A site superintendent in Chandler, Arizona, watched a 36-ton Hitachi excavator load articulated dump trucks for eight hours last November without a human touching the controls. The machine ran a Bedrock Robotics retrofit kit: LiDAR, cameras, GNSS, and a compute box bolted to an excavator the crew had been running manually for three years. Same bucket, same tracks, same hydraulic lines, but a completely new brain bolted where the lunch cooler used to sit. It worked.
That deployment, part of a 130-acre manufacturing facility prep with Sundt Construction, moved over 65,000 cubic yards of earth using supervised autonomy, which means a human monitored the machines remotely but did not operate them. Bedrock calls it the largest supervised autonomy deployment for mass excavation in the US construction industry. The excavators ranged from 20 to 80 tons, and the retrofit kit installed in a single day, reversibly, so the machines could flip back to manual operation between autonomous tasks.
Three Kits, Three Price Points, Real Numbers
Forget the press releases. Seriously. Here is what the retrofit autonomy companies are actually selling, what they charge, and what the independently verifiable data says.
Bluelight Machines sells retrofit kits for rollers and haul trucks at $40,000 to $50,000 per unit, with rental and subscription options. CEO Dick Zhang told Equipment World the kits install in hours, take minutes to learn, and operate as a "buddy system" where the autonomous machine pairs with a human-operated dozer or grader. Zhang's numbers: 14% higher site-wide output by reallocating operators to higher-value tasks, with idle time dropping from 70% to 7%. Over one billion square feet compacted autonomously with zero injuries across four continents, with hundreds of live machines running today.
Gravis Robotics debuted the Gravis Copilot at CONEXPO 2026 in Las Vegas, a manufacturer-agnostic retrofit platform that works across mixed OEM fleets. Deployed in seven countries with partners including Hitachi, Develon, Holcim, and Taylor Woodrow. Their published numbers from live deployments: 30% increase in operator productivity, 97% bucket fill rates, and estimated annual net savings of $74,000 or more per machine. A 60-mile autonomous oil and gas pipeline project with Techint Group in Argentina ran multiple 36-ton excavators digging six-foot trenches through compacted soil and shallow bedrock. Gravis raised $23 million in its most recent round, a fraction of Bedrock's war chest.
Bedrock Robotics is the gorilla in the room, founded by ex-Waymo engineers, led by CEO Boris Sofman, and funded with $350 million total including a $270 million Series B in February 2026 co-led by CapitalG and Valor Equity Partners, with NVIDIA's venture arm participating, which pushed the company's valuation to $1.75 billion. Their system learns by watching experienced operators remotely and replicating their patterns. Compatible with excavators from 20 to 80 tons. Currently deployed on active sites in Arizona, Texas, and Arkansas. Bedrock targets fully operator-less excavator deployments by end of 2026, which would make it the first company to put heavy construction equipment into production without anyone sitting in the cab, watching a screen, or holding a kill switch within arm's reach.
And then there is the incumbent, moving slower but carrying bigger iron. Caterpillar unveiled autonomous excavators, haul trucks, dozers, and compactors at CES 2026, backed by 30-plus years of autonomy research and a fleet that has moved 11 billion tonnes across 380 million kilometers, numbers that make every startup's deployment stats look like a rounding error. Cat pledged $25 million over five years for workforce education related to autonomous roles. When both startups and the world's largest equipment manufacturer converge on the same technology simultaneously, the market is past the question of whether this works. It does.
Residential Math: Tighter Than You Think
Most of the deployment data comes from large infrastructure projects. Hundred-acre sites. Sixty-mile pipelines. Strip mines where a single haul truck runs 20 hours a day. Residential site prep is a different animal, and the economics deserve a calculator rather than a press release.
Site preparation for a single residential lot runs $15,000 to $50,000, according to TrueBuiltHome, covering clearing, grading, excavation, and utility trenching. Equipment operator labor for the grading and compaction portion typically accounts for $8,000 to $12,000 of that cost on a standard lot in flat-to-moderate terrain, and those grading and compaction hours are precisely the tasks that autonomous kits are designed to absorb.
Run the math on a 20-lot subdivision. If you realize a 14% productivity gain on equipment tasks (Bluelight's number), you save roughly $1,120 to $1,680 per lot on grading and compaction labor, numbers that sound modest until you multiply them across every lot in the plat. That is $22,400 to $33,600 in total savings. Not nothing. Break-even on a $50,000 kit lands break-even on a $50,000 kit somewhere between the first and second subdivision you grade with it.
Use Gravis's 30% productivity number and the same 20-lot project, and the savings climb to $48,000 to $72,000, meaning the kit pays for itself before you finish grading the first community. Numbers like that sell themselves.
Both numbers come with a caveat large enough to park a D6 in. Do the math anyway.
Why Residential Sites Are Harder
Infrastructure projects are where autonomous equipment shines because the work is repetitive, the sites are vast, and the obstacles are mapped in advance. Residential lots are none of those things, and that gap between a 130-acre manufacturing pad and a 7,500-square-foot infill lot matters more than any productivity percentage on a marketing slide. with a three-foot setback from the property line, utility stubs poking out of the ground, a retaining wall on the uphill side, a neighbor's fence 12 feet from where the compactor needs to turn around. That is a typical infill lot in a suburban subdivision, and every one of those obstacles creates a scenario where the machine either needs human intervention or risks damage that exceeds the labor savings.
None of the three retrofit companies have published case studies from residential projects. Every deployment highlighted in their marketing materials involves infrastructure, commercial site work, mining, or pipeline construction. When I asked about residential applications, the answer from every direction was some version of "our technology is capable of working on smaller sites," which is marketing for "we haven't proven it there yet," and you should note that gap before writing any checks.
Insurance is the other unresolved variable, because autonomous equipment operating on an active residential construction site raises liability questions that the insurance industry has barely started answering. Who is at fault when a supervised autonomous excavator damages a water main that the utility company mislocated: the operator who was monitoring remotely, the retrofit kit manufacturer, or the general contractor who authorized autonomous operation? Your current general liability policy almost certainly does not contemplate this scenario, and adding coverage for autonomous equipment operations, a product category that most underwriters have never evaluated and for which actuarial loss data essentially does not exist, may cost enough to narrow the labor savings gap that justified the kit in the first place.
When the Kit Wins Without the Math
Here is what the spreadsheet misses. The real value proposition for residential builders is not the 14% or 30% productivity gain. It is the Tuesday morning when your roller operator does not show up and you have a compaction spec that the inspector is checking at 2 PM. You cannot hire a replacement in six hours. You cannot slip the inspection without slipping the foundation pour without slipping the framing schedule without blowing the closing date.
One operator supervising one autonomous roller and running a second machine manually means your compaction gets done regardless of who called in sick, regardless of the staffing agency's hold time, regardless of every variable between you and a compaction spec that an inspector is checking at 2 PM. Schedule saved. That is worth more than $50,000 to a builder who has lost a closing date because a roller sat idle for three days while the staffing agency searched for an operator with a CDL and the right endorsements.
For subdivision developers running 50 or more lots per year with dedicated grading crews, the financial math works now, especially if you are leasing the kits at $3,000 to $4,000 per month instead of buying outright. For a custom builder doing five homes a year on varied lots with tight site conditions, the kit is probably a curiosity for another 18 to 24 months, until residential-specific deployments generate the case studies and insurance products that close the remaining gaps.
What You Should Actually Do
If you run a subdivision with 20-plus lots and your site prep subcontractor regularly misses schedule due to operator shortages, ask Bluelight or Gravis for a pilot. Both offer rental and subscription pricing, so you can test without committing $50,000 to a kit you might not keep. Insist on seeing their residential site data before signing anything longer than a three-month trial. If they do not have residential data, you are the pilot, and you should negotiate pricing that reflects the risk of being first, because the lessons learned on your lots will be worth more to them than your subscription fee.
If you run a custom home business with fewer than ten projects a year, watch this space but do not buy anything yet. The technology is real, the infrastructure deployments prove it works, but residential proof points are not there yet. When Bluelight or Gravis publishes a case study from a 30-lot suburban subdivision with documented cost savings, that is when the conversation shifts from interesting to actionable for smaller operators. Not before.
If you are buying a home and wondering whether autonomous site prep affects the price you pay, it does not. Not yet. Site prep is a line item buried in the builder's total cost, and the labor savings from autonomy are small relative to the total project budget. What it might affect is timing, because the builder who cannot find operators is the builder who misses every milestone downstream of grading, and grading is the first domino in a chain that ends at your closing date.
What This Analysis Did Not Prove
All productivity and savings claims in this article come from the companies selling the kits. Bluelight's 14% figure, Gravis's 30% and $74,000 annual savings, Bedrock's deployment volume data: none have been independently verified by third-party researchers or published in peer-reviewed studies. Construction Dive, Equipment World, and other trade outlets have reported the claims but not audited them.
The residential break-even calculation extrapolates from infrastructure deployment data to residential lot conditions. No retrofit autonomy company has published residential-specific productivity data, cost savings, or deployment case studies. The $8,000 to $12,000 operator labor estimate for grading and compaction on a standard residential lot is derived from contractor pricing surveys, not from metered data on autonomous versus manual operation on identical lots.
Long-term maintenance costs for retrofit kits remain unknown. The kits include LiDAR sensors, compute hardware, cameras, and GNSS receivers operating in dusty, high-vibration environments. Component replacement intervals and associated costs have not been published by any of the three manufacturers.
Insurance and liability frameworks for autonomous construction equipment on residential sites are effectively nonexistent. The regulatory landscape for supervised autonomy on construction sites varies by jurisdiction and remains unsettled in most states.