Where can I get monthly semiconductor production data?

A monthly production reading is what a national statistical authority says a country's electronics factories made this month - expressed as an index level against a fixed base year, a percent change against last month and the same month last year, and an operating rate for the installed capacity. It is the supply-side scoreboard for the chip cycle: demand-side series like billings tell you revenue, production indexes tell you whether factories are actually running harder.

• Federal Reserve G.17 Industrial Production - Electronics - US NAICS 334 computer and electronic products, cut into computers (3341), communications equipment (3342) and semiconductors and related electronic components (3344), each carrying a seasonally adjusted index on a 2017=100 base, two horizons of percent change, capacity utilization and capacity growth, with history reaching 1972.
• UK ONS Manufacturing and Production Industry - the monthly Index of Production, 1,134 series columns over 1,340 monthly date rows, of which 84 match computer, electronic, optical or electrical equipment keywords, alongside annual ProdCom manufacturers' sales and the TOPSI turnover series.
• JEITA Electronics Shipment & Production Statistics - monthly Japanese domestic unit shipments by product category since 2000, volume rows interleaved with year-over-year percent-change rows, roughly 15-20 categories per yearly table.

Of the sixteen primary datasets cataloged for electronic equipment & instruments, these three are the only ones that measure production on a monthly rhythm. Everything else in the industry pool orbits them.

One row per industry line per month, exactly as they land in your warehouse - July 2026 observations verified during cataloging:

reference_month=2026-07  revision_status=p   (index base 2017=100)

series                                          naics  weight  ip_index  yoy_pct
Computer and electronic products                334    4.98    140.6      +9.9
Computers and peripheral equipment              3341   0.21    183.6      +7.3
Communications equipment                        3342   0.45    231.8     +13.7
Semiconductors & related electronic components  3344   1.49    191.9     +11.9

capacity_utilization                            naics  rate
Computer and electronic products                334    77.2
Semiconductors & related electronic components  3344   75.4

Read the weight column before anything else: semiconductors alone carry 1.49 percent of the entire US industrial production index, seven times the weight of computer assembly at 0.21 - which is why a chip downturn moves national manufacturing headlines while a PC downturn barely dents them.

Then read the tension between growth and operating rate. Semiconductor output rose 11.9 percent year over year, yet the line runs at 75.4 percent utilization, below the group average of 77.2. Capacity is being added almost as fast as volume climbs, so headline growth overstates tightness. That pairing - index beside utilization - is the whole point of pulling both measures instead of one.

A second national frame measured the same way every month, plus turnover in money rather than only volume. The ONS shelf holds three companion products: the monthly Index of Production, annual ProdCom manufacturers' sales, and the TOPSI turnover series. Division 26 - computer, electronic and optical products - is where chip work lives, and it arrives in both bases:

Title        : 2026 JUN
CDID         : K23O
Series name  : IOP: 26:Manufacture Of Computer Electronic And Optical Products : CVMSA
Unit         : Index            (chained volume measure, seasonally adjusted)
Series value : 109.7

Title        : 2026 FEB          <- same series K23O, four months earlier
Series value : 106.5

Series name  : UKProdTO: 26 - Manuf of Computer, Electronic and Optical Products TOTAL (£m)
Title        : 2026 JUN          <- turnover series; GBP millions, split TOTAL / EXPORT / HOME

Division 26 real output moved from 106.5 in February to 109.7 in June 2026 - four monthly steps of seasonally adjusted volume on one stable CDID join key. Every series carries that identifier, which is the difference between a shelf you can query and a folder you must re-parse. A head-to-head comparison lays out where the two national records overlap and diverge; both score 9 out of 10 on the catalog rubric against a 7.81 mean across 1,744 datasets.

Far enough to cover every cycle a model is likely to be thrown at. The G.17's long-run statistics reach 1972, five decades in which the US electronics industry absorbed several full boom-and-bust cycles; full spans ship in Datadory deliveries rather than just the recent months of the current release. The ONS IoP panel spans 1,340 monthly date rows - deep enough to hold the financial crisis, the pandemic collapse and rebound, and the post-2022 energy shock inside one continuous grid. JEITA's archive runs 2000 through the current year, twenty-seven yearly tables with no gap.

Two vintage notes matter more than raw depth. First, the G.17's latest month arrives preliminary (revision_status=p) and prior months arrive revised - vintage flags ride on every observation so a model never trains across a splice. Second, the autumn 2026 annual revision moves the G.17 reference base from 2017=100 to 2022=100 and restates history; deliveries keep pre- and post-rebase vintages flagged rather than quietly merged. JEITA's 2016 edition splits around a methodology change, which anyone stitching a long series handles deliberately - once, here, instead of per project.

Indexes measure direction, not money and not borders. Two companion records complete the picture, and both sit in the same industry pool.

For dollar scale, UN Comtrade Plus TradeFlow - Electrical Machinery & Instruments (HS 85/90) serves bilateral reporter x partner x six-digit commodity flows with CIF and FOB values, weights and quantities, at monthly and annual frequency reaching back to 1962. Verified rows show US imports of HS 8541 diodes and transistors at $17.99 billion CIF in 2022, and China-to-US exports of HS 8542 integrated circuits at $2.33 billion FOB in 2023. For a reconciled world total, the Observatory of Economic Complexity - Electrical Machinery & Equipment puts chapter 85 trade at $3.44 trillion in 2024, China exporting $1.06 trillion of it. The practical chain: take the G.17's monthly semiconductor line to time the cycle, then anchor the scale with trade values.

For a ready-made competitiveness frame, the World Bank High-Technology Exports Indicator computes high-tech exports as a share of manufactured exports for 200+ economies annually from 2007 through 2024 - China printed 27.77 percent in 2022 against 30.22 the year before.

Four, all structural rather than incidental.

First, granularity: none of these records names companies. The G.17 stops at NAICS industry lines and market groups, ONS at SIC divisions and ProdCom products, JEITA at product categories - firm-level output requires corporate filings, a different dataset class entirely.

Second, geography is fixed at one country per source. Multi-region coverage means holding the US, UK and Japan panels side by side and respecting that they measure different concepts - production indexes versus shipment volumes. There is no state split inside the G.17 and no regional split inside the IoP grid.

Third, concept discipline: a television assembled in Japan for export never enters JEITA's domestic shipment table, so demand readings cannot be reconciled against output indexes without also pulling trade flows. Fourth, no wafer-fab-specific unit count exists anywhere in the panel - the closest is the G.17's semiconductors and related electronic components subline inside NAICS 334.

Ranked by how directly the panel answers the day job:

1. Investors and quant researchers. The monthly electronics print is core cycle evidence - semiconductor output and utilization lead hardware revenue by quarters, and five decades of history covers every backtest window. See electronic equipment & instruments data for investors and quants.
2. Data scientists and ML engineers. A monthly, revision-flagged exogenous panel drops into forecasting models without reshaping; see demand forecasting.
3. Market researchers and consultants. Official denominators behind every electronics market-sizing deck - one definition per series, no reconciliation exercise.
4. Component buyers and supply-chain planners. Utilization and capacity growth flag whether fabrication capacity is tightening before procurement negotiations price it in.
5. Competitive-intel teams. When the sector index moves ten percent and a company reports the same move, that is beta, not execution - the index proves the difference.
6. Journalists and academics. One citable official indicator instead of stitched trade-group press releases that each define 'electronics' differently.

The constant across all six is the join key - NAICS code or CDID plus reference month - which lets every persona line the same series up against their own numbers without a mapping project.

Because the hard part was never the first number - it is the tenth, and the eleventh month after that. Wide layouts whose row-one titles must become typed columns and row-two identifiers become join keys. Preliminary months revised underneath models already trained on them. A base year moving from 2017=100 to 2022=100 mid-stream. Volume rows sitting beside percent-change rows waiting to corrupt any average computed downstream. Three countries, three concepts, three calendars.

Each is survivable once; none is fun to re-solve every month. Datadory handles them upstream: the wide grids arrive normalized with the same fields and types every edition, consecutive deliveries diff cleanly instead of fighting a new spreadsheet shape, vintage flags keep pre- and post-rebase history separable, and units are typed per row so thousand-unit counts never blend with percentages. Delivery runs daily, weekly, or hourly through API responses, scheduled files, or warehouse-native loads.

Every sample ships with the field dictionary attached and fill rates noted, scoped to the lines, divisions, categories and months you nominate - request a sample and real rows come back before anything recurring switches on.

Go straight to the records: the Federal Reserve G.17 Industrial Production - Electronics, UK ONS Manufacturing and Production Industry and JEITA Electronics Shipment & Production Statistics dataset pages each carry sample rows, the full field dictionary and coverage chips. For browsing rather than reading, the electronic equipment & instruments data hub indexes all sixteen primary records with quality scores, and best electronic equipment & instruments datasets ranks the top ten side by side. Neighboring threads go deeper on adjacent questions: semiconductor billings by region covers the demand-side revenue read from 1986 forward, Japan consumer electronics shipment statistics unpacks the JEITA archive in full, and US electronics trade data by port works the trade layer beneath the production indexes.