From, Daleep Singh, PGIM Fixed Income, Vice Chair and Chief Global Economist
MACRO
Last Monday’s announcement of a deal between OpenAI and Advanced Micro Devices provides another reminder that the boom in artificial intelligence capex appears to be on a parabolic trajectory. At this point, the leading tech companies are on pace to invest nearly $400B this year on the infrastructure needed to build and train AI models. With an estimated cumulative total of up to $7T globally by 2030, the expected capex is set to be among the largest on an industrial scale in history. The size of the investments also provide context to our recently published economic scenarios, which places a greater weight on the right tail of the distribution relative to the left side.
With U.S. real GDP growth of about 1.5% in Q2 2025, data center-related capex accounted for about 35% of that growth, and that share may climb to more than 50% through the second half. When factoring in growth multipliers and spillovers, AI capex investments have the potential to rival the productivity boom from the late 1990s.
Indeed, anecdotal studies suggest a median productivity uplift of about 30% in instances where generative AI has been deployed, and when applied to the current productivity baseline of 1.8%, that could amount to a productivity increase of 50 bps. The Congressional Budget Office indicated that a sustained uplift of that magnitude could reduce its projected debt-to-GDP from 156% over its extended baseline to 113% over the same timeframe.
The surge in AI-related capex eventually points to the ongoing sustainable-boom or looming-bust debate, and we see several points on each side. In terms of the boom, the demand is being driven by highly profitable firms with low leverage and dominant market positions. Furthermore, initial signs indicate the potential for rapidly accelerating adoption on a global scale considering a low marginal cost of deployment amidst the fixed assets already in place. Finally, the geopolitical race for AI supremacy will only intensify over time with the U.S. and China likely vying for the lead.
On the flip side, at this point, AI-related revenues are still only a fraction of the capex thus far, questions linger about stranded assets given the investments, and potential spillover effects to constituents include electricity prices and potential labor force displacement.
Also read: Pengana Global Private Credit Trust Offers Additional Units
DEVELOPED MARKET RATES
Developed market yields generally declined by less than 10 bps last week, including those in the U.S. as they usually do during partial shutdowns of the Federal government. Online markets are indicating a potential shutdown lasting 11-35 days as the most likely outcome in terms of a timeframe.
Despite the relative quiet on the trading and data front, there were some interesting anecdotal observations from last week. A significantly large call buying program was executed last week, implying an expectation for the U.S. 10-year yield to potentially drop below 4%. The buyer reportedly spent $200M in premium to set the position.
In the MBS market, mortgage volatility continues to decrease as spreads tightened modestly. Origination has slowed after a recent rate decline, helping firm up the market.
