After major political realignment in the US, Franklin Templeton’s Fixed Income CIO Sonal Desai expects wider deficits, resilient growth, further Fed hikes, and continued upward pressure on bond yields.
“Let’s start with AI. In my view, the AI infrastructure buildout will be slower after next year and more expensive. Data centers are less likely to benefit from fiscal incentives and more likely to have to contribute to build new power generation, or pay subsidies to help defray the higher cost of power due to their additional demand, and maybe commit to boosting local employment. These challenges are not insurmountable, but they could increase the cost of the build out.
“The adverse fallout on economic activity, however, will likely be moderate. The often-repeated argument that US growth is now almost entirely dependent on AI investment is vastly overstated,” says Desai.
The contribution of AI capital expenditures (capex) to gross domestic product (GDP) growth peaked in the third quarter of this year, when it accounted for 1.4 percentage points out of an annual growth rate of 2.1%. What we and other analysts can measure as AI capex, however, captures more than just AI investment. National accounts statistics provide data on information processing equipment, software, research and development, and data centers as the categories most closely associated with AI. This is the combined measure of capex that peaked at 1.4 percentage points last quarter. Between 2015 and 2022, before generative AI came onto the scene, this same measure averaged about 0.5 percentage points. The excess attributable to AI is therefore short of one percentage point.
Moreover, a lot of the related hardware is imported and therefore needs to be subtracted to get the net contribution to growth. This adjustment would at least halve the contribution of AI capex to growth, bringing it to a more realistic 0.4 to 0.5 percentage points, somewhat less than one-quarter of the US economy’s growth rate. Clearly this is a meaningful contribution, but far from the only driver of US growth.
According to Desai, “AI capex will not disappear, and these estimates show that even if it were cut in half, it would reduce US GDP growth by less than a quarter of a percentage point, which would still leave the US economy running at about potential. Since AI capex accounts for just over 7% of nominal GDP, the fallout would be modest, including in terms of employment (the same sectors account for about 5% of private payrolls).
“Finally, some of the slowing AI capex would plausibly be made up by an acceleration in other areas that have more recently been crowded out by the AI boom. The bottom line for me is that the shift in the political wind will make the AI infrastructure buildout slower and more expensive, but without major implications for growth and inflation.”
She adds “If concerns about the growth impact of AI slowdowns are overblown, I think the idea that a divided government will narrow the fiscal deficit is wishful thinking. If there is one thing that unites the two political sides, it is their desire to spend more. A divided house will not spend less. It will likely spend more.
“The growth outlook remains healthy, in my view. Looser fiscal policy and potentially lower energy prices, if the Iran crisis is resolved, would be positive for economic activity and would outweigh the limited headwind from a potentially slower pace of AI infrastructure investment.”
“As far as the Fed is concerned, a resilient economy, a strong labor market, and entrenched inflationary pressures will matter more than a potential marginal reduction in price pressures from energy and tariffs. I therefore think the Fed remains on track to deliver another two to three rate hikes, starting most likely in December.
“I also think pressure on bond yields will remain to the upside. As I have argued before, I think real yields are not rich yet. The neutral nominal fed funds rate is around 4%-4.50%, in my view, and adding about 1 percentage point of term and risk premium would give us a fair value of 5%-5.50% for 10-year US Treasuries.
“I expect continued loose fiscal policy will keep generating substantial funding requirements for the government. And as I mentioned above, I think any slack in funding demand created by a potential slowdown in AI infrastructure is likely to be taken up by investment in other sectors of the economy. Upward pressure on yields is therefore likely to persist,” says Desai.
“In credit, attractive yields do not necessarily translate into adequate compensation for risk. Despite recent moderate widening, quality-adjusted high-yield spreads sit near the 29th percentile of their range over the past three decades, compared with the 60th percentile over the past five years. Dispersion is significant, however, with CCC spreads near the 99th percentile of their five-year range, signaling material distress among weaker borrowers.
“While I do not see evidence of broad deterioration in corporate fundamentals, refinancing risks are increasingly important. I favour durable cash flows, manageable refinancing needs and sufficient compensation for downgrade and default risk.
“The opportunity is to earn income without underwriting a refinancing problem.”
