Goldman Sachs: The AI Capex Retreat Is Equivalent to an Alternative Form of Quantitative Easing for the Bond Market

On Thursday, as earnings reports from several tech giants raised market doubts about the sustainability of AI capital expenditure, US Treasury yields fell sharply across the curve, with yields on 2-year to 10-year Treasuries all dropping more than 15 basis points in a single day, the largest one-day decline in more than a year, giving Treasuries their best single-day performance since August 2025.

The 10-year Treasury yield fell below 5% for the first time in nearly a month.

In his morning note to institutional clients, Goldman Sachs' Rich Privorotsky characterized the phenomenon as "remarkable reflexivity"—every pullback in AI capital expenditure means a reduction in potential bond supply, thereby directly relieving pressure on the bond market.

In Privorotsky's view, when the scale of debt issuance by hyperscale cloud computing companies has become comparable to that of US Treasuries, any contraction in AI capital expenditure is essentially equivalent to reducing potential supply pressure on the bond market, with an effect akin to quantitative easing.

Earnings Trigger Selloff, Multiple Signals Shake Market Confidence

The trigger for this round of tech stock declines came from a series of mixed earnings signals. According to a breakdown by Goldman Sachs' Privorotsky, the market came under pressure from at least five factors.

First, although Lam Research (LRCX) delivered strong results, it failed to break through the market's implicit upper limit of expectations, and its guidance did not offer significant upside; shares of Japan's Tokyo Electron subsequently fell more than 6.99%;

Second, Tesla cut its full-year capital expenditure guidance from more than $25 billion to about $20 billion, a clear divergence from the positive signals previously sent by SpaceX, and the market was particularly sensitive to the reaction to the capital expenditure reversal;

Third, Meta's overall earnings performance was decent, and it raised the lower bound of its full-year capital expenditure guidance to $135 billion, but this means fourth-quarter spending alone would need to reach $49 billion to $59 billion, and the scale raised market doubts about execution feasibility;

Fourth, Texas Instruments delivered excellent results, with a full recovery in analog chips and strong data center demand, but the stock reaction was muted;

Fifth, DeepSeek's open-source reasoning model DeepInfer claims it can substantially improve inference efficiency on existing computing power, thereby unlocking existing capacity and prompting the market to reassess the necessity of incremental computing demand.

Privorotsky's summary was:

None of the above is disruptive on its own, but taken together, they are enough to shake already fragile market sentiment.

He noted that before Wednesday's modest rebound, the semiconductor sector had closed lower in 10 of the previous 11 trading sessions, and that rebound "already looked weak."

Goldman Sachs Prime data show that the share of net exposure to AI-related assets in total global net exposure has fallen from a peak of about 31% this summer to about 25%, and the long-short ratio has compressed from more than 4.2 to about 3.2.

Dragged down by the above news, the Philadelphia Semiconductor Index fell more than 7% on Thursday. Intel and AMD fell more than 8%, and Qualcomm fell more than 3%.

Privorotsky's analysis touches on a deeper market structural logic: the scale of AI capital expenditure has already directly affected the supply-demand balance in the bond market.

Meta's case is instructive. In the third quarter alone, Meta added $76 billion in off-balance-sheet commitments (mainly lease liabilities), bringing total off-balance-sheet commitments to $747 billion.

At the same time, Meta has joined Amazon in the camp of negative free cash flow. Meta's chief financial officer warned on the earnings call that "massive investment" will push up the cost of services, and as soon as he said this, Meta's stock price gave back most of its after-hours gains.

In Privorotsky's view, when the scale of debt issuance by hyperscale cloud computing companies has become comparable to that of US Treasuries, any contraction in AI capital expenditure is essentially equivalent to reducing potential supply pressure on the bond market, with an effect akin to quantitative easing.

He thus drew a conclusion unsettling to stock bulls: if the interest rate problem can only be alleviated when AI stocks fall, the bond market is in fact objectively "benefiting from" the continued slump in the AI trade, and the Federal Reserve, busy with rate hikes, will not step in to rescue the market on its own initiative.

Multiple Tailwinds Converge for Treasuries, Option Costs Soar

The spread between the French 10-year government bond yield and the German yield on the same maturity once widened to more than 170 basis points, the highest level since the 2011 European debt crisis, and combined with the wave of protests over the cost of living in France, this drove funds to seek safe haven in US and German government bonds.

In addition, market expectations for a Federal Reserve rate hike next year have also moderated somewhat, with the market now fully pricing in a 25 basis point hike in January next year rather than the previously expected December.

Against this backdrop, the cost of hedging with US Treasury options soared to the highest level since March on Thursday, and after the 10-year Treasury yield fell below 5%, traders spent about $7 million in option premiums to establish positions betting that the 10-year yield would fall to around 4.6% within a month, with a contract size of about 65,000 lots.

Meanwhile, CME data show large-scale new short positions in long-bond futures, and the relevant shorts faced passive covering pressure as Treasuries rose, further fueling the rally.

Although Brent crude rose back above $112 per barrel after the attack on Saudi Arabia, in the view of Michael Chang, a rates derivatives strategist at Citi:

Treasuries are showing genuine safe-haven asset characteristics, which is an encouraging signal for the rates market. Risk Disclosure and Disclaimer

Markets carry risk, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Investment based on this is at the user's own risk.