This pairing is unusual on a materials site because it is rarely a real choice. Nobody stands in front of a bracket wondering whether to print it in rubber. What people are actually asking is one of two narrower questions: can my printer even run flexible filament, and what does it cost me to make this part flex.
Both have concrete answers.
Your extruder decides whether you have the choice
TPU is pushed, not pulled. A rigid filament transmits the extruder's push down the tube like a rod; a soft one behaves like a spring, compressing and buckling anywhere the path is not tightly constrained. The stored weakness list for TPU 95A names bowden extruders first for exactly this reason.
- Direct drive with a constrained path. TPU 95A prints without drama at modest speeds. Softer grades need care but are possible.
- Direct drive with a gap between the gears and the heatbreak. Filament escapes into the gap and jams. This is fixable with a printed guide or a different extruder, and it is the single most common reason people conclude "my printer cannot do TPU".
- Bowden. TPU 95A is achievable with patience, very low speeds and retraction turned nearly off. Anything softer is not worth attempting.
Retraction is the specific enemy: every retraction stretches the filament in the path, and the extruder has to take up that slack before extrusion resumes. This is why flexible prints ooze and string on machines that are immaculate with PLA, and why the usual advice is to reduce retraction distance rather than increase it.
What flexibility costs in hours
Take a 45 mm machine foot: 42 cm³ of plastic, 0.2 mm layers, 25% infill, three walls, a 0.4 mm nozzle. Printed at 60 mm/s — an ordinary speed for a rigid filament — the estimator gives 3.94 hours across 225 layers.
Drop the speed to 25 mm/s, which is a realistic working ceiling for TPU on a mainstream direct-drive machine, and the same geometry takes 9.369 hours. Nothing about the part changed. Nearly five and a half hours of extra machine time for the same 42 cm³ of plastic is the real price of flexibility, and it lands on the machine-time line rather than the filament line. The print time estimator holds the TPU run; raising the speed back towards the rigid figure shows how much of that gap a better extruder recovers and how much it does not.
Note that flow is not the limit here — the requested flow at 25 mm/s is far below the hotend's ceiling. The speed limit on flexible filament is mechanical, in the extruder, not thermal in the melt zone. That is why a high-flow hotend does not help and a stiffer, shorter filament path does.
What that does to the cost
Feed the two print times through the same cost model. Both runs bill the printer by the hour at 50 cents and take power at the 17.5-cent United States rate. The PLA version — 52 g at $20/kg, 3.94 hours at 100 W, with 5% of attempts written off — comes to $3.24, of which only $1.04 is plastic.
The TPU version, 50 g at $35/kg over 9.369 hours with a 12% failure allowance: $7.50, of which $1.75 is plastic and $4.68 is machine time. The filament premium is barely a factor. Over two-thirds of the difference is the hours, and hours are what you are really buying when you choose a flexible material.
Getting flex out of PLA instead
Sometimes the honest answer is to keep the rigid material and make the geometry compliant.
- Spring geometry. A printed leaf spring or a compliant flexure in PLA deflects usefully and returns. This is how printed clips, catches and battery contacts are normally done.
- Thin sections. Anything under about a millimetre thick bends noticeably in PLA. Whether it survives being bent repeatedly is a different question.
- Vase-mode walls. A single-wall printed cylinder is surprisingly compliant radially while staying stiff axially.
The failure mode to respect is fatigue. PLA tolerates being bent; it does not tolerate being bent ten thousand times. If the part is a one-time snap fit, PLA is fine. If it is a hinge that opens daily, it is not, and polypropylene rather than TPU is often the better answer — see PP against PETG for what that costs elsewhere.
Getting rigidity out of TPU instead
The reverse trick is underused. TPU's effective stiffness in a printed part is set as much by geometry as by the polymer: a solid TPU block is firm, a gyroid-infilled one at low density is squashy, and the same spool covers both. If a part needs to be soft in one region and firm in another, infill density does that in a single print without a second material.
TPU 98A exists for the same reason. It is barely rubbery in the hand, but it survives impacts and cold that snap PLA, and it prints faster and more predictably than the softer grades.
The parts that genuinely are a choice
A short list, and it is short on purpose: printer feet and vibration dampers, phone and tool cases, wheels and tyres, gaskets and seals, cable strain reliefs, and grips on printed handles.
For everything on that list the question is not which material is better. It is whether the part's job is to hold a shape or to absorb something — a load, a shock, a leak. Hold: PLA, in a few hours. Absorb: TPU, and clear your afternoon.