The elements are not costly, according to Palmer Luckey. It is the change and the control that have made it very costly. In his account the next big jump in price does not resemble another production line. It resembles software capable of planning, scheduling, verifying, automating the messy middle of industry, mines, mills, factories and job sites, until the cost of converting raw materials into finished hardware goes down to one of the floors. The most vivid example provided by Luckey is the purchase of a pickup truck at the sort of money that is usually spent on phone upgrade, a provocation that was intended to provoke a question: is it because of the manufacturing being more or less machine-driven, that physical goods remain expensive?
The answer is one statistic that manufacturers are neurotically following: labor per unit. In the auto industry worldwide, labor consumes 65 to 70 percent of the total conversion costs, which is a high percentage that can easily change in hands of slight productivity improvement to propagate up the pricing and margins. A survey across over 250 assembly facilities revealed that most mainstream and high-volume auto makers were at an average labour cost of $880/vehicle, whereas Chinese auto makers were at an average of $585 – a difference due to newer plants, scale and highly tuned processes. When the near-term superpower of AI is eliminating minutes per task, fewer handoffs, fewer work-back loops, fewer schedule failures, it is going to the cost stack where it is most concentrated, not the stack of costs most visible to consumers.
But the concept of “near-zero” transformation costs by Luckey is bumped up against a hard physical law that software can not desire to annul: energy and the infrastructure that makes it available.
AI is taking on the role of an industrial burden, rather than an enabler. The data centers have become “factories of calculation”, and their demand on the upstream feeds on the same concrete, steel, copper, switch gear, and manual electrical labor that factories and housing demand. One AI campus can consume hundreds of megawatts at any given moment and project timing now depends more on substation and transmission schedules than on the deliveries of server shipments. Transformer lead times of 1836 months utility grade and queues of interconnection of up to 35 years make the power the constraining item to new capacity, despite the availability of capital.
It is important because the vision of affordability created by Luckey presupposes not only the automated factories, but also less expensive steel, less expensive energy, and less expensive recycling. However, the U.S. is facing an electricity construction problem: to sustain the projected demand in the AI era, it may need to add 75-100 GW of new generation by 2030. The realistic short term path to adding firm power suggests it is majorly towards natural gas, a route that continues to run into the sidewalls of permitting, grid integration, and local opposition, precisely the same route of institutional and regulatory bottlenecks Luckey says is swelling the sticker prices of today.
In the meantime, the shortage of labor is driving factories to another form of automation: human space-size robots instead of having to redesign plants. Humanoid systems have already been piloted in automotive processes and factories, mostly on low risk material movement processes. Their attractiveness is to the point: they are being made to fit the human world, Humanoid Global CEO Shahab Samimi said. They are capable of using tools, moving materials, and collaborating with people, they do not make factories be redesigned. The gateway to cost is still there; Samimi based current units on the price of about $50,000, and wider use depended on the fall to the price of between $5,000 and 10,000.
What is being contested is the point that is being put across by Luckey in his seasonal-car thought experiment: buy, use, then recycle; high efficiency. Whether AI can reduce the cost of the thinking work within manufacturing, or not, is not a question but whether energy systems, supply chains and compliance regimes can take a wave of new industrial throughput without turning reliability and power accessibility into the new luxury goods.
