Plutux
Trump Wants the World's Lowest Interest Rates. The AI Boom Just Borrowed $167 Billion to Make Sure He Can't Have Them insight cover
Markets / EventJPM · NVDA · AMZN18 min read

Trump Wants the World's Lowest Interest Rates. The AI Boom Just Borrowed $167 Billion to Make Sure He Can't Have Them

Four hyperscalers spent $511 billion on capital expenditure in the year to June 30, 2026 — 75% more than a year earlier — and funded the gap with roughly $167 billion of net new debt. That is a private credit-demand shock landing on top of 3.4% headline inflation and a $1.17 trillion federal interest bill, and it is why futures price a September rate hike rather than the cuts the White House is demanding. The rate that actually governs the AI buildout is the 10-year at 4.77%, and political pressure on the Fed pushes that number up, not down.

Published Sep 6, 2026Updated Sep 6, 2026

Fed funds target range

3.50%–3.75%

Unchanged since December; next decision Sept 15–16

Odds of a September hike

~62%

Futures pricing after the Aug jobs report, Sept 4, 2026

Headline CPI

3.4%

12 months to July 2026; core at 2.5%

Hyperscaler capex

$510.7B

TTM through Jun 30, 2026, up 75% year over year

On Aug 31, President Trump said the United States "should have the lowest interest rates anywhere in the world." On Sept 4 he escalated, threatening to halt trade with deficit countries unless the Fed cuts. Markets responded by raising the odds of a September rate hike to about 62%.

Fed funds target range

3.50%–3.75%

Unchanged since December; next decision Sept 15–16

Odds of a September hike

~62%

Futures pricing after the Aug jobs report, Sept 4, 2026

Headline CPI

3.4%

12 months to July 2026; core at 2.5%

Hyperscaler capex

$510.7B

TTM through Jun 30, 2026, up 75% year over year

10-year Treasury yield

4.77%

Sept 3, 2026 — 115bp above the policy midpoint

Federal interest expense

$1.17T

Fiscal 2026 to date, through Jul 31, 2026

The event

The Demand Escalated From "Cut" to "Lowest in the World" — and the Market Moved the Other Way

This is not the familiar pressure campaign. Asking for a cut is a request about the next 25 basis points; asking for the lowest policy rate on earth is a request to run US rates below Japan's and Switzerland's while inflation sits at 3.4%. The gap between the demand and the pricing has never been wider.

  • Aug 31: Trump said the US "should have the lowest interest rates anywhere in the world," pairing it with a claim that US growth could reach 20%.
  • Sept 4: On Truth Social he wrote, "We should have the LOWEST RATE of any country in the World … LOWER THE RATE OR I'LL STOP TRADING WITH COUNTRIES WITH WHICH WE HAVE A DEFICIT."
  • Same morning, August payrolls came in at +162,000 — roughly three times consensus — with unemployment steady at 4.1% and participation up to 61.6%.
  • Futures odds of a 25bp hike on Sept 16 rose to about 62% from roughly 55% before the report, per Reuters.
  • Fed Chair Kevin Warsh told the Kansas City Fed's Jackson Hole conference that higher rates may be needed with inflation stuck above 2% for five and a half years.

We should have the LOWEST RATE of any country in the World … LOWER THE RATE OR I'LL STOP TRADING WITH COUNTRIES WITH WHICH WE HAVE A DEFICIT.

President Donald Trump, Truth Social, Sept 4, 2026, as reported by Reuters
The enforcement mechanism contradicts the goal. Halting trade with deficit countries is an import-supply shock, and import-supply shocks push the very CPI print that keeps the Fed from cutting. The threat, if executed, makes the demand less achievable.

The inflation picture

Core Inflation Is Almost Fine. Everything Around It Isn't

The honest read of the July CPI report is split. Core is 2.5% — close enough to target that a rate-cut case exists on paper. But headline runs a full point higher at 3.4%, driven by a 14.7% year-over-year energy surge, and the Fed's preferred PCE measure was expected at above 3% heading into Jackson Hole.

US consumer price index, 12 months ended July 2026 (released Aug 12, 2026)
ComponentYear-over-yearWhat it tells the Fed
All items3.4%Above target for five and a half consecutive years
Core (ex food and energy)2.5%The strongest argument the White House has
Energy+14.7%Oil-led; renewed strikes on Iran added to the move
Electricity+4.2%Grid cost pass-through, partly data-center driven
Food+3.0%Tariff and input-cost pass-through still live

A central bank can look through a relative price shock once. It cannot look through one that arrives alongside a tariff regime, an oil shock, an electricity cost cycle and 162,000 monthly job gains. Each channel on its own is survivable; stacked, they remove the cover for a cut.

The constraint nobody leads with

The AI Buildout Is a $511 Billion Claim on National Savings

Here is the part of the macro story that Fed commentary keeps treating as a separate subject. In the twelve months to June 30, 2026, four companies alone spent $510.7 billion on property and equipment — up 75% from $291.3 billion the year before. That is a real investment boom competing for the same pool of capital as the Treasury.

Capital expenditure, trailing twelve months to June 30, 2026

Microsoft figures are fiscal 2026 (year ended June 30, 2026); the others are the four quarters to June 30, 2026.

Unit: US$ billions

Amazon

+61% year over year

173

Alphabet

+98% year over year

132.4

Microsoft

+80% year over year

115.9

Meta Platforms

+71% year over year

89.3

Cash flow no longer covers it. Combined operating cash flow across the four was $660.3 billion, so capex now absorbs 77% of every operating dollar they generate, leaving $149.6 billion of free cash flow against a $510.7 billion spend. Amazon ran negative free cash flow of $11.6 billion over the period.

How the buildout is financed — trailing twelve months to June 30, 2026 (US$ billions)
CompanyCapexOperating cash flowFree cash flowNet debt issued
Amazon$173.0$161.4-$11.6$75.7
Alphabet$132.4$185.7$53.3$70.1
Microsoft$115.9$182.9$67.0-$3.0
Meta Platforms$89.3$130.3$41.0$23.7
Combined$510.7$660.3$149.6$166.6
The four largest AI spenders raised roughly $167 billion of net new debt in a single trailing year. Reuters put total hyperscaler AI-related issuance at $220 billion in 2026 alone. An investment boom that borrows raises the neutral rate — it does not lower it.

That is the mechanism the "lowest rates in the world" demand runs into. Rates equilibrate desired saving against desired investment. When the private sector's investment demand jumps by $219 billion a year and the government is already absorbing savings to fund a $1.17 trillion fiscal-year interest bill, the market-clearing rate rises regardless of who chairs the Fed.

Transmission

The Fed Doesn't Set the Rate That Prices the AI Buildout

Even a compliant Fed could not deliver what the White House wants, because the buildout is financed at the long end and in corporate credit — not at the overnight rate. The 10-year sits at 4.77%, 115 basis points above the policy midpoint, and investment-grade corporates yield about 5.4%.

The policy rate is the lowest rate in the stack — and the least relevant to a data center

Fed funds midpoint and Treasury yields as of Sept 2–3, 2026; average rate on federal debt through Jul 31, 2026; investment-grade corporate index yield per Reuters, Aug 21, 2026.

Unit: %

Avg rate on federal debt

Fiscal 2026 to date

3.5

Fed funds (midpoint)

The only rate the FOMC sets directly

3.6

2-year Treasury

Sept 2, 2026

4.3

10-year Treasury

Sept 3, 2026

4.8

Investment-grade corporates

Where hyperscaler bonds clear

5.4

Long yields are policy expectations plus a term premium, and the term premium is where credibility gets priced. Estimates put the 10-year term premium near 0.72%–0.97% at the start of September, and Reuters tied the latest leg higher partly to investors demanding more to hold long US debt. Pressuring the Fed widens that premium rather than compressing it.

  • On the ~$167 billion the four hyperscalers issued in the last year, every 100bp of extra funding cost is roughly $1.7 billion of annual interest — a permanent tax on the buildout.
  • On the $220 billion of 2026 AI-related issuance Reuters tracks, 100bp adds about $2.2 billion a year in coupon before a single additional dollar is borrowed.
  • Federal interest expense already ran to $1.17 trillion in fiscal 2026 through July, at an average rate of just 3.447% — well below where new paper is being issued today.
  • That average rate is the real fiscal risk: refinancing $40 trillion of debt into a 4.77% ten-year is a slow-motion repricing that a lower funds rate barely touches.
The uncomfortable conclusion for the White House: the campaign to force cuts is most likely to steepen the curve and raise the cost of the AI capex it celebrates, because the AI buildout borrows at the long end where political risk is priced.

Second-order effects

AI Capex Has Become an Inflation Input, Not Just a Growth Input

The buildout does not only compete for capital — it competes for electrons, and that shows up in the CPI. Electricity prices rose 4.2% in the year to July, and grid capacity costs are being repriced upward across the largest US power market.

  • PJM's capacity auction for the 2028–29 delivery year cleared at the $325/MW-day price cap across the entire footprint in July 2026, with a reserve shortfall.
  • The prior cycle cleared at $329.17/MW-day for 2026–27, up from $28.92/MW-day for 2024–25 — a more-than-tenfold repricing in two years.
  • Data-center demand accounted for roughly $6.3 billion of those capacity costs across 13 states, per reporting on the July 2026 auction.
  • Those costs land in household bills, then in CPI — tightening the constraint on the Fed that AI capex is simultaneously loosening on growth.

That loop creates identifiable winners in the physical layer. GE Vernova posted $41.4 billion of trailing revenue with quarterly revenue up 21.9% and earnings up 32.8%; Vertiv grew quarterly revenue 24.1% with earnings up 53%; Constellation Energy grew quarterly revenue 23.0%. Scarce power and scarce electrical equipment are repricing faster than the compute itself.

The physical supply chain of the buildout — company-reported figures for the quarter ended June 30, 2026
CompanyRole in the chainQuarterly revenue growthTrailing P/E
NVIDIAAccelerated compute+105.9%29.2x
VertivPower and thermal management+24.1%61.0x
GE VernovaTurbines, grid equipment+21.9%27.0x
Constellation EnergyNuclear and merchant generation+23.0%29.3x

Positioning

If the Fed Wins the Argument, Financials Get Paid and Housing Pays

Higher-for-longer is not a neutral outcome — it is a transfer. JPMorgan Chase reported $25.5 billion of net interest income in the June 2026 quarter, up 9.9% from $23.2 billion a year earlier, and net income of $21.2 billion, up 41% year over year, on diluted EPS of $7.57.

The mirror image sits in housing. D.R. Horton reported flat quarterly revenue and earnings down 4.8% year over year, and the shares at $142.75 trade about 23% below their 52-week high of $184.55 and below their 200-day average. A 4.77% ten-year is already doing the damage a hike would formalize.

Rate-sensitivity scorecard under a higher-for-longer path
ExposureMechanismDirection
JPMorgan ChaseAsset repricing lifts net interest income; heavy AI-related bond underwritingPositive
NVIDIADemand intact at 63.7% net margin, but customers now fund purchases with debtPositive with a new financing dependency
Amazon, Alphabet$145.8B combined net debt issued in a year; long-end costs hit the buildoutMixed
GE Vernova, Constellation EnergyPower scarcity repricing; capacity auctions at the capPositive
VertivOrders boom, but a 61x multiple and 2.08 beta are discount-rate sensitiveMixed
D.R. HortonMortgage rates track the 10-year, not the funds rateNegative

The next three weeks decide it

Week of Sept 7

August CPI and PPI

Reuters calls it the swing input for the meeting

Sept 15–16

FOMC decision

~62% priced for a 25bp hike

Current range

3.50%–3.75%

Held since December 2025

Street path

4.00%–4.25% by year-end

Two 25bp hikes, per Nationwide's view cited by Reuters

Fed leadership

Chair Kevin Warsh

Signalled at Jackson Hole that higher rates may be needed

Both horizons

What Moves First, What Compounds, and What Would Prove This Wrong

  • Days to weeks: the August CPI print is the single variable. A core reading that holds at or below 2.5% reopens the case for a hold and flattens the curve; a hot headline confirms the hike.
  • Weeks to a quarter: if trade threats are executed against deficit countries, expect an import-price impulse with a two-to-three-month lag — the opposite of what the demand intends.
  • One to three years: hyperscaler debt issuance is the metric to watch. Reuters already flags investor fatigue at $220 billion of 2026 AI issuance; the next $220–300 billion clears at whatever term premium politics leaves behind.
  • Structural: if AI capex keeps compounding at anything near 75%, the neutral rate stays elevated and the 2010s low-rate regime does not return regardless of who sits on the Board.

The strongest counterargument is in the data above, not outside it. Core CPI at 2.5% is genuinely close to target, and the 14.7% energy spike is partly a geopolitical oil shock tied to renewed strikes on Iran — the classic definition of something a central bank should look through. If oil retraces, the hawkish case thins to one payroll print.

  • Risk to the thesis #1: oil normalizes, headline converges to core near 2.5%, and the Fed holds — bond-proxy losers like D.R. Horton rerate first.
  • Risk to the thesis #2: AI capex digestion. Any hyperscaler guiding capex down would cut the private credit-demand shock at its source and pull long yields lower.
  • Risk to the thesis #3: personnel. Board turnover that delivers cuts anyway would likely widen the term premium rather than lower long rates — a steepener, not a bull market in duration.
  • Risk to the thesis #4: labor cracks. Hourly earnings growth of 3.1% is not an inflationary wage spiral; a weak September payroll changes the entire framing.
Facts: the rate range, the CPI components, the capex and debt figures, the auction clearing prices. Inference: that AI capex is now a macro-level credit-demand shock. Speculation: that a trade halt gets executed. The first two carry the argument without needing the third.

Where This Lands in Portfolios

JJPMorgan ChaseJPM--
--Vol --
-
Bullish
  • Q2 2026 net interest income of $25.5B was up 9.9% year over year; a September hike extends the asset-repricing tailwind into 2027.
  • Underwrites the $220B of 2026 AI-related bond supply — fee income rises with the same issuance that keeps long rates elevated.
  • Near-term risk: an executed trade halt would hit credit costs and trading revenue before it hits net interest income.
NNVIDIANVDA--
--Vol --
-
Bullish
  • Trailing revenue of $303B at a 63.7% net margin, with the July-quarter top line up 105.9% — demand is not yet rate-constrained.
  • New dependency: buyers now fund purchases with debt, so a 100bp widening in credit spreads becomes a demand variable for the first time.
  • Over one to three years, order momentum survives higher rates; over quarters, any hyperscaler capex guide-down transmits here within one reporting cycle.
AAmazonAMZN--
--Vol --
-
Mixed
  • Spent $173.0B on capex in the year to June 30, 2026 while burning $11.6B of free cash flow — the most rate-exposed balance sheet of the four.
  • Raised $75.7B of net new debt in the same window; each 100bp of funding cost is roughly $757M of annual interest on that stack alone.
  • Cloud revenue growth must outrun financing cost — a 1–3 year test, not a quarterly one.
GAlphabetGOOGL--
--Vol --
-
Mixed
  • Capex of $132.4B was up 98% year over year, the steepest acceleration of the group, and pushed June-quarter free cash flow to negative $5.9B.
  • $70.1B of net debt issued over the trailing year converts a former net-cash story into a term-premium story.
  • Still generated $185.7B of operating cash flow — the cushion is real, but it is shrinking against the spend.
GGE VernovaGEV--
--Vol --
-
Bullish
  • PJM capacity clearing at the $325/MW-day cap with a reserve shortfall signals years of turbine and grid orders, not a single cycle.
  • June-quarter revenue rose 21.9% and earnings 32.8% on $41.4B of trailing revenue; net cash balance sheet limits rate sensitivity.
  • Watch the 27x trailing multiple: the earnings are cyclical even if the demand driver is structural.
CConstellation EnergyCEG--
--Vol --
-
Mixed
  • June-quarter revenue up 23.0% as data-center power demand adds roughly $6.3B to regional capacity costs across 13 states.
  • But quarterly earnings fell 46.8% year over year and capex absorbs 92.7% of operating cash flow — the growth is not yet converting to free cash.
  • Utilities carry duration: a 4.77% ten-year caps the multiple even as the power thesis strengthens over 1–3 years.
VVertiv HoldingsVRT--
--Vol --
-
Mixed
  • June-quarter revenue up 24.1% and earnings up 53% place it directly in the capex flow of the $510.7B buildout.
  • A 61x trailing multiple and 2.08 beta mean a 25bp hike hits the valuation faster than it hits the order book.
  • Binary catalyst: the Sept 16 decision, then Q3 orders — fundamentals and discount rate move in opposite directions.
DD.R. HortonDHI--
--Vol --
-
Bearish
  • Mortgage rates track the 10-year at 4.77%, not the funds rate, so even a political win on cuts fails to rescue affordability.
  • June-quarter revenue was flat with earnings down 4.8%; shares at $142.75 sit 23% below the 52-week high and under the 200-day average.
  • Short-term catalyst: a soft August CPI is the fastest path to relief; absent that, a September hike extends the drawdown.

Plutux is not an investment adviser. Market data and AI-generated analysis are for information and education only, not investment advice. Disclaimer

© Plutux Technology Limited 2026