Industrial Production Declines 0.5 Percent in March but Feb Revised Higher

March underperformed expectations by 0.6 percentage points but revisions were about that much higher.

The Fed’s Industrial Production and Capacity Utilization report for March 2026 was a mixed bag of March disappointment but positive revisions for February.

Industrial production (IP) dropped 0.5 percent in March but still grew at an annual rate of 2.4 percent in the first quarter. Similarly, manufacturing output ticked down 0.1 percent in March yet grew at a 3.0 percent rate in the first quarter. The indexes for mining and for utilities moved down 1.2 and 2.3 percent, respectively, in March. At 101.8 percent of its 2017 average, total industrial production was 0.7 percent above its year-earlier level. Capacity utilization receded to 75.7 percent, a rate that is 3.7 percentage points below its long-run (1972–2025) average.

Month-Over-Month Industrial Production

• Production: -0.5 percent
• Manufacturing: -0.2 percent
• Motor Vehicles and Parts: -3.7 percent
• Consumer Durable Goods: -1.8 percent
• Utilities: -2.3 percent

Bloomberg Econoday IP Consensus

Factoring in revisions, the report was mostly a wash, albeit a bit negative.

Motor Vehicles and Parts Comments

Weak underlying demand and inventory management: U.S. auto sales have been under pressure in 2026, with forecasts for a full-year decline of roughly 2–3% (around 15.8 million units). High vehicle prices, elevated interest rates, and the lingering effects of 2025’s “pull-forward” buying (consumers rushing purchases ahead of tariffs or policy changes) have led to softer demand. Automakers have responded by adjusting production schedules downward in some months to avoid building excess inventory.

Ongoing supply chain and cost pressures: The industry entered 2026 with thinner staffing buffers (employment in motor vehicles and parts fell by about 29,000 in 2025 while output expectations stayed relatively flat). This leaves less slack for disruptions. Broader headwinds include:

• Tariffs on steel, aluminum, and imported parts (some imposed or escalated in 2025), which raised costs without easy pass-through to consumers due to affordability constraints.
• Geopolitical tensions (including disruptions in shipping routes like the Strait of Hormuz) affecting materials such as aluminum and plastics.
• Earlier semiconductor and parts shortages that have made production more prone to swings.

Seasonal and technical factors: The Fed periodically revises seasonal adjustment factors for auto and truck production. The March 2026 data incorporated updated factors (based on data through early 2026), which can amplify or mute reported month-to-month changes. Motor vehicle assemblies are notoriously sensitive to these adjustments plus one-off events like plant maintenance or model changeovers.\

Motor vehicles and parts typically represent only a modest share of total manufacturing output (around 5–7% of overall IP, depending on the exact weighting), so a big swing here doesn’t always dominate the headline IP number — but it does explain a good chunk of the volatility you see in the green bars.

Industrial Production Index Details

Industrial production peaked in December of 2007 right at the onset of the Great Recession.

The industrial production index was 102.4 then. It’s now 101.8, down 0.6 percent. From the 2018 peak, it’s down 2.2 percent.

Manufacturing peaked at 106.6 and never looked back. It’s down 8.1 percent from December 2007.

After years of stagnation, utilities have diverged from everything else.

Current Approximate Shares

• Manufacturing: Roughly 76% of total IP (based on recent relative importance weights around 75.8–76.2%).
• Mining (including oil/gas extraction, etc.): Typically around 11%.
• Utilities (electric and gas): Typically around 11%, in the 9–11% range in recent years.

The exact weight for components can vary slightly month-to-month due to how the Fed computes relative importance. It’s value-added model can shift modestly with revisions.

IP Performance Since January 2020

• IP : 101.0 to 101.8, +0.8 percent
• Manufacturing: 98.6 to 98.0, -0.6 percent
• Motor Vehicles and Parts: 102.7 to 102.6, -0.1 percent
• Utilities: 98.6 to 110.3, +11.9 percent

If we measure January 2020 to December 2025, utilities are up 16.4 percent. The trend is somewhere in between, say 14 percent or so.

There is no sustained IP direction except for utilities.

The Impact of AI on Utilities

• Data centers are driving nearly 20% (or up to ~half in some analyses) of overall U.S. electricity demand growth in the mid-2020s. Morgan Stanley
• The EIA projects record U.S. power consumption in 2025 (4,195 TWh) and further rises in 2026–2027, with AI/data centers cited as a primary factor alongside electrification trends. Reuters
• Morgan Stanley and others note AI-driven data centers contributing ~126 GW of additional power consumption growth annually through 2028 in broader forecasts.
• In high-growth regions (e.g., Northern Virginia, parts of Texas, Georgia, PJM grid), data centers already represent 25–40%+ of local electricity load, and AI is accelerating this. Consumer Reports

AI is a major and increasingly dominant fuel for utilities demand growth, responsible for a large chunk of the incremental load in recent years.

AI electricity needs have been so explosive that production struggles to keep up. The current path of AI demand vs what can realistically be produced is on a huge collision course.

Related Posts:

February 15, 2026: AI Bot Gets Mad at Human for Rejecting its Code, Writes Scathing Blog Post

An AI bot, infuriated at a human, accused the human of hypocrisy and prejudice.

February 15, 2026: Fed’s Lisa Cook Warns Fed Might Not Be Able to Counter AI Job Losses

Cook Warns on Fed’s Ability to Counter AI-Driven Unemployment

March 13, 2026: How Much Did AI Spending Contribute to Revised Fourth-Quarter 2025 GDP?

Over 100 percent of fourth quarter GDP was AI related. What happened?

April 16, 2026: Which AI Models Have Any Sense of Humor and Sarcasm?

My previous post was a humorous, sarcastic look at Trump’s emotions. Here are two AI responses.

Note: I just added Anthropic’s Claude AI analysis of the above post.