Brian Potter challenges a deeply held assumption about the future of work: that the high cost of custom, low-volume production is an immutable law of economics. In this piece from Construction Physics, he argues that the convergence of advanced AI, flexible robotics, and cloud-like infrastructure could shatter the historical link between volume and efficiency, turning physical manufacturing into a utility as elastic as Amazon Web Services. For busy leaders watching inflation eat into margins, this isn't just theory; it's a potential blueprint for a world where the cost of a one-off repair or a bespoke prototype drops to near-zero.
The Repetition Trap
Potter begins by dismantling the logic that has driven industrial efficiency for a century. He points out that historically, making things cheaper required doing the exact same thing over and over again. "Higher production volume means larger, more efficient factories," Potter writes. "It means more opportunities to use dedicated, high-speed, continuous process production equipment." This is the fundamental reason why a new car gets cheaper every year, while fixing a dent in that same car gets more expensive. The car is a product of mass repetition; the repair is a unique, low-volume event that resists automation.
He illustrates this with a stark comparison: "Cars are manufactured via a repetitive, high-volume process that spits out nearly identical models by the hundreds of thousands or millions. Car manufacturers can justify spending billions of dollars designing a new model of car and the process for making it, because that cost will get spread out over a huge number of cars." In contrast, repairing a car involves "a huge number of different models and model years, each damaged in different ways," making it impossible to justify the capital expenditure for a dedicated, automated repair line. This framing is powerful because it isolates the variable that has long been the bottleneck: the inability to amortize the cost of automation setup over a large enough batch of identical units. It echoes the logic behind Design for Manufacturability discussed in Potter's companion deep dives, where the goal was always to force the unique into the mold of the repetitive.
"If you're only going to run your process once, or just a handful of times, these opportunities are considerably narrowed."
Critics might argue that this view underestimates the sheer complexity of physical reality compared to digital code. While software can be cloned instantly, physical materials have friction, tolerances, and supply chain constraints that don't vanish just because an algorithm gets smarter. Potter acknowledges this, noting that even the Toyota Production System, which introduced flexibility to mass production, didn't change the "fundamental logic" that volume drives down cost. However, he suggests that logic is now being rewritten.
The AWS of Physical Things
The core of Potter's argument rests on a shift in automation technology. For decades, robots were dumb, rigid, and expensive to program. They could weld a chassis perfectly a million times, but if the model changed, the line had to stop. Now, with computer vision and AI, machines can adapt on the fly. Potter sees a future where the "minimum efficient scale"—the size a factory needs to be to be competitive—shrinks dramatically. He envisions two paths forward: tiny, flexible microfactories, or massive, centralized hubs that produce millions of different products with the same efficiency as a single product line.
He coins the term "AWS for everything" to describe the latter. Just as Amazon Web Services allows a startup to rent computing power without building a data center, Potter imagines a world where a company rents time on a massive, flexible production line. "You can very quickly use Amazon's infrastructure to perform whatever computing task you're interested in... without needing to build or operate any of that infrastructure yourself," he notes, drawing the parallel to physical goods. The key is a "software control layer" that can route materials and tasks dynamically. "If your software is smart enough, and your equipment flexible enough, you can set up some new process to produce some new widget on the fly," Potter writes.
This is where the argument gets most provocative. It suggests that the future of manufacturing isn't about building more factories, but about building smarter, shared infrastructure. He points to existing players like SendCutSend, which uses software to nest hundreds of different customer parts onto a single sheet of metal, achieving near-perfect material utilization. "The key with high mix is that it actually works at scale," says SendCutSend founder Jim Belosic, quoted by Potter. "The larger volume of high mix, the easier things get... We only do one setup, for potentially dozens or hundreds of customers, we do one load into the machine, we only retrieve the material once."
"The larger volume of high mix, the easier things get... We only do one setup, for potentially dozens or hundreds of customers, we do one load into the machine, we only retrieve the material once."
Potter extends this logic beyond metal cutting to biotech, citing a vision of "cloud labs" where centralized automation runs experiments for thousands of researchers. "If you're running enough experiments such that your robots are constantly humming, you can justify producing your own reagents," he explains. This creates a flywheel: lower costs attract more demand, which justifies more automation, driving costs down further. It's a compelling narrative of deflationary pressure on the physical world, mirroring the deflationary trends we've seen in software.
The Human Element and the Road Ahead
However, Potter is careful not to paint a picture of a fully automated, human-less future. He suggests that in the near term, the "connective tissue" of these flexible systems might still be human workers, directed by software. "You can imagine humans acting as much of the 'connective tissue' in this sort of production process, being directed by software telling them where to go and what to do to maximize utilization," he writes. This is a pragmatic take that avoids the sci-fi trap of total replacement, acknowledging that human dexterity and judgment are still superior for certain variable tasks. It recalls the Baumol effect discussed in his other work, where services that resist automation (like care or repair) become relatively more expensive; Potter's thesis is essentially a roadmap for breaking that effect by making the "service" of manufacturing itself highly automatable.
A counterargument worth considering is the energy and capital intensity of this model. Building the "huge warehouses filled with all sorts of different machines" requires massive upfront investment. While the marginal cost per unit might drop, the fixed costs could create new barriers to entry, potentially consolidating power in the hands of a few infrastructure providers rather than democratizing production. Potter hints at this, noting that these operations "wouldn't displace traditional mass-production style processes" but would exist alongside them, suggesting a hybrid future rather than a total revolution.
Bottom Line
Potter's most compelling insight is that the barrier to low-volume production is no longer physical capability, but economic justification—a gap that AI and flexible robotics are rapidly closing. The argument's greatest vulnerability lies in the assumption that software intelligence can fully overcome the physical friction and supply chain rigidity of the real world, but the trajectory is undeniable. Readers should watch for the emergence of these "cloud manufacturing" platforms, as they will likely be the first to break the decades-long trend of rising costs for custom goods and services.
"If your software is smart enough, and your equipment flexible enough, you can set up some new process to produce some new widget on the fly, automatically working out what the process steps need to be and how to route the material through the various machines."
The strongest part of this analysis is its ability to translate abstract tech trends into concrete economic outcomes, showing how the "AWS" model could finally solve the cost puzzle of car repairs, custom parts, and bespoke research. The biggest risk is underestimating the time and capital required to build the infrastructure, but if Potter's vision holds, the next decade could see the most significant shift in industrial economics since the assembly line.