A single object with no seams is the cheapest structure to defend in a review. It has no joints to argue about, no fasteners to fatigue, no assembly tolerance to stack up across seventy interfaces. That is the quiet objective function beneath the loudest machine in car manufacturing. When Tesla began pressing the rear underbody of the Model Y as one aluminium casting in late 2020, the headline was the machine — a 410-to-430-tonne Idra press exerting 55,000 to 61,000 kilonewtons of clamping force, as the Giga Press entry on Wikipedia documents in Idra’s own OL 5500 / 6200 datasheets. The concept was older, and more interesting than the tonnage: replace an assembly of stamped-and-welded parts with a single solved geometry.

This is not really a story about cars. It is a story about part consolidation — the design decision to let one continuous volume carry a load that used to be shared across many bracketed, bolted, spot-welded members. Every joint you delete is a stress concentration you never have to analyse, a jig you never have to build, and a tolerance you never have to hold. The Giga Press is simply the most expensive way anyone has yet found to say that out loud.

←TODAY: A 6,000-tonne press turns ~70 welded parts into one 80-kg aluminium casting in ~85 seconds. →3012: The joint becomes a design smell — structures are grown as single load paths, not assembled from a bill of materials. Fulcrum: Consolidation only pays when the geometry is derived from the load, not merely shrunk to fit a bigger machine.

What it is: Single-piece casting — the automotive press calls it megacasting, the structural engineer calls it part consolidation — is high-pressure die casting (HPDC) scaled to the size of a vehicle floor. Molten aluminium alloy is forced into a reusable steel die under intense pressure, held while it solidifies, then ejected as a near-net-shape part. What changes at Giga Press scale is not the mechanism but the ambition: instead of casting a bracket, you cast the whole subframe as one topology, with the ribs, bosses and crash rails already in the geometry.

Why it works: Three physics facts do the work. First, joints are where structures fail — a weld or a fastener is a discontinuity in the load path, and a discontinuity is a stress concentration. Remove the joint and the load flows through continuous material instead of arcing around a seam. Second, the alloy has to be castable and structural at once: Sandy Munro’s 2021 teardown reported the large Tesla chassis parts as roughly 89.5% aluminium and 8.5% silicon — the silicon is what lets the melt flow into thin ribs before it freezes, placing it near Aluminum Association alloy AA386. Third, the machine size is not vanity, it is arithmetic. Clamping force must exceed the cavity pressure times the projected area of the part, or the die blows open on the shot. A vehicle-floor footprint at HPDC pressures lands you at tens of thousands of kilonewtons — hence 55,000 kN and up. The quality then lives in the details most people skip: a low vacuum of ~50 millibar pulled inside the closed die to stop trapped-gas porosity, held by a 4,000-litre tank made by Fondarex in Switzerland, and argon rotary-degassing to strip hydrogen from the melt. Porosity is the enemy because a void is a crack that has not started yet.

The deeper lineage is topology. The question “where should material be, and where should it be absent, to carry this load with least mass?” is exactly the problem A.G.M. Michell posed in 1904 and that modern topology optimisation — SIMP, BESO — solves numerically. A megacasting is only worth its die cost if the geometry inside it is a genuine load path, not a shrink-wrapped copy of the old welded assembly. Cast the answer to the wrong objective function at 6,000 tonnes and you have simply industrialised a bad idea.

Origins: Die casting itself is 19th-century; the leap here is scale. Idra Group in Italy listed base-specification Giga Press machines in its 2018 catalogue — general manager Riccardo Ferrario noted in 2018 that Idra had built a 4,200-tonne press before competitors and had 5,500- and 6,500-tonne models on the books. In 2019 Tesla’s Jerome Guillen described a “giant, giant, giant machine” to “make full-size cars in the same way that toy cars are made”; Elon Musk confirmed the purchase in April 2020, and Fremont’s DCM1 ran trial Model Y castings by September 2020. The idea then spread past Tesla: Toyota announced large-casting adoption for EVs in June 2023, Japanese supplier Ryobi projected a 20% body-cost reduction, and by November 2023 two 9,000-tonne presses were ordered for a Volvo EV plant at Košice, Slovakia. The frontier detail worth watching is the mould: a metal die can cost around $1.5 million and takes many iterations, so Tesla turned to binder-jet 3D printing in industrial sand to prototype casts — and to form hollow subframes by removing sand cores after casting.

In practice: For a Swiss studio, the transferable idea is not the press — it is the discipline. Precast and prefabricated concrete has always chased the same objective: fewer site joints, more factory-controlled continuity, the load path resolved before the crane arrives. The binder-jet sand mould is the bridge that makes it real for architecture, and it sits squarely in the digital-fabrication tradition ETH Zurich’s DFAB work has pushed for a decade — 3D-printed formwork lets you cast a genuinely optimised, doubly-curved element instead of the rectangular one the plywood mould could afford. There is a real trade-off, stated plainly: single-piece casting swaps repairability and mould flexibility for joint elimination, and a minor collision that once bent a bracket now condemns a subframe. Tesla itself paused single-piece casting in May 2024 and reverted to three-piece on its current lineup — the system met its economic limit, and that is data, not defeat. The parallel to the coordinated single model in our Archicad — En Obra concept panel is exact: consolidation is powerful precisely because it removes seams, and dangerous for the same reason. Monday move: take one detail in a live project that carries too many parts — a steel connection with a dozen plates, a repeated façade bracket — model the consolidated single-element version, and cost the one printed mould against the joints, welds and inspections it deletes. Do it once and the instinct sticks.

Hack: Size the press from the part, not from the brochure. Clamping force is not a spec you admire — it is projected area times cavity pressure, and it tells you why a floor-sized casting needs a building-sized machine. Render it in four lines and the tonnage stops being mythology:

proj_area_m2 = 1.6          # rear underbody footprint, m^2
cavity_pressure_MPa = 90    # HPDC intensification, ~70-100 MPa
force_MN = proj_area_m2 * cavity_pressure_MPa   # MPa*m^2 == MN
print(round(force_MN), "MN ->", round(force_MN * 102), "tonne-force")

Run it and you get roughly 144 MN, about 14,700 tonne-force — change the footprint or the pressure and the machine resizes in front of you. That is the whole economic argument of megacasting in one equation: the part sets the press.

The advice that outlives any press is the one my generation learned late. A generated or cast geometry with no recorded objective function is a beautiful guess you cannot defend in a structural review. Before you consolidate anything into one continuous element — in aluminium or in concrete — write down which load path the form is a solution to, and keep that derivation somewhere the mould file cannot take with it when it goes dark. Cast the reasoning, not just the ribs.

Source: en.wikipedia.org