Original Lens and publication context

A Buy Side Lens reading of Goldman Sachs's H2 macro roadmap: fund the railway before pricing the software.

  • AI is a capex cycle with a railway-scale physical footprint, not a software cycle — and infrastructure booms raise the cost of capital where software cycles lowered it.
  • The funding has left the balance sheet: hyperscaler credit issuance is running at roughly ten times the historical norm, with the equity channel now tapped alongside, competing with fiscal borrowing for a waning duration bid.
  • June payrolls (three-month average 111k, health services the only job creator, no wage pickup) deflated the July hike; any tightening is optional and capped at two to three moves.
  • Physical bottlenecks — power, grid, skilled construction labour, cooling — are potential headwinds to inflation normalisation if the stated capex numbers are realised.
  • The trade: the front end is adequately priced; fade rallies in the belly and long end of DM curves — price the build before the benefits.

I. Railway, Not Software

Goldman Sachs's macro desk opens its H2 roadmap by attacking the market's favourite frame. AI-as-productivity-story has supported the rates outlook all year: disinflation eventually, cheaper compute, better margins, possibly fewer jobs, a lower neutral rate down the road. The desk's view is that this is likely right about the destination — and dangerously silent about the journey.

What is happening is not a software cycle. It is a capex cycle, closer in physical footprint to a railway build than to the dotcom expansion everyone keeps citing. The distinction carries the entire rates argument: the software cycle lowered equilibrium rates because it scaled on almost no capital, while infrastructure booms raise them because they compete for scarce capital — power, grid connections, skilled construction labour, cooling, and above all funding. Tech investment as a share of GDP is now at a record, above the dotcom peak, and consensus capex estimates have only moved one way.

EXHIBIT 1 — THE BUILD KEEPS GETTING BIGGER. LEFT PANEL: TECHNOLOGY INVESTMENT AS A SHARE OF US GDP IS AT AN ALL-TIME HIGH — ABOVE EVEN THE DOTCOM PEAK. RIGHT PANEL: CONSENSUS HYPERSCALER CAPEX ESTIMATES FOR 2027 HAVE JUMPED FROM $614BN TO $942BN IN JUST EIGHT MONTHS. EVERY UPGRADE TO THE BUILD IS AN UPGRADE TO THE FUNDING IT NEEDS.
EXHIBIT 1 — THE BUILD KEEPS GETTING BIGGER. LEFT PANEL: TECHNOLOGY INVESTMENT AS A SHARE OF US GDP IS AT AN ALL-TIME HIGH — ABOVE EVEN THE DOTCOM PEAK. RIGHT PANEL: CONSENSUS HYPERSCALER CAPEX ESTIMATES FOR 2027 HAVE JUMPED FROM $614BN TO $942BN IN JUST EIGHT MONTHS. EVERY UPGRADE TO THE BUILD IS AN UPGRADE TO THE FUNDING IT NEEDS.

Supplied archive exhibit · original publication context.

We keep describing AI as a software cycle while building it like a railway. Rates markets need to price the build before they price the benefits.

II. The Funding Leak

The funding has now visibly left the private balance sheet. Credit issuance by the hyperscale complex is large and still expanding — on the desk's tracking, issuance across dollars, euros, and other currencies is running at an order of magnitude above the pre-2025 norm — and the equity channel is now being tapped alongside it. This is the mechanical heart of the note: every incremental bond sold to fund a data centre is duration supply competing with government paper for the same, visibly waning, duration bid. The fiscal supply story that has weighed on DM long ends all year has acquired a second, private-sector leg.

EXHIBIT 2 — HOW THE BUILD IS BEING PAID FOR. BOND ISSUANCE BY THE FIVE BIGGEST CLOUD COMPANIES IS NEAR $170BN ALREADY IN 2026 — ROUGHLY TEN TIMES A NORMAL YEAR — AND NOW SPANS DOLLARS, EUROS, STERLING, AND YEN. IN PLAIN WORDS: THE AI BUILD HAS OUTGROWN INTERNAL CASH FLOW AND MOVED ONTO THE BOND MARKET, WHERE IT COMPETES WITH GOVERNMENTS FOR THE SAME INVESTORS.
EXHIBIT 2 — HOW THE BUILD IS BEING PAID FOR. BOND ISSUANCE BY THE FIVE BIGGEST CLOUD COMPANIES IS NEAR $170BN ALREADY IN 2026 — ROUGHLY TEN TIMES A NORMAL YEAR — AND NOW SPANS DOLLARS, EUROS, STERLING, AND YEN. IN PLAIN WORDS: THE AI BUILD HAS OUTGROWN INTERNAL CASH FLOW AND MOVED ONTO THE BOND MARKET, WHERE IT COMPETES WITH GOVERNMENTS FOR THE SAME INVESTORS.

Supplied archive exhibit · original publication context.

Layer on the physical bottleneck risk if the stated capex numbers are realised — power, grid, skilled construction labour, cooling — and the desk sees potential headwinds to inflation normalisation itself. The burden of proof from here sits with the productivity gain: whether it emerges, and how long it takes. Until it does, the known deserves more weight than the promised: more competition for capital, private and public, credit and equity.

The desk is candid that the theme is hard to trade cleanly — AI may still deliver the disinflation and lower neutral rate the software framing implies. But an enormous physical build must be funded first, and the funding has to come from somewhere. The conclusion: this caps how far rallies in the belly and long end of DM curves can extend from here.

Figure 01 · Conceptual framework

Why an AI buildout reaches the yield curve

Why an AI buildout reaches the yield curvePhysical investment. Compute, buildings, power and cooling Financing demand. Cash flow, bonds and equity Competition for capital. Private issuance joins sovereign supply Required return. Funding costs influence investment choices Productivity payback. Benefits must justify the capital committedPhysical investmentCompute, buildings, power andcoolingFinancing demandCash flow, bonds and equityCompetition for capitalPrivate issuance joins sovereignsupplyRequired returnFunding costs influenceinvestment choicesProductivity paybackBenefits must justify thecapital committed
Reading the figureThe source’s cost-of-capital mechanism. It is a hypothesis about transmission, not a measured forecast of interest rates.

III. The July Hike Fizzle

Payrolls took the air out of the July hike balloon. The three-month average of nonfarm payrolls is down to 111k from 188k; the tick lower in unemployment was nullified by a falling participation rate; revisions plus a negative June print for hospitality leave health services as the only real job creator. The softness is corroborated everywhere the desk looks: soft vacancies, a declining labour-market differential, muted hiring intentions, and critically no wage pickup.

EXHIBIT 3 — WHO IS ACTUALLY HIRING. THE RED LINE IS HEALTHCARE — THE ONLY SECTOR CONSISTENTLY ADDING JOBS. THE BLUE AND ORANGE LINES ARE THE REST OF THE PRIVATE ECONOMY, HOVERING AROUND ZERO. IN PLAIN WORDS: STRIP OUT HOSPITALS AND CARE HOMES, AND AMERICA IS BARELY CREATING JOBS. THAT IS WHY A JULY RATE HIKE SUDDENLY LOOKS UNNECESSARY.
EXHIBIT 3 — WHO IS ACTUALLY HIRING. THE RED LINE IS HEALTHCARE — THE ONLY SECTOR CONSISTENTLY ADDING JOBS. THE BLUE AND ORANGE LINES ARE THE REST OF THE PRIVATE ECONOMY, HOVERING AROUND ZERO. IN PLAIN WORDS: STRIP OUT HOSPITALS AND CARE HOMES, AND AMERICA IS BARELY CREATING JOBS. THAT IS WHY A JULY RATE HIKE SUDDENLY LOOKS UNNECESSARY.

Supplied archive exhibit · original publication context.

EXHIBIT 4 — THE UNDERLYING TREND IN JOB GROWTH. THIS SMOOTHS OUT THE MONTHLY NOISE. TREND JOB CREATION COLLAPSED TO NEAR ZERO IN MID-2025, RECOVERED THROUGH 2026, BUT SITS AT A MODEST 75–120K A MONTH — AN ECONOMY THAT IS GROWING, NOT OVERHEATING. NOTHING HERE FORCES THE FED'S HAND.
EXHIBIT 4 — THE UNDERLYING TREND IN JOB GROWTH. THIS SMOOTHS OUT THE MONTHLY NOISE. TREND JOB CREATION COLLAPSED TO NEAR ZERO IN MID-2025, RECOVERED THROUGH 2026, BUT SITS AT A MODEST 75–120K A MONTH — AN ECONOMY THAT IS GROWING, NOT OVERHEATING. NOTHING HERE FORCES THE FED'S HAND.

Supplied archive exhibit · original publication context.

Sintra reinforced the shift. Chair Warsh gave away little — the Greenspan-era communication style is clearly deliberate — but characterised inflation risks as having come down, and his AI commentary pointed at the destination (supply-side effects that could lower neutral rates) rather than the cost-of-capital journey this note is built on. The combination compresses hike risk without eliminating it: the labour data is soft, not disastrous, and it buys the Fed time.

The desk's position: the Fed could hike, but it is optional. A hike would burnish inflation-fighting credibility at seemingly minimal cost. It is less clear it is necessary — energy is doing the work, sequential inflation should fall from here helped by PCE methodology changes — and there is a deeper question of whether hikes are even the right tool for a capex-driven cycle, when the rate-sensitive sectors hikes actually reach, housing above all, are already struggling. If the Fed goes, there is no case for a sustained cycle beyond two to three moves. Inflation prints decide everything.

EXHIBIT 5 — THE MOMENTUM BASKET BREAKS. THE 'MOMENTUM BASKET' BUNDLES THE MARKET'S HOTTEST STOCKS — THIS CYCLE, CHIPS AND MEMORY. IT DROPPED 17.3% IN TWO DAYS, THE SHARPEST BREAK SINCE THE 2020 PANDEMIC CRASH (DASHED ARROW). WHEN THE MOST CROWDED TRADE BREAKS THIS HARD, LEADERSHIP USUALLY CHANGES.
EXHIBIT 5 — THE MOMENTUM BASKET BREAKS. THE 'MOMENTUM BASKET' BUNDLES THE MARKET'S HOTTEST STOCKS — THIS CYCLE, CHIPS AND MEMORY. IT DROPPED 17.3% IN TWO DAYS, THE SHARPEST BREAK SINCE THE 2020 PANDEMIC CRASH (DASHED ARROW). WHEN THE MOST CROWDED TRADE BREAKS THIS HARD, LEADERSHIP USUALLY CHANGES.

Supplied archive exhibit · original publication context.

IV. What Is Priced

The strip carries roughly 38bp — about one and a half 25bp hikes — through March 2027, front-loaded into the September-to-January window, and nothing thereafter. Against the desk's own view (optional, shallow, data-dependent), the front end is adequately priced: not worth fighting in either direction. The trade therefore lives further out the curve, where the cost-of-capital argument does its work.

EXHIBIT 6 — WHAT THE BOND MARKET EXPECTS, MEETING BY MEETING. 'BPS PRICED' IS THE TIGHTENING BUILT INTO MARKET PRICES BY EACH FED MEETING DATE: ABOUT 38BP — ROUGHLY ONE AND A HALF 25BP HIKES — BY MARCH 2027, AND NOTHING AFTER. IN PLAIN WORDS: MARKETS EXPECT AT MOST ONE OR TWO HIKES, THEN DONE. THAT IS THE BAR THE DATA HAS TO BEAT.
EXHIBIT 6 — WHAT THE BOND MARKET EXPECTS, MEETING BY MEETING. 'BPS PRICED' IS THE TIGHTENING BUILT INTO MARKET PRICES BY EACH FED MEETING DATE: ABOUT 38BP — ROUGHLY ONE AND A HALF 25BP HIKES — BY MARCH 2027, AND NOTHING AFTER. IN PLAIN WORDS: MARKETS EXPECT AT MOST ONE OR TWO HIKES, THEN DONE. THAT IS THE BAR THE DATA HAS TO BEAT.

Supplied archive exhibit · original publication context.

EXHIBIT 7 — THE JULY HIKE BALLOON DEFLATES. PRICED 2026 FED TIGHTENING IN MOTION: EXPECTATIONS DROPPED SHARPLY ON THE WEAK JUNE PAYROLLS AND HAVE SETTLED NEAR 30BP (CIRCLED) — BARELY MORE THAN ONE HIKE. TWO WEEKS AGO THE DEBATE WAS JULY; NOW IT IS WHETHER THE FED MOVES AT ALL.
EXHIBIT 7 — THE JULY HIKE BALLOON DEFLATES. PRICED 2026 FED TIGHTENING IN MOTION: EXPECTATIONS DROPPED SHARPLY ON THE WEAK JUNE PAYROLLS AND HAVE SETTLED NEAR 30BP (CIRCLED) — BARELY MORE THAN ONE HIKE. TWO WEEKS AGO THE DEBATE WAS JULY; NOW IT IS WHETHER THE FED MOVES AT ALL.

Supplied archive exhibit · original publication context.

V. The Trade

SLEEVE

POSITION

HORIZON

DRIVER

Front end

Adequately priced at ~1.5 hikes through Q1 2027; hold, do not fight.

Data-dependent

Optional-hike regime

Belly & long end

Fade rallies; dual duration supply (fiscal + private capex funding) caps the upside.

H2 2026

Cost of capital

Curve

The implied structure: front end anchored or rallying, long end pinned by supply — steepening bias.

H2 2026

Supply vs policy

Watchlist

Inflation prints into September; hyperscaler issuance calendar; power/grid/labour bottleneck data.

Rolling

Decides the hike

This desk note accompanies the consolidated Buy Side Lens report "Built Like a Railway, Priced Like Software" (BSL-018, 07 July 2026), which carries the comparison against the equity view.