Macro Check: The Mark AI Leaves in Official Statistics
This board answers: whether the AI pull shows up in official US macro data, how concentrated it is, and which series is the true footprint of AI capex.
Read from this board: The AI pull is highly concentrated: computer and electronics orders are up about 15% year on year while other durable goods are up only 6%; business equipment investment up more than 30% is the true macro footprint of AI capex, and factory orders only catch the slice not assembled overseas.
334 new orders, YoY
15.4%
2026-06｜computer and electronics, 3-month smoothed
Other durables, YoY
6.2%
2026-06｜excluding 334, 3-month smoothed
334 unfilled orders
$157B
2026-06｜stock, record high
Equipment investment
$798B
2026Q2｜BEA information processing, annualized
How to read this board
Core view: Before 2020 this set of series was read to call the electronics cycle; now the same series answer a different question: how concentrated the AI pull is in official macro data. Two readings. M3 is orders received by US factories, and most AI servers are assembled overseas, so 334 orders up about 15% understate the AI footprint; BEA equipment investment is what US businesses spend, imports included, so up more than 30% is the real footprint, and the gap between the two is the supply chain sitting abroad. Durable goods excluding 334 are up only 6%: the traditional economy is mild and AI is highly concentrated, the macro-data version of sorting into layers before drawing a conclusion.
- The warning combination is BEA equipment investment growth clearly rolling over together with 334 unfilled orders turning down; both at once marks the top of the investment cycle. One series weakening alone is usually a definition issue.
- In the past the two growth lines pulled apart mostly because the blue line collapsed in a recession (2009, 2020); this time the orange line is surging on its own, a different animal.
- Nominal orders at a new cycle high but still below the 2000 peak measure 26 years of offshoring and hardware deflation, not the whole of demand.
- M3 is published with about a five-week lag and BEA with GDP; both are seasonally adjusted. Read direction and the gap between layers, not single-month noise.
Term: 334 is the computer and electronic products industry in the Census Bureau's M3 factory orders, where AI server orders land; but most servers are assembled overseas, so it only captures the part US factories receive. BEA information-processing equipment investment is what US businesses actually spend on equipment, imports included, the direct footprint of AI capex in the GDP accounts.
1Concentration: computer and electronics versus other durables
Orange is 334, computer and electronic products, where AI-related orders land; blue is all durable goods excluding 334. New orders, year on year, three-month smoothed. In June 2026 the raw monthly figures were about 17% against 7%. In the past the two lines pulled apart mostly because the blue line collapsed in a recession (2009, 2020); this time the orange line is surging on its own: the AI pull is concentrated and the traditional economy is mild.
Concentration: Computers and Electronics vs the Rest of Durables (New Orders YoY, 3-Month Average)#
How to read this chart
Orange is NAICS 334, computers and electronics, where AI lands; blue is all other durables. Three-month smoothed. Past wide gaps mostly came from the blue line collapsing in recessions (2009, 2020); this time the orange line is surging on its own.
Source: US Census Bureau M3 factory orders via FRED; FinSight compilation · Updated 2026-10-02
2The US-factory view: nominal orders and full capacity
Left: nominal 334 new orders at $31.1B a month, a new cycle high but still below the $41.2B dot-com peak of June 2000 (dashed line), the measure of 26 years of offshoring plus hardware deflation; this is the US-factory view, not the whole of AI demand. Right: unfilled orders at a record $157.4B, nearly 30% above the 2019 average, orders received faster than they can be worked off; the semiconductor and electronic components production index is nearly 70% above 2019, so domestic capacity is running flat out too.
Nominal Orders vs 2000: NAICS 334 New Orders ($B/Month, Seasonally Adjusted)#
How to read this chart
This is the US-factory view, not the whole of AI demand.
Source: US Census Bureau M3 factory orders via FRED; FinSight compilation · Updated 2026-10-02
Signs Capacity Is Maxed Out: NAICS 334 Unfilled Orders and Semiconductor Industrial Production#
How to read this chart
Left axis: unfilled orders, $157.4B in June 2026, 28% above the 2019 average. Right axis: the semiconductor and electronic components production index, 69% above 2019.
Source: US Census Bureau M3 and Federal Reserve G.17 via FRED; FinSight compilation · Updated 2026-10-02
3The money view: equipment investment is the true footprint of AI capex
BEA information-processing equipment investment ran at an annualized $798B in the second quarter of 2026, up about 31% year on year; it includes imported equipment, so it is more sensitive than factory orders, and it is the direct footprint of AI capex in the GDP accounts. Software investment was $815B, up about 9%. The slope of the equipment curve since 2023 is the body of this capex cycle.
Follow the Dollars: BEA Information Processing Equipment vs Software Investment (Nominal SAAR, $B)#
How to read this chart
Equipment investment includes imports, so it is more sensitive than factory orders. Software investment is $815B, up 9.2%. The slope of the equipment curve since 2023 is this CapEx cycle itself.
Source: US Bureau of Economic Analysis via FRED; FinSight compilation · Updated 2026-10-02