3D printed temple: how India completed the world's first one

• 1
  Key takeaways
• 2
  What the 3D printed temple actually is
• 3
  How concrete extrusion printing works, layer by layer
• 4
  Where the process stops working well
• 5
  What this teaches about 3D printing for metal parts
• 6
  When to print, when to machine
• 7
  How to inspect a layer-built part
• 8
  Concrete printing vs. metal printing vs. CNC machining

The 3D printed temple sits in Siddipet, Telangana. Construction company Apsuja Infratech worked with 3D printing firm Simpliforge Creations, and the structure was reported complete on 6 June 2023. It is not one monolithic block. The design splits into three separate shrines, each printed as its own shell and finished on site.

That split matters more than the headline. A printer with a fixed gantry has a working envelope, so the designers broke a large religious complex into pieces that fit inside it. The same decision shows up in metal work every week: when a part exceeds the machine travel, you either redesign it as an assembly or move to a different process.

The wall itself is a hollow shell. A nozzle extrudes a cementitious mix in beads roughly 20–40 mm wide, and the printer stacks them. There is no formwork, no vibration, and no rebar cage placed before the pour. Reinforcement, where present, is added between layers or into a cavity the printer leaves open.

So the honest description is this: a large concrete shell, printed in segments, assembled, then detailed by hand. The 3D printed temple is a construction milestone, but it is not a tolerance milestone. Nothing about it is held to ±0.1 mm, and it does not need to be.

• 1
  Three shrines, printed separatelyEach module fits the printer envelope, then gets joined on site.
• 2
  Hollow wallsMaterial goes only where the shell needs it, which cuts weight and cure time.
• 3
  Hand finishingCornices, openings and surfaces are cleaned up after printing.

The mix leaves the nozzle as a soft paste with a yield stress high enough to hold its shape once it lands. That is the whole trick. Ordinary concrete slumps; printable concrete has to stand as a vertical bead the moment it is deposited, then stiffen fast enough that the bead above does not squash it.

Print speed and bead geometry are locked together. If the head moves too fast, the bead thins and the wall loses section. Too slow, and the bead piles up thicker than the model calls for, so the next pass has less room. Operators tune flow rate against traverse speed until the bead cross-section matches the toolpath within a few millimeters.

The vertical rhythm is called the layer time. Every point on the wall has to wait long enough for the layer below to reach a green strength that can carry the next pass. On a big footprint, that means the head may need to travel a long path per layer, or the print pauses. Pausing creates a cold joint, and cold joints are where the structure is weakest.

Curing continues long after the machine stops. Printed concrete loses moisture fast because so much surface is exposed, so most builds get covered or misted. Shrinkage during that window is what drives the curling and cracking you see at the top of tall printed walls.

• 1
  Yield stressThe mix must hold a vertical edge with no support.
• 2
  Flow vs. speedBead width is set by the ratio, not by either value alone.
• 3
  Green strengthThe layer below has to carry the layer above, still wet.

Overhangs are the first wall you hit. A concrete bead cannot bridge air, so anything past roughly 30–45° from vertical needs either a support structure, a printed corbel, or a redesign. The temple's domes and arches work because they curve inward gradually; a sharp cantilever would not.

Openings are the second. A door or window means the bead path has to end and restart, and every restart is a seam. In practice you either print a lintel in place, insert a precast element, or leave the opening oversized and cut it later.

Flatness is the third. Layer stacking leaves visible ridges and small undulations across the face. A printed wall is not a machined face. If a fixture or a bracket has to bolt flush to it, you plan for grouting, shimming, or a cast-in plate placed during the print.

None of these are failures. They are the boundary of the method. The useful question is always whether your part's critical features fall inside that boundary or outside it.

The physics rhyme. Metal powder-bed printing also stacks thin layers, typically 20–60 μm, and those layers also need the previous one to hold them. The difference is that a laser or electron beam fuses each layer fully, so the bond is metallurgical rather than adhesive. That is why metal printing holds ±0.05–0.1 mm on small features while concrete holds centimeters.

Support is the shared headache. A metal part with a 60° overhang prints onto sacrificial support that gets cut away and refinished by hand. Add up that labor and the cost curve turns. For a one-off bracket with a handful of overhangs, printing wins. For a thousand identical brackets, a 5-axis machining center with a Ø400 mm rotary table usually wins on both unit cost and repeatability.

Scale changes the answer again. GreatLight runs 127 high-precision CNC machines, including 16 simultaneous 5-axis centers and a maximum processing size of 4,000 mm. When a printed metal part would need a build chamber larger than that, or when the geometry is mostly prismatic, subtractive is the cheaper path.

The rule of thumb we give engineers: print the geometry you cannot cut, machine the geometry you can. Holes, flats, slots, threads and sealing faces belong on a spindle. Internal channels, lattice, and organic ribs belong in a printer. Mixed parts often get both, with printed blanks finished on a mill.

• 1
  Layer bond typeConcrete relies on adhesion; metal printing fuses the layer.
• 2
  Support removalEvery overhang becomes manual labor on the back end.
• 3
  Hybrid routePrint the near-net shape, then machine the critical faces.

Start with feature count, not part count. If a part has more than about ten critical faces that must locate against something else, printing the whole thing and then chasing those faces is slow. You print a blank with stock and let a 3-axis or 5-axis machine bring the faces in.

Then look at material. The printable metals most shops run are titanium alloys, stainless steels and aluminum-silicon powders. If your part is 6061-T6, 7075, 17-4PH or C360 brass, you can buy certified bar stock and cut it today. There is no reason to wait weeks for powder and a build slot.

Then look at the tolerance callout. Anything at ±0.005 mm, or a surface at Ra 0.2–0.8 μm, is machining territory. Printed surfaces on metal land near Ra 8–12 μm as-built, so every functional face needs a finishing pass anyway.

Finally, look at the schedule. A machined prototype from bar stock can start within 24 hours and ship in 3–5 days. A printed metal part involves a build queue, stress relief, support removal and then the same finishing operations. Printing buys geometry, not speed, unless the geometry is impossible any other way.

Layered parts fail at the seams, so inspection targets the seams. For printed concrete, that means coring samples and checking bond strength between layers, plus a moisture and shrinkage reading during the first days of cure. A wall that looks fine on day one can crack on day ten.

For printed metal, the checks are density, porosity and dimensional drift. A CT scan catches internal voids that a surface inspection misses entirely. Coupons printed alongside the part give you tensile data for the same build conditions.

For machined parts, inspection is more direct. At GreatLight every shipment goes through a raw material check, in-process monitoring and a final inspection, with reports on request. The qualification rate we hold is 99.99%, and that number is only meaningful because the checks happen at three points, not one.

Whichever route you take, write the inspection plan before the first part is made. Deciding what to measure after the fact usually means measuring the wrong thing very carefully.