37 Million Tons of Wood Get Demolished Every Year. A Robot De-Nails 16,000 Board Feet a Day. Your New Lumber Has a 45% Tariff on It.

Somewhere in Oakland, California, a robotic arm studies a piece of demolition lumber with a 3D camera, locates a rusted nail embedded at an angle no human eye would catch, and extracts it with the precision of a surgeon and the grip of a crowbar. The machine calls itself "Big Bird." It processes 16,000 board feet of salvaged wood per day, enough to frame an entire house.

Meanwhile, the lumber you'd buy new just got more expensive again. Gordian's RSMeans data puts framing lumber at $916.62 per MBF as of April 2026, the ninth consecutive quarter of year-over-year increases, and wholesale SPF 2×4s are running $490 when the same wood cost $383 two years ago.

What changed? The tariff stack happened.

A 45% Wall

The United States imports roughly one-third of its softwood lumber, and Canada supplies 85% of those imports. American sawmills, running at just 64% capacity according to NAHB data, cannot close the gap because this isn't a production choice but a capacity constraint. Mill closures. Equipment aging. Decades of underinvestment.

And it now comes with a serious bill. In 2025, Commerce Department officials more than doubled antidumping and countervailing duties on Canadian lumber from 14.5% to 35%, and a Section 232 proclamation that October added another 10% on all imported softwood, invoking national security. Combined rate: 45%.

For a builder framing a 2,000-square-foot house with approximately 14,000 board feet of lumber, the tariff-driven price increase alone adds roughly $1,500 compared to two years ago. NAHB called it "additional headwinds for an already challenged housing market."

Headwinds is generous.

37 Million Tons, Meet the Robot

Every year, demolition activities in the United States generate 37 million tons of wood debris, according to EPA data. Most of it ends up in landfills or incinerators, not because the lumber has degraded beyond structural utility, but because nobody had the technology or the economic incentive to remove the fasteners embedded in every board.

Pulling fasteners from demolition lumber by hand is brutal, slow work — a skilled laborer might process a few hundred board feet per day. But several things have changed simultaneously: lumber costs spiked 28%, tariffs stacked to 45%, and a startup backed by Google Ventures built a machine that uses computer vision to locate every embedded fastener in a board and robotically extract them.

Urban Machine, founded by construction veteran Eric Law, operates modular de-nailing systems from two 40-foot trailers that deploy directly to demolition sites. AI scans each board, identifies fasteners by type and orientation, and removes them robotically. A final metal detector confirms the wood is clean. At full capacity, the line processes 16,000 board feet of cleaned lumber per day.

At that rate, assuming $30 per hour fully loaded for eight operators, direct labor costs run approximately $0.12 per board foot, about $120 per thousand. Machine costs and transport add more, but the raw material is free.

Urban Machine plans to expand to six US metropolitan areas with a fleet of 12 systems, backed by a seed round from Lower Carbon Capital and GV in 2022.

Beyond De-Nailing

De-nailing is one bottleneck. Just one. Coordination is the harder problem: demolition happens unpredictably, species and dimensions vary wildly, and the volume of usable lumber at any single site depends on factors nobody controls.

Cambium, a Maryland-based startup with $28.5 million in funding including an $18.5 million Series A led by VoLo Earth Ventures, attacks this coordination problem with software by connecting tree care services, trucking, and sawmills into a single supply chain for what it calls "Carbon Smart" wood.

"If you call my uncle and try to sell him wood software, good luck," CEO Ben Christensen told TechCrunch. "Offer to buy 40,000 board feet every 60 days? Different conversation."

Global wood demand is projected to triple by 2050, but the supply chain remains fragmented and analog.

A Missing Standard

Here is where the story gets uncomfortable for anyone hoping reclaimed lumber will replace new SPF in residential framing tomorrow. It won't.

Structural lumber in the United States must be graded to standards maintained by the American Lumber Standard Committee or the National Hardwood Lumber Association, and every board sold for structural use carries a grade stamp certifying its species, moisture content, and load-bearing capacity before a building inspector will approve it. No equivalent grading standard exists for salvaged dimensional lumber, the IRC doesn't address it, and no certification body has proposed a pathway for approving reclaimed wood in load-bearing residential framing applications governed by US building codes. Insurers won't touch it. Building inspectors won't sign off.

Research backs the caution: a 2024 study published in Engineering Structures found that secondary timber consistently shows a lower modulus of rupture than primary timber, owing to aged surfaces, contaminants, and nail holes. Even after de-nailing, planing, jointing, and full refurbishment, reclaimed wood generally does not meet the structural loads that primary engineered timber supports.

This is precisely where AI-driven grading and certification could rewrite the equation.

Cornell University's AR3-Lumber project, funded through the Atkinson Academic Venture Fund, is developing an AI-driven visual grading and certification system for reclaimed lumber in structural applications. Their goal: a production-ready system that grades salvaged wood based on actual measured properties rather than assumed grades, bridging what the researchers call "the gap between circular economy principles and practical industry applications."

Separately, Maestro, born out of MIT's Senseable City Lab, scans raw logs with AI to find optimal board orientations that fit together like puzzle pieces, reducing waste in cross-laminated timber production by up to 30%. Maestro's approach works with irregular shapes (warps, splits, old bolt holes), which is exactly what reclaimed wood tends to be.

Where the Math Works Today

Reclaimed lumber already commands a premium in green building markets because both LEED and the Living Building Challenge reward reclaimed wood use. In these markets, builders pay more for reclaimed. Not less. Green building certification drives the premium here, not cost savings.

For US residential construction, three scenarios pencil out: urban infill near demolition sources, projects pursuing green building certifications where reclaimed wood earns credits offsetting any premium, and non-structural applications like sheathing and blocking where the grading standard gap is irrelevant.

None of this yet works for production homebuilders who need guaranteed supply of code-stamped SPF on a fixed schedule, because a framing crew running six houses a month cannot pause to sort through variable-dimension salvage when their schedule demands uniform 2×4s arriving on a flatbed at seven in the morning. That market still belongs to the lumber yard.

What Needs to Happen

Three things would accelerate the crossover. A nationally recognized grading standard for reclaimed structural lumber, developed with ALSC involvement, would unlock insurance underwriting and code compliance, and AI visual grading like Cornell's would make that standard enforceable at scale. And continued tariff pressure on Canadian imports would keep narrowing the price gap that has historically made reclaimed lumber uncompetitive on cost alone.

The first two are years away. The third is happening now.

At $490 per thousand board feet and climbing, the question for builders is no longer whether reclaimed lumber makes environmental sense. The 37 million tons rotting in landfills answer that. What remains is whether AI can make the grading, processing, and supply chain reliable enough that a framing crew trusts it as much as the SPF stacked in their lumber yard. Machines are getting there. Standards aren't.

Limitations

This analysis relies on publicly available pricing data from RSMeans and Madison's Lumber Reporter. Processing cost estimates for Urban Machine are calculated from publicly stated crew sizes and assumed labor rates, not company-disclosed financials. No peer-reviewed cost comparison of AI-processed reclaimed versus new structural lumber exists at residential scale, and Cornell's AR3-Lumber is a research project, not a commercial product.