Goldman Sachs Asset Management is going underweight on the very companies it once called the safest credits in America: the hyperscalers.

The call comes as Big Tech's AI borrowing binge is rewriting the US corporate bond market.

According to Goldman Sachs Research, six companies — Amazon, Alphabet, Nvidia, Meta, Oracle and SpaceX — issued roughly $182 billion in bonds in the first half of 2026. A year earlier, that number was $13 billion. A 14x jump in 12 months from firms that were supposed to be self-funding the AI buildout from cash.

The problem, Goldman argues, is cash flow no longer covers it.

Hyperscaler capex is now running at close to 100% of operating cash flow in 2026 and is projected to exceed it in 2027. Goldman estimates the four largest hyperscalers will spend $725 billion in capex this year, and over $1.1 trillion in 2027. More than a third of that 2027 spend — $420 billion in bond issuance alone, up 60% from 2026 — will be funded with debt, a record.

It doesn't stop at bonds. Alphabet, Amazon, Meta, Microsoft and Oracle have also racked up $1.5 trillion in data center leases, with $1 trillion tied to leases that haven't even started yet and therefore don't show up as debt on balance sheets. Microsoft alone disclosed $329 billion in future leases through 2033, Oracle $260 billion. Goldman says this understates leverage.

The flood is hitting prices. AI-related bond spreads are hovering around 115 basis points over Treasuries, versus 78 bps for the broader investment-grade index, according to Goldman and ICE BofA data cited by Reuters this week. Investors are demanding extra yield just to absorb the supply.

That's why GSAM is pulling back.

Head of credit and portfolio managers at the asset management arm say they are being "very selective" on hyperscaler debt. "Our degree of investment conviction needs to be very high because of the coming supply and because of the lack of visibility into that return on invested capital," Brown Advisory's fixed income head told Reuters, echoing a sentiment now widespread.

Two numbers explain the caution:

1. Concentration: Five hyperscalers now make up ∼4% of the entire US IG bond index. Tech is 20% of all US IG issuance — one in five dollars lent.

2. Off-balance sheet leverage: Even if Microsoft, Meta, Alphabet and Amazon levered to 2x net debt, Goldman says they have $1.3-1.4 trillion of debt capacity and remain AAA/AA rated. But Oracle already sits at BBB- after $130 billion in debt, and Alphabet just posted its first negative free cash flow quarter since 2004.

Goldman Sachs credit trading desk internally flagged a "carnage" scenario if Fed policy tightens under new Chair Kevin Warsh — higher financing costs would hit the most capital-intensive growth engine in the S&P 500, which now accounts for 42% of market cap and 50% of expected 2026 earnings growth.

For now, Goldman still says hyperscalers have "tremendous debt capacity." But GSAM's message to clients is clear: the AI trade in equities may be booming, but in credit, you're being paid to be cautious — and underweight.

#US30YearYieldHighestSince2004 #BinanceWillListHyperliquid(HYPE) #FedOctoberRateHikeOddsRiseTo69.7% #BitcoinFallsToAround$84,600ThisWeek #USWeighsPromotingDollarStablecoinsAbroad