Structural (castings + PEEK)

Aluminium structural castings are the clearest non-bottleneck in the whole bill of materials, they are fully elastic, and they never bind. A humanoid’s frame and housings are a modeled fifteen kilograms or so of die-cast and gravity-cast aluminium¹, so even a worldwide fleet of forty-seven million robots by 2040 needs only about seven hundred and five kilotonnes of castings in total⁷, a rounding error against a global aluminium-castings base already running between twenty-six and thirty-one million tonnes every year²³. Run the four-series gap and the volume crossover simply never arrives, in any region: cumulative global output stays roughly nine hundred times ahead of the fleet’s cumulative need across the whole horizon, China about four hundred times ahead, and even the smallest bloc, United States output alone, about one hundred and thirty times ahead. All three crossover years are legitimately null⁷. The robot slice is less than one percent of a single year’s casting output, transportation already consumes more than half of that base³, and electric-vehicle gigacasting is scaling the industry far faster than humanoids ever could pull it¹. The only things to watch sit one and two levels upstream and never bite at these tonnages: primary-aluminium smelting is energy-intensive and concentrated in China, and the specialty polymer PEEK that rides alongside the castings is a genuinely thin market⁶. Neither gates the fifteen kilograms of commodity aluminium. This is the link the report includes to prove the model is honest: when a part is genuinely abundant, the four-series gap shows no wedge at all.

The structural castings are the robot’s skeleton: the load-bearing frame members, joint housings and limb shells, made the way the auto industry makes them, high-pressure die casting and gravity casting of aluminium alloy, the same gigacasting techniques now pressing single-piece electric-vehicle underbodies on six-thousand-to-sixteen-thousand-tonne presses¹. We model about fifteen kilograms of aluminium castings per robot, the frame and housings, which is a modeled engineering estimate anchored to that gigacasting structural trend rather than a published per-robot figure¹. Because this is tracked in mass, not piece count, fifteen kilograms per robot multiplied across the worldwide fleet gives the total casting tonnage that fleet needs. Riding alongside the aluminium is a much smaller amount of the high-performance polymer PEEK, on the order of two kilograms per robot, which is the genuine specialty material to watch, but it is substitutable and not imminent, and it does not change the aluminium picture⁶.

Because a casting is built into a robot and stays there for its service life, structure does not wear out and there is no replacement tail¹, the mass the fleet needs is a cumulative stock: the total number of robots ever built worldwide multiplied by fifteen kilograms each, not a per-year flow. Integrating the consensus shipment ramp⁷, the worldwide installed fleet reaches about five hundred and eighty-five thousand robots by 2030, ten million by 2035 and forty-seven million by 2040, requiring roughly nine kilotonnes, one hundred and fifty kilotonnes and seven hundred and five kilotonnes of aluminium castings respectively. In the chart below the black line is that worldwide cumulative need; the grey band is cumulative global casting production, and the gap never opens, because the casting base dwarfs the robot slice by hundreds of times. That absence of a wedge is the finding, not an error.

Production is the only thing that differs between the Global, China and US views. The worldwide humanoid fleet and the casting mass it needs stay identical in all three; only the supply line changes, asking whether one bloc’s foundries alone could equip the whole global fleet. Global aluminium-castings output runs from about thirty-one thousand kilotonnes a year today, that is roughly thirty-one million tonnes², with more than half going to transportation³, grown a conservative three percent a year (a mature commodity carrying an electric-vehicle gigacasting tailwind) toward roughly forty-eight thousand kilotonnes by 2040. China is the largest aluminium-castings hub in the world⁴, Asia-Pacific holds about half of the market, with two-and-a-half to three thousand automotive foundries, several pressing over one hundred thousand tonnes a year each, and gigacasting already settled and scaled there¹, running near fourteen thousand kilotonnes a year and growing with the base toward about twenty-three thousand by 2040. United-States-located output is a mature domestic niche, about five thousand kilotonnes a year, American aluminium die-casting alone is roughly three-and-a-half million tonnes of an eighteen-million-tonne global die-casting total⁵, growing slowly toward about six thousand two hundred kilotonnes by 2040, because no robot-driven reshoring is needed. These post-2030 figures are a modeled mature-growth extrapolation of the casting base, not a reported forecast, but the verdict holds under any plausible growth path, because the robot slice is under one percent of the base.

It does not, and that is the point of including this link. The crossover, the year cumulative production drops below cumulative need⁷, never occurs, in any region. Cumulative global output reaches something like six hundred and twenty-seven thousand kilotonnes by 2040 against a fleet need of only about seven hundred and five, roughly nine hundred times the coverage². China alone covers the worldwide fleet about four hundred and twenty times over⁴, and even the smallest bloc, United States output by itself, covers it about one hundred and thirty times over⁵. All three crossover years are legitimately null, and the global gap stays closed across the whole horizon. This is the opposite of the precision-mechanical links: there the durable, never-replaced stock relentlessly outruns a rate-limited supply; here the supply base is so vast, so cheap to expand, a new die-casting cell builds in about a year, and the per-robot mass so small that the fleet’s entire lifetime demand is absorbed without a ripple. A humanoid’s fifteen kilograms of castings is, to the aluminium industry, statistical noise.

There is no binding step at the foundry. Aluminium structural castings sit on a vast, elastic auto and industrial light-metal base², the most supply-abundant material in the robot. If anything is worth watching it is one and two levels upstream, and neither bites at robot tonnages. First, the metal itself: primary-aluminium smelting is highly energy-intensive and roughly fifty-five to sixty percent concentrated in China, so the strategic exposure is to electricity and trade policy on raw ingot, not to casting capacity, and at seven hundred kilotonnes of finished castings by 2040, robots are far too small a draw to move that market. Second, the polymer that rides alongside the aluminium: PEEK is a genuinely thin specialty market, only about eight-and-a-half kilotonnes consumed worldwide a year, with Victrex holding more than half and plants taking two-and-a-half to three years to build⁶. At roughly two kilograms per robot, a fleet of ten million robots would call for about twenty kilotonnes of PEEK, which approaches the whole current supply, so PEEK is the real specialty watch-item beside the castings, though it is substitutable and not imminent, and it does not gate the fifteen kilograms of commodity aluminium. The honest verdict on this link is therefore the simplest in the report: aluminium structural castings are fully elastic, they never cross, and the gigacasting build-out the electric-vehicle industry is already paying for guarantees the foundries will be there. The four-series model shows no wedge here precisely because there is no shortage to show.

A cobot’s aluminum arm runs about 19 kg; a traditional industrial arm, far heavier, carries on the order of 280 kg of structural aluminum, so the ~6-million industrial-arm fleet adds the most aluminum tonnage of the embodiments by 2040. But aluminum is among the most elastic materials in the chain (a vast auto/aerospace/packaging base), so it never binds; the embodiments only add volume. Quadrupeds are lighter again, about eight kilograms of aluminum apiece, so even a quadruped fleet of about 5.15 million by 2040 adds only some forty thousand tonnes, still statistical noise against a base of tens of millions of tonnes a year. Drones are lighter still and mostly plastic and carbon fibre: only about a tenth of a kilogram of aluminium apiece, so even a cumulative drone fleet of roughly four hundred and seventy-eight million by 2040 (units built, since about half are single-use military first-person-view craft) adds only some forty-eight thousand tonnes, again noise against that same base.