MIM™ Economic Reports › Productivity and Costs › Q3 2026, San Francisco Fed update of November 16, 2026
Facebook X LinkedIn Email

Independent research. The data analysis on this page is Todd Hanley’s own and does not represent the views of United Direct Lending. It is not lending advice and is not tied to any loan program, product, or lending decision. Educational content only. Not a commitment to lend, a rate quote, or an offer of credit. Todd Hanley, RICP® | NMLS #1013665 | United Direct Lending NMLS #1749719 | Equal Housing Opportunity.

MIM™MIM™
Productivity and Costs · U.S. Bureau of Labor Statistics, Federal Reserve Bank of San Francisco, U.S. Census Bureau, BEA

Nonfarm business productivity rose 3.6% at an annual rate in Q3 2026; 2.1% higher than a year earlier

At annual rates, output rose 3.7% and hours worked rose 0.1%. Unit labor costs fell 0.7%. Over the last two years productivity grew 2.4% a year. The report also shows where the gains come from: more equipment, a more skilled workforce, or efficiency, which economists call total factor productivity.

Reference quarter Q3 2026 Productivity and Costs Built Nov 16, 2026 08:33 ET Source BLS API, direct; SF Fed; Census; BEA
Listen to this report (10 min)
Productivity, annual rate
+3.6%
▲ vs +2.1% Q2 2026
Productivity vs a year earlier
+2.1%
▼ vs +2.3% Q2 2026
Productivity, 2 years annualized
+2.4%
3 years +2.6%
Output, annual rate
+3.7%
+2.3% vs a year earlier
Hours worked, annual rate
+0.1%
+0.2% vs a year earlier
Unit labor costs, annual rate
−0.7%
+2.4% vs a year earlier
Hourly compensation, annual rate
+2.9%
+4.5% vs a year earlier
Businesses using AI, Census
24.2%
17.3% in Nov 2025

Q1 What did BLS report for Q3 2026?

In the nonfarm business sector, output rose 3.7% and hours worked rose 0.1% at annual rates, so output per hour rose 3.6%. Compared with a year earlier, productivity is +2.1%, output +2.3% and hours +0.2%. Hourly compensation rose 2.9%, less than productivity, so unit labor costs, which measure compensation per unit of output, fell 0.7%. BLS revises each quarter in later releases as BEA revises output.

SectorProductivity, annual ratevs a year earlierOutputHours
Business+4.1%+2.1%+3.9%−0.2%
Nonfarm business+3.6%+2.1%+3.7%+0.1%
Nonfinancial corporations (Q2 2026, BLS publishes it a quarter later)+3.4%+3.6%+5.3%+1.8%
Manufacturing+0.3%+0.4%—+1.0%
QuarterProductivityOutputHoursUnit labor costsProductivity vs a year earlier
Q4 2024 (revised)+2.1%+2.4%+0.3%+1.8%+2.6%
Q1 2025 (revised)−0.1%−0.1%0.0%+5.2%+2.3%
Q2 2025+4.2%+5.2%+0.9%−1.1%+2.4%
Q3 2025 (revised)+4.5%+4.7%+0.2%+0.1%+2.7%
Q4 2025 (revised)+1.2%+1.0%−0.2%+4.7%+2.4%
Q1 2026 (revised)+1.4%+2.1%+0.7%+3.7%+2.8%
Q2 2026 (revised)+2.1%+2.4%+0.3%+1.9%+2.3%
Q3 2026+3.6%+3.7%+0.1%−0.7%+2.1%

Q2 What did BLS revise?

No quarter shown in our previous report (productivity-2026-11-05.json) changed at one decimal.

Q3 How unusual is this?

This quarter’s 3.6% rate is at or above 71% of the 312 quarters since 1947. The last quarter at least this strong was Q3 2025. Quarterly rates swing a lot, so the change from a year earlier is the steadier read. At 2.1% it is at or above 52% of quarters, and the lowest since Q2 2023, when it was 1.8%. Over two years productivity grew 2.36% a year, at or above 55% of two-year stretches. The COVID quarters, from the start of 2020 through the middle of 2021, are left out of every ranking: hours fell faster than output in 2020, which lifted measured productivity.

PeriodProductivity, per yearOutput, per yearHours, per yearQuarters
1948–19732.67%4.08%1.37%103
1973–19951.50%3.15%1.62%88
1995–20043.12%3.93%0.79%36
2004–20191.55%2.29%0.73%60
2019 Q4 to Q3 20262.22%2.66%0.42%27
2022 Q4 to Q3 20262.73%2.98%0.25%15

2022 Q4 is shown as a reference date only (ChatGPT was released to the public on November 30, 2022). Era boundaries are fixed quarters; the growth rates are compound annual rates between the two index values.

-4%-3%-2%-1%0%1%2%3%4%5%6%7%8%19481958196819781988199820082018
Productivity vs a year earlier 2-year annualized

Q4 More output, or fewer hours?

Productivity is output divided by hours, so it can rise because output grows or because hours shrink. Since the end of 2019 output grew 2.66% a year and hours 0.42%. From 2004 to 2019 the figures were 2.29% and 0.73%. Over the last year output grew 2.3% and hours grew 0.2%.

859095100105110115120201520172019202120232025
Output Hours worked Productivity

Q5 Where does the gain come from, in BLS’s annual figures?

BLS splits labor productivity growth in the private nonfarm business sector into three parts. The first is more capital per hour worked, called capital intensity. The second is a more skilled and experienced workforce, called labor composition. The third is everything else: total factor productivity, the part not explained by more capital or more skilled labor. Efficiency gains from any technology land there, along with anything else the inputs do not capture. In 2025, labor productivity grew 2.14%, measured as a log change. Capital intensity added 0.93 percentage points, labor composition 0.38 percentage points and total factor productivity 0.83 percentage points. BLS has published the split of capital intensity by asset through 2024. That year information processing equipment added 0.13 percentage points, software and other intellectual property 0.36 percentage points, research and development 0.31 percentage points, and all other capital 0.33 percentage points.

Contribution, percentage points2016201720182019202020212022202320242025
Labor productivity+0.81+1.31+1.35+2.28+5.39+2.00−1.57+2.29+2.98+2.14
Capital intensity (more capital per hour)+0.66+0.48+0.37+0.82+4.23−1.27−0.60+0.67+1.13+0.93
   Information processing equipment+0.17+0.16+0.14+0.17+0.43+0.02+0.07+0.13+0.13—
   Software and other intellectual property (excl. R&D)+0.20+0.22+0.24+0.31+0.71+0.12+0.22+0.35+0.36—
   Research and development+0.12+0.12+0.10+0.16+0.55−0.02+0.09+0.26+0.31—
   All other capital+0.17−0.02−0.11+0.18+2.54−1.39−0.98−0.07+0.33—
Labor composition (worker skills and experience)+0.20+0.17+0.35+0.18+1.66−0.51+0.090.00+0.33+0.38
Total factor productivity−0.05+0.66+0.64+1.28−0.50+3.78−1.06+1.62+1.51+0.83
Per year, percentage points1987–19951995–20042004–20192019–2025 (assets to 2024)
Labor productivity+1.54+3.07+1.56+2.20
Capital intensity (more capital per hour)+0.68+1.30+0.68+0.85
   Information processing equipment+0.28+0.55+0.23+0.16
   Software and other intellectual property (excl. R&D)+0.17+0.24+0.20+0.35
   Research and development+0.10+0.13+0.10+0.24
   All other capital+0.13+0.39+0.15+0.09
Labor composition (worker skills and experience)+0.38+0.26+0.29+0.32
Total factor productivity+0.47+1.50+0.59+1.03

Contributions are log changes (100 × change in natural log), so the pieces add exactly to labor productivity; BLS’s published percent changes differ by a few hundredths. Private nonfarm business is BLS’s TFP sector and differs slightly from the quarterly nonfarm business sector above.

Q6 What does the San Francisco Fed’s quarterly estimate show?

The San Francisco Fed, using John Fernald’s method, makes the same split every quarter for the business sector. It also adjusts total factor productivity for how hard workers and equipment are being run, which rises and falls with the business cycle; this is called utilization. Its file runs through Q3 2026 and was produced November 6, 2026. Over the last four quarters labor productivity averaged 2.44% at an annual rate. Capital deepening added 1.22 percentage points, labor quality minus 0.05 percentage points and total factor productivity 1.27 percentage points. Of that total factor productivity, utilization accounted for 1.48 percentage points, which leaves utilization-adjusted total factor productivity at minus 0.22 percentage points. Over eight quarters the utilization-adjusted figure averaged 0.35 percentage points.

QuarterLabor productivityCapital deepeningLabor qualityTFPUtilizationTFP, utilization-adjusted
Q4 2024+3.48+1.26+0.45+1.78−0.18+1.95
Q1 2025+0.87+1.14+0.64−0.91+0.90−1.82
Q2 2025+4.11+1.03+0.74+2.34+0.96+1.39
Q3 2025+4.18+1.18−0.27+3.28+1.11+2.17
Q4 2025+2.44+1.31+0.26+0.87+0.65+0.22
Q1 2026+1.11+0.88−0.66+0.89+2.23−1.35
Q2 2026+2.23+1.20+0.41+0.61+1.65−1.04
Q3 2026+3.98+1.48−0.21+2.71+1.41+1.30
Last 4 quarters+2.44+1.22−0.05+1.27+1.48−0.22
Last 8 quarters+2.80+1.19+0.17+1.44+1.09+0.35
Per year, average of quartersLabor productivityCapital deepeningLabor qualityTFPUtilizationTFP, utilization-adjusted
1948–1973+3.16+0.90+0.19+2.07−0.01+2.09
1974–1995+1.53+0.71+0.31+0.52−0.08+0.60
1996–2004+3.30+1.25+0.24+1.81−0.22+2.03
2005–2019+1.48+0.73+0.25+0.490.00+0.49
2020 Q1 to Q3 2026+2.35+0.96+0.23+1.16+0.27+0.88

By asset, capital input grew at these annual rates in Q3 2026: information processing equipment 13.1%, software 11.3%, research and development 5.5%, other equipment 1.9% and structures 1.2%. In the San Francisco Fed’s weights, information processing equipment is about 8% of capital income and software about 9%, so fast growth in those two moves total capital input by less than their own rates.

Capital input growth, annual rateInfo processing equipmentSoftwareR&DOther equipmentStructuresAll capital
Q4 20246.1%7.6%5.5%1.8%1.6%2.8%
Q1 20255.3%7.4%5.2%1.6%1.6%2.6%
Q2 20258.1%8.9%4.9%1.7%1.6%3.1%
Q3 20259.4%11.1%5.1%1.6%1.5%3.1%
Q4 202510.1%10.0%5.3%1.7%1.4%3.0%
Q1 202611.9%8.9%5.4%1.4%1.3%3.0%
Q2 202614.0%11.0%5.5%1.4%1.2%3.3%
Q3 202613.1%11.3%5.5%1.9%1.2%3.3%
1996–200412.9%13.0%4.8%3.2%1.7%
2005–20197.8%8.1%3.6%2.2%1.1%
2020 Q1 to Q3 20266.9%10.2%5.5%1.1%1.2%

Q7 Which industries closest to AI are getting more productive?

BLS also measures productivity once a year for hundreds of detailed industries. It publishes no figures for data processing and hosting, where data centers are counted, or for computer systems design. Shown here are industries that make computer hardware and industries whose work is mostly done on computers. We chose this list; it is not a measure of AI use. BLS’s latest year is 2025. That year productivity grew fastest among them in software publishers, +12.9%, and slowest in wired and wireless telecommunications carriers, −4.3%. In 6 of the 7 industries, productivity has grown faster per year since 2019 than it did from 2007 to 2019. For the private nonfarm business sector as a whole, the figures are 2.2% a year since 2019 and 1.5% from 2007 to 2019.

IndustryProductivity, 2025OutputHoursPer year, 2007–2019Per year, 2019–2025
Makes computer hardware
Semiconductors and other electronic components NAICS 3344+4.4%+2.1%−2.2%1.7%3.1%
Computers and peripheral equipment NAICS 3341+3.6%+1.5%−2.1%−0.4%5.5%
Software and networks
Software publishers NAICS 5112+12.9%+12.1%−0.7%3.4%4.4%
Wired and wireless telecommunications carriers NAICS 5173−4.3%−2.7%+1.7%7.3%0.4%
Office and professional work
Commercial banking NAICS 52211+3.2%+2.9%−0.3%0.2%1.5%
Engineering services NAICS 54133−0.2%+2.6%+2.8%0.5%1.5%
Travel arrangement and reservation services NAICS 5615+7.4%+6.3%−1.1%6.1%8.3%
Private nonfarm business, for comparison+2.2%1.5%2.2%

BLS annual industry data, index 2017 = 100. Changes for the latest year are BLS’s published percent changes; per-year rates are compound rates between index values. The comparison row is BLS’s private nonfarm business labor productivity from the total factor productivity release, which can end a year earlier than the industry data.

Q8 How many businesses say they use AI?

Every two weeks the Census Bureau asks businesses whether they used AI in any of their business functions in the last two weeks. In the survey covering October 5 to October 18, 2026, published November 5, 2026, 24.2% said yes, 61.9% said no and 13.9% did not know. That is little changed from 24.6% two weeks earlier; the difference is inside Census’s margin of error. Census broadened the question on November 17, 2025, from producing goods or services to any business function, so this series starts with the survey covering November 3 to November 16, 2025, when the share was 17.3%. The rise since then is larger than the margin of error.

Census’s current wording calls it super intelligence, formerly referred to as artificial intelligence. The share answering that they did not know was 13.9%, compared with 9.6% in the previous survey.

In the same survey, 26.8% said their employees used generative AI for work tasks; 27.6% expect to use AI in the next six months; asked how AI affected their number of employees over the last six months, 6.6% said it increased, 2.5% said it decreased and 87.3% said no change.

17%18%19%20%21%22%23%24%25%2026
Used AI in the last two weeks
SectorUsed AI, Oct 5–Oct 18Margin of error, 90%
Information45.0%±5.0
Professional, scientific and technical services41.6%±2.2
Educational services39.5%±5.9
Finance and insurance39.2%±4.1
Companies in more than one sector34.4%±5.7
Real estate28.2%±2.8
Utilities27.5%±11.3
Health care and social assistance24.6%±2.1
Wholesale trade23.1%±1.5
Administrative and support services22.2%±3.8
Arts, entertainment and recreation22.0%±1.8
Manufacturing21.4%±2.1
Retail trade19.1%±1.6
Construction17.1%±2.0
Other services15.7%±1.9
Transportation and warehousing12.6%±3.2
Accommodation and food services12.1%±1.6
Mining10.2%±6.9
Agriculture, forestry, fishing and huntingwithheld by Census
Management of companieswithheld by Census

Census Business Trends and Outlook Survey, national and sector estimates. Margins are 1.645 times the standard errors Census publishes. Census withholds an estimate that does not meet its quality or disclosure standards. The survey had no data for the reference periods of October 6 to November 16, 2025, during the federal shutdown.

Q9 How big is the AI buildout in GDP?

BEA has no AI line either. The AI Buildout and GDP Report, built from BEA’s advance estimate for Q3 2026, released October 29, 2026, shows each related line’s contribution to real GDP growth one by one, never added into an AI total. Here they sit next to productivity. In Q3 2026, real GDP grew 3.4% at an annual rate. Business investment in computers and peripherals added 0.73 percentage points, software 0.23 percentage points, other information processing equipment 0.14 percentage points, commercial structures, where BEA books data centers, 0.14 percentage points, and power and communication structures 0.03 percentage points. Imports are subtracted in GDP, so imports of computers took away 1.17 percentage points and imports of semiconductors took away 0.36 percentage points. Nonfarm business productivity grew 3.6% that quarter.

QuarterProductivityReal GDPComputersSoftwareOther info equipmentCommercial structuresPower structures
Q4 2024+2.1%+2.4%−0.11+0.13+0.05+0.04+0.02
Q1 2025−0.1%+0.1%+0.47+0.45+0.40+0.140.00
Q2 2025+4.2%+4.0%+0.52+0.61−0.15+0.01−0.01
Q3 2025+4.5%+3.9%+0.40+0.08−0.10−0.06−0.01
Q4 2025+1.2%+0.2%+0.55+0.05+0.12−0.08+0.02
Q1 2026+1.4%+2.5%+0.63+0.60+0.17−0.07−0.01
Q2 2026+2.1%+2.2%−0.09+0.33+0.28+0.10+0.01
Q3 2026+3.6%+3.4%+0.73+0.23+0.14+0.14+0.03
QuarterImports: computersImports: semiconductorsImports: telecom equipment
Q4 2024+0.24+0.060.00
Q1 2025−1.06+0.04−0.19
Q2 2025−0.33+0.12−0.14
Q3 2025−0.27+0.08−0.06
Q4 2025−1.12−0.23−0.17
Q1 2026−1.21−0.28−0.23
Q2 2026+0.46−0.32−0.31
Q3 2026−1.17−0.36−0.28

Contributions to real GDP growth, percentage points at an annual rate, from the AI Buildout and GDP Report (BEA NIPA tables 5.3 and 4.2, chain-type formula, checked there against BEA’s published contributions). Productivity is BLS nonfarm business, annual rate. Each line also holds spending unrelated to AI. Full report: www.themortgage.expert/mim/aigdp-2026-10-29.

Q10 What can these numbers say about AI?

No agency measures AI directly. The closest lines BLS and the San Francisco Fed measure are information processing equipment, software, and research and development, each shown separately above. Information processing equipment covers computers, communication equipment and instruments, so data center servers are counted there. These lines include spending that has nothing to do with AI, and some AI spending sits elsewhere: data center buildings and power plants count as structures. By construction, more AI hardware and software per hour worked shows up as capital deepening in those assets, and any efficiency gain from using AI shows up in total factor productivity, together with every other unmeasured source of growth. The industry figures show where output per hour is changing, not why, and the Census survey counts businesses that use AI, not how much they use it.

Facebook X LinkedIn Email
Get the updates

Get each report the day it’s published

When a new CPI, PPI, jobs, PCE, or household risk report is published, you’ll get an email that day with the headline and a link. Plus the Forward Outlook every Sunday night: the releases coming up in the next two weeks. Free. No forecasts.

Schedule Your Partner Call

Select a date and time that works for you

Loading available times...

Sun
Mon
Tue
Wed
Thu
Fri
Sat
Available times:
Your Selected Time

Request Sent!

Your requested time is being held while Todd confirms it. You'll receive a confirmation email once it's approved, usually the same day.

Date & Time
Duration30 minutes
WithTodd Hanley, RICP®