About $800 billion will be spent this year on data center build-out and AI infrastructure, and Mona Mahajan expects that to rise to over $1.1 trillion next year.
A maturing trade is usually one where the spending rolls over. Mahajan's reading of the latest quarter of capital expenditure guidance is the opposite: the numbers are not going down, they are not staying steady, they are going up.
"So all of that together leads us to the kind of the conclusion that the AI story is probably maturing, not yet rolling over, not yet showing signs of even decelerating."
Mahajan sets investment strategy and asset allocation at Edward Jones, which SIFMA's president introduced on air as a $2.6 trillion wealth manager serving households across North America, and she was recorded for the monthly Market Snapshot before the September inflation print and before the Federal Open Market Committee met.
The full interview is covered here so you can skip it. 44 minutes of audio, 21 minutes of reading.
Here are the 14 insights that matter.
👤 Guest: Mona Mahajan, Principal at Edward Jones, where she runs the investment strategy and asset allocation teams for a wealth manager serving households across North America
🎙️ Host: Kenneth E. Bentsen Jr., President and CEO of SIFMA
👥 Also on: Heidi Learner, Director of Research at SIFMA, who writes its monthly market metrics and trends report
📰 Published: 15 September 2026 on the SIFMA Podcast feed · recorded earlier
🟢 Spotify | 🟣 Apple Podcasts | 🔗 Episode page | ⏱️ 44 min | ✅ Time saved: 23 min
Key Takeaways
AI capital spending is still climbing, from about $800 billion this year to over $1.1 trillion next
Backlogs at the semiconductor makers, the hyperscalers and the cloud businesses are all still strong
The productivity winners have not moved yet, and that is where the next three to five years go
Industrials, healthcare and financial services are the sectors she names
Wage growth has fallen below inflation, which kills the wage-price spiral argument
Year-over-year wage gains went from 3.5% to 3.1%
The stock market cares about the speed of a rate rise more than the level
The 10-year is up 60 to 65 basis points this year against over 200 in 2022, when the S&P 500 fell about 20%
Emerging markets are now a concentration trade too: three semiconductor companies are about a third of the index
Investing at an all-time high beat investing on any other day over one, three and five years
Only the three-month return was worse, and there were over 700 all-time-high days between 1990 and 2025
AI construction is now a state-politics problem, not only a capital-allocation one
Roughly 15 state legislatures have introduced bills to impose moratoriums on data center construction
The AI build-out is already visible in GDP and still invisible in productivity
A Fed paper put its Q1 2026 contribution at nearly three-quarters of a percentage point, up from a quarter point in Q1 2024
Corporate debt issuance is up more than 26% year to date and credit spreads have not widened
1. A Live Fed, No Wage Spiral
Bentsen opened on the Federal Open Market Committee meeting scheduled for 15 and 16 September, and on Mahajan's own description of that meeting as live.
Mahajan said the Fed is weighing three things: the labor market, inflation, and inflation expectations. The first two are the dual mandate; the third is what decides whether the first two stay manageable.
On the labor market she saw nothing to worry a central banker. Hiring picked up in the August jobs report and came in well above expectations, the unemployment rate stayed steady at 4.1%, and wage gains have been steady, so no wage-driven inflation is coming through.
Inflation is where the difficulty sits. It has been above the Fed's 2% target for over five years, headline inflation has crept higher on volatile oil prices, and the one piece of good news is that core inflation has come in at a steadier level.
Inflation expectations are contained, on market-based indicators. Her point about why that matters is behavioral: if expectations become unanchored, consumers, households and corporations act accordingly.
Her call was a bias to stay on hold unless the inflation figure due that Friday came in outsized. She also noted that the Fed does not want to dampen demand or the labor market to a large extent.
Bentsen pressed on real wages, since wage growth is now below inflation. Mahajan agreed the position has flipped — for much of the last couple of years real wages were positive against headline CPI, and consumers are no longer seeing real wage growth, which does dampen demand to some extent.
What the Fed is guarding against is the loop, not the level. She walked through it: demand for labor gets so high that corporations pay higher wages, that flows into the cost of goods and the cost of the end product, inflation follows, and wages chase it again.
"Yes, they're below inflation, but the good news is they're not spiraling out of control." Year-over-year wage increases came in at 3.5% and are back at 3.1%.
Learner put the market's own number on the meeting. On Fed funds futures, roughly a 62% probability of a 25 basis point move that would take the target from 3.5 to 3.75% up to 3.75 to 4% — the same probability as at the end of July, and up substantially from about 35% in mid-August.
"So again if you are to bet, I would bet on a rate hike, at least according to what the market consensus points to." — Heidi Learner
2. The 4.5–5% New Normal
Bentsen asked how much of the acceleration in long-term rates over the previous two to three weeks was stronger growth and how much was fiscal or inflation risk.
Mahajan said the term premium has risen but is not extended. She glossed the term as the extra yield an investor is paid to compensate for the risks in the market, and said it is off its lows but not back at even 12-month highs.
Her answer was that four things are pushing in the same direction. Strong economic growth and especially strong earnings growth this year; worries about rising debt, deficits and the US fiscal position; a global yield story showing signs of life; and large-cap AI companies tapping the debt market and competing in investment grade corporate bond issuance.
She put the 10-year Treasury yield at about 4.80 on the day, and said the market briefly touched 5% in 2023 and bounced straight off it.
"But in our mind, a 4.5 to 5% range is very much the new normal." She contrasted it with the last 10 to 15 years, when the Fed funds rate was closer to zero and Treasury yields were closer to one and a half to two percent.
The consequence cuts both ways. Higher yields mean a higher cost of borrowing for consumers and corporations, and a 4.8% 10-year — or anything shorter along the curve — becomes more interesting for a saver, someone close to retirement, or anyone looking for income.
Learner agreed on the risk premium and put more of the weight on inflation. She said the rise in 10-year rates is not solely down to inflation risk, growth or employment, but to a widening of the risk premium, and that the fiscal picture has to be part of that.
"So I'm gonna vote for a blend of all of the above." — Heidi Learner
3. No Crowding Out Yet
Bentsen asked whether government and corporate borrowers are bumping into each other over the same pool of capital for the first time since the early-to-mid 1990s.
Mahajan sized the competition. Forecasts point to upwards of $500 billion of AI corporate issuance, against a Treasury market she put at maybe two to two and a half trillion dollars a year on average — a big chunk on a percentage basis.
The AI borrowers are not weak credits. She said the companies tapping the debt market in that space have strong free cash flow metrics and are strong credits in debt markets.
"And so I think generally that supply-demand picture paints or points to yields moving higher." More supply means higher yields are needed to attract the incremental demand.
The buyer bases are not the same, which limits the collision. Pension funds and foreign investors tend to buy sovereigns rather than corporates.
Learner said the effect has not shown up where it would show up first. Option-adjusted spreads have not risen meaningfully over the last few months.
She would separate how much of the rise in total borrowing costs is the move in Treasuries and how much is credit spreads themselves.
"Another thing I think I'd watch rather than credit spreads more broadly is looking at CDS." — the price of default protection, particularly for names whose debt load has been rising.
Bentsen raised the offset he would expect: a contraction in mortgage-backed issuance as mortgage rates rise. Learner said corporate debt issuance is about 26% higher calendar year to date than the same period in 2025, with nothing similar happening in the mortgage market, and that higher mortgage rates are scaling back both purchases and refinancings.
4. The Bond Market Tightens
Bentsen asked whether, with long rates elevated and the Fed on hold, the bond market is doing the tightening on the Fed's behalf — something he said Chairman Warsh has more than suggested.
Mahajan called it a live debate and agreed with the mechanism. "To some extent higher yields, as we talked about, do translate to a higher cost of borrowing for both consumers and corporations. And that is a de facto form of tightening of financial conditions."
The move has spread down the curve. She said the long end has moved higher and the short end is starting to catch up, as oil markets have been volatile and inflationary pressures have come back to the forefront.
She raised a second question the market is asking: whether the Fed needs a hike for credibility rather than for the economy. Relaying Kevin Warsh, she said he has stated multiple times that the Fed still has work to do, that inflation is the number one focus, and that 2% is still the goal.
"On the other hand, does he need to raise rates to just have some credibility behind those words?"
Her bottom line was that a hike does not break the story. Whether the Fed raises once or even twice this cycle, she does not think it derails the broader economic and earnings picture, and she noted the consumer has been quite resilient through most of the rate rise.
5. Pace Beats Level
Bentsen turned to equity valuations, said he had often thought the economy ran effectively at a 6% long bond in the past, and asked whether higher yields are a long-term negative for the stock market.
"One of the things that we've looked at in the past is it's not only the level of Treasuries that the stock market looks at, it's also the pace at which we get to that higher level."
"So if you're moving rapidly higher, that is worse for equity markets to absorb than if it's a more gradual pace."
She put this year against 2022 to show the difference. The 10-year is up about 60 to 65 basis points in 2026; in 2022 it rose over 200 basis points in a short period, with the Fed tightening, and the market had a severe reaction and fell around 20%, close to bear market territory.
This year the S&P 500 is still up 11 to 12% despite the 60 basis point move.
The number the market watches is 5% on the 10-year. Above it, she said, quantitative buyers tend to come in on the view that the yield is attractive — and it is also the level at which valuations start to re-rate lower if the yield stays above it for prolonged periods.
Her comfort comes from the earnings half of the equation. Markets are driven by earnings growth and valuation expansion; valuation expansion has not been much of the story this year, but earnings growth is anticipated to be above 30% for the S&P 500, a rate not seen since 2021 coming out of the COVID crisis.
Learner made the discounting point plainly: a higher level of rates applied to a company's future cash flows should, all things being equal, produce a lower valuation. She added that the Fed controls only the Fed funds rate, and it is the 10-year that matters for most corporate borrowers, for mortgage rates and as a benchmark for other borrowing.
What she would add to level and pace is uncertainty — the whipsaw reactions to changing rate-hike probabilities.
"So, to the extent that we do see a rate hike in 2026, it's not one and done." Looking out a year, she said the market is pricing a fairly steep probability of consecutive or subsequent hikes, and may have to play catch-up.
6. The Dot Plot's Future
Bentsen asked whether the dot plot is finished under the new Fed regime, and whether individual bank presidents and governors will give more of their own forecasts.
Mahajan said Warsh has been critical of the dot plot to some extent and has probably asked his committee to look at whether to keep it.
She placed it in a wider pullback from Fed communication. She noted that Chairman Powell instituted the press conference at every meeting, and said cutting that back could be a first step in reducing the transparency and communication element of the Fed.
What she wants kept is the quarterly Summary of Economic Projections — the forward look at GDP, inflation and unemployment that comes every quarter alongside the dots.
Her reason is that it is a comparison tool, not a forecast. It is a meaningful way to assess how markets and investors are seeing the world against how the Fed is seeing it.
"So my hope is that they keep elements of this forward communication and guidance in place and maybe tweak or remove other elements of it."
7. AI Capex Still Rising
Bentsen quoted her own written description of the AI trade as maturing, not breaking, and asked what a mature AI trade looks like and whether leadership migrates away from semiconductors and infrastructure.
Her starting point was how narrow the gains have been. This is the fourth year of an S&P 500 up double digits, and in the last three years at least the gains have come from a concentrated set of sectors and a concentrated set of names, with the Magnificent Seven the drivers.
The capex numbers from the latest quarter are the evidence that it is not breaking. "They're not going down, they're not staying steady, they're actually increasing."
About $800 billion of capital expenditure on data center build-out and AI infrastructure this year, rising to over $1.1 trillion next year.
Backlogs are strong across the chain — semiconductors, hyperscalers, the cloud businesses — and demand continues to outpace supply.
"So all of that together leads us to the kind of the conclusion that the AI story is probably maturing, not yet rolling over, not yet showing signs of even decelerating." The pace of growth may be coming in a bit, but there are no negative growth rates.
The gap she sees is in the stock market, not in the spending. "But what we have not yet seen is a real meaningful increase in the sectors that may gain from productivity." The first few years rewarded the building blocks: semiconductors, hyperscalers, data center players.
The productivity gainers she names are industrials, through manufacturing; healthcare; and financial services, her own field. Asked whether this is a post-AI world, her answer was no — she sees a long runway over the next three to five years and expects a broadening into those sectors.
Her current positions are industrials and communication services, which she described as a balance between a cyclical broadening theme and a sector that happens to hold Google and Meta.
"We don't want to give up on the AI trade wholesale, but we want to make sure we have exposure to the next generation as well."
8. The Data Center Backlash
Bentsen said there is a big debate in many states about the growth of data centers, and asked whether the market is pricing the possibility of a curtailment.
Mahajan said it is hard to put numbers around it now, and that midterm elections could make it more prominent in certain parts of the country.
She read the backlash as a symptom of maturity rather than a threat to it. As the build-out reaches its end stages, she said, that is when the questions about ramifications, regulations and boundaries start being asked.
"And keep in mind, we went through similar phases when we went through the industrial revolution or the internet revolution as well." The work is establishing guardrails, regulations and boundaries between the effect on households on the ground and the benefit of the broader technology.
Her timeline is this election cycle or the next six to 12 months, and she called it a healthy sign that it is part of the narrative now.
Learner supplied the count. "I think it's something like 15 state legislatures have introduced bills to impose moratoriums on data center construction."
The concern behind those bills, on her reading, is the consumer's pocketbook — whether the build-out raises energy costs that have already been increasing at a rapid rate.
She said a supply constraint is possible at a certain point, but that for now the demand for new data center construction is still there.
9. AI Shows Up in GDP First
Bentsen asked what the capex spending is doing to the broader economic data — GDP, productivity and capital spending.
Learner said it is unambiguous in the GDP data, and cited a Fed paper from July that examined this exact question.
Even adjusting for the high import content of the equipment, the net contribution from spending on software, data centers, power facilities and computer and peripheral equipment came to nearly three-quarters of a percentage point of GDP in Q1 2026. That is up from half a point in Q1 2025 and a quarter point in Q1 2024.
The productivity side of the same paper is where the evidence runs out. It found what she called micro-level experiments showing productivity gains from AI tools, but no translation into aggregate productivity gains for the US as a whole yet.
Her explanation for the gap is measurement and bottlenecks. It is easy to measure how much time AI saves a programmer, or how much more code he can generate in a given time frame; that does not mean proportional gains for the firm if there are still bottlenecks elsewhere in it.
"But there is some evidence that measured productivity gains really lag investment by several years." So the answer may not arrive for some time.
Where she thinks the effect shows up before productivity is employment. Professional and business services, a sector with high exposure to AI, is not back to the peak levels of one or two years ago.
"So professional and business services, for example, is 1.4% lower today than it was more than two years ago." She noted this is against an economy that is still growing and clearly not in recession.
10. EM's Concentration Risk
Bentsen asked whether investors buying emerging markets — now heavy in Korea and Taiwan — are getting diversification or another version of the AI trade.
"And look, it's not only Korea and Taiwan make up a huge part of EM, three companies alone make up about a third of the EM basket." Those three are semiconductor companies.
Her conclusion was that the US concentration problem has been reproduced in emerging market equities.
She still counts it as a diversifier. For investors who want an alternative to the US technology trade, she said the space is growing rapidly and becoming highly competitive with US players, US models and US large language models.
Valuations support it too. Relative discounts are relatively lower on a historical view, and she said the same earnings trend and earnings uptick apply over the next 12 to 18 months.
She set out the three lenses Edward Jones uses on equities. By market capitalization, US large caps for the continued technology and AI exposure and US mid caps for the catch-up and broadening. By region, the US, emerging markets — which she is monitoring for a possible near-term peak — and international value in developed markets outside the US, naming Europe and Japan. By sector, US industrials and communication services.
On the bond side she offered a rule of thumb for balanced investors. Mahajan said "one of the best predictors of long-term investment grade bond returns is where its starting yield is at," and that today's elevated starting yield points to a positive total return profile.
11. Buying at All-Time Highs
Bentsen described research she has done on data from 1990 to 2025 showing higher long-term equity returns when investing at all-time highs than on any other day, called the result counterintuitive, and asked her to explain the method.
The question that prompted the work is the one she gets from clients. The market is back at all-time highs despite the noise and the volatility — should they keep investing at these levels or wait for a pullback?
The finding, in her words, is that "if you invested at all-time highs versus any other day in the market, your one, three, and five-year returns were actually higher investing at an all-time high than investing any other day in the market."
The exception is short-horizon money. Three-month returns from investing at a high are lower.
"But otherwise, investing at all-time highs is not a bad thing." She added that recent years have had a strong momentum effect, so an investor buying at a high is probably also buying momentum.
The practical instruction is to leave a dollar-cost-averaging plan alone. If an investor is putting money in every month, quarter or year, there is no need to alter that because the market is at a high — the history and the data do not support it.
Mahajan said investors "are notoriously not great at timing market bottoms and market tops, and nor do they need to." She noted other think tanks have done similar research.
Bentsen asked the obvious counter-question: what is the longest stretch since 1990 with no new high? She named two — between 2000 and 2007, after the dot-com crash, and a second she put at 2008 to 2013 after the financial crisis.
Across the whole 35-year period she counted over 700 days that were all-time highs, so every new high along the way counts.
The caveat is the deep, prolonged, recessionary bear market. "And on average, it does take two to three years to recover many of those losses."
Her reason for not expecting one is the trigger list. Those episodes tend to happen in or entering a recession, when the Fed is raising rates aggressively, or on an unknown like the pandemic. She sees no recession on the horizon, hopes the Fed raises once or twice at most, and said the unknown unknowns are harder to handicap.
12. The Risks She Watches
Bentsen asked what the market is underpricing and what would make her less constructive on equities.
Inflation is first. Headline inflation has been above 3%, on elevated energy and food prices, and it has not yet seeped into core.
Core inflation is the number that would change her view, because it is the part the Fed can reach. The Fed cannot control oil supply, but raising rates can dampen demand and reach the core; a meaningful acceleration in core inflation would be a big risk.
"If inflation expectations start to run away and become unanchored, we think that would be a cause for the Fed to really step in and make some more dramatic moves in its rate policy."
"Third one I'll point to is the geopolitics." She said it is very hard to handicap and that the uncertainty continues to be an overhang on markets.
13. Midterms and Gridlock
Bentsen closed by asking what investors should watch this month, with midterm elections less than two months away.
The Fed meeting on 16 September comes first, and then attention shifts to the midterms.
The pattern she cited goes back to the 1930s and is consistent: the period before an election is volatile, and the last couple of months after it tend to rally. Election day this year is 3 November.
Part of the post-election rally is uncertainty lifting; part of it is the seat math. The incumbent party tends to lose seats in the House and Senate regardless of which party it is, and that creates gridlock in Congress.
"And so why do markets actually favor that gridlock? Well, because it means less risk of or less chance of new regulation, new legislation, a little bit more operating transparency."
Her expectation is volatility into the midterms after a strong first half, then a better year end. She hedged it: "Now, history doesn't always repeat itself."
14. Hyperscaler Debt Is Covered
Bentsen asked Learner what she had covered in the latest SIFMA Insights market metrics and trends report.
The month's subject was corporate issuance, which is up more than 26% year to date — notable, she said, against rising interest rates and the talk about growing hyperscaler issuance.
She looked at debt ratios and found the information technology subsector of the S&P 500 stronger than the index as a whole. Even with accelerating capital expenditure, free cash flow per share and returns on capital were higher for information technology than for the S&P 500.
Net debt to EBITDA in the sector is rising, and the sector's median free cash flow coverage ratio is nearly 15 times.
"So we think there's substantial capacity to meet obligations based on internally generated cash flow, and we're comfortable with the abilities of these firms to service their current debt loads."
"So at least for now the high level of hyperscaler debt issuance doesn't raise too many concerns for us."
Bonus Insights
Bentsen's own framing of the crowding-out question was historical — he reached back to the early-to-mid 1990s for the last time he remembers government and corporate borrowers going after the same pool of capital.
He also pushed back on his own premise about rates and equities, noting that the economy ran effectively at a 6% long bond in the past, and that it has been a long time since it was there.
He made the point that fixed income investors are not monolithic — there are reasons rates investors and credit investors behave differently, so the pool of capital they draw on is not one pool.
His read on where the AI trade sits is that the market is still pricing the capital investment side of the trade, not the finished product output, which is where the productivity gains would reach a broader group of sectors.
Mahajan called the questions about a mature AI trade "all million dollar questions" before answering them.
The whole conversation was recorded before the inflation reading she kept pointing to. She described the figure due that Friday, 11 September, as the final data point that might sway the Fed one way or the other.
Mahajan's bottom line is that the AI trade is maturing rather than breaking — spending is still rising, the market has paid only the infrastructure half of it, and the next three to five years belong to the sectors that use the technology rather than build it, with a 4.5 to 5% 10-year Treasury as the backdrop rather than a threat.
Products, Companies & Tools Mentioned
Edward Jones (Mahajan's firm, introduced by the host as a $2.6 trillion wealth manager serving households across North America; it likes US large and mid caps, emerging markets, international value, industrials and communication services)
Google and Meta (Named as the reason communication services carries AI exposure alongside her cyclical industrials position)
Federal Reserve (Weighing a 25bp hike, reconsidering the dot plot, and the author of the July paper on what the AI build-out is contributing to GDP)
SIFMA (The host's organization; its research desk produces the monthly market metrics and trends report the episode closes on)
Books & Resources Mentioned
The AI Buildout and the Economy – Federal Reserve Board (The July 2026 FEDS Note that is Learner's source for the nearly three-quarters of a percentage point contribution to Q1 2026 GDP and for the gap between micro-level productivity gains and aggregate ones)
SIFMA Insights Market Metrics and Trends (Learner's monthly report; this edition covers corporate issuance and the free cash flow coverage of the S&P 500's information technology subsector)
Edward Jones' research on investing at all-time highs, 1990 to 2025 (The study behind the finding that one, three and five-year returns were higher from investing at a high than on any other day)
Watch the full episode:
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: