, and Goldman expects rate moves, not spreads, to keep driving returns ^[11]. Price and value have separated: the quantity of AI capex credit outstanding has grown sharply, while compensation for holding it has stayed roughly where it was.

The euro leg is smaller than the noise around it suggests. US companies have about $46 billion of bonds outstanding in the euro zone, close to 10% of gross new issuance, which puts the implied euro gross calendar near $460 billion ^[12][22], and AI issuers are 6% of European volume against 24% in the US, a factor of four ^[5][21]. Amazon and Alphabet lead the issuer table there ^[12], and Goldman notes hyperscalers are increasingly tapping euro credit in a way that may narrow the euro's technical advantage ^[7].

The counter-case is respectable, and the same source carries it: these are strong balance sheets with real cash flows, which is why the paper clears ^[15], and Goldman's own 2027 estimate of $2.4 trillion is only $100 billion above 2026, about 4.3% ^[23], a plateau at a high level rather than an escalating borrowing curve. The material covers debt only, not equity, so the tidy claim that credit has displaced equity as the funding channel for the buildout goes beyond what is evidenced here; what the numbers do show is that a quarter of the US investment-grade new-issue calendar now has one theme behind it ^[4].

The thesis fails if the very busy September that Goldman's Thursday report anticipates clears at unchanged spreads ^[6], because that would make the ECB's crowding argument a theory rather than a price, and it fails on a slower schedule if net supply lands nearer the old $850 billion than the new trillion ^[2].