How Formula 1 teams actually use 3D printing
The interesting parts are not on the Formula 1 car. They are the tools, cores, and models that let a team test more ideas per week than the team next door. Here is what gets printed, in which polymers, why those polymers, and which of the widely repeated numbers you should treat as marketing rather than data.
| Reference part | Hydraulic line bracket |
|---|---|
| Material | Carbon fiber reinforced polyamide |
| Process class | Material extrusion (MEX), per ISO/ASTM 52900 |
| Reported lead time | 2 weeks to ~4 hours |
Why would you print a hydraulic line bracket instead of machining it?
Start with the least glamorous component on a 2017 Formula 1 car: a small bracket whose only job is to restrain a hydraulic line. Machined conventionally, that part carried a lead time of roughly two weeks. Printed in a chopped carbon fiber reinforced polyamide, it came off the machine in about four hours and went straight onto a race car.[1]
The mechanism matters more than the ratio. Almost none of those two weeks is cutting time. It is queueing behind other jobs, CAM programming, work holding and fixture design, stock procurement, deburring, inspection, and shipping. Additive does not make the material removal faster because there is no material removal. It deletes the setup chain, and setup is where small one-off parts lose their schedule.
The tradeoff is anisotropy. A short fiber reinforced extruded part is strong along the bead and weak across the layer interface, so the governing design case is almost always interlayer tension or interlaminar shear, not the tensile number on the datasheet. That makes build orientation a controlled engineering parameter rather than an operator preference, and it needs to be recorded on the drawing the same way a grain direction would be.
| Designation | What it governs |
|---|---|
| ISO/ASTM 52900 | Additive manufacturing terminology. This is the document that makes FDM, SLA, and PolyJet into material extrusion, vat photopolymerization, and material jetting. Use the process class names in specifications, since the popular names are trademarks. |
| ISO/ASTM 52921 | Terminology for coordinate systems and test methodologies. This is how you state build orientation unambiguously on a print. |
| ISO 527 and ASTM D638 | Tensile properties of plastics. Any quoted strength for a printed polymer is meaningless without the method, the specimen geometry, and the build orientation attached to it. |
| ASTM D570 | Water absorption of plastics. Polyamide and polyetherimide are both hygroscopic, and moisture pickup changes both printability and mechanical response. |
Can a 3D printed tool survive an autoclave cure cycle?
Yes, within limits, and this is the highest value application in the whole workflow. A rear wing flap needs a layup tool. The flap itself is a pre-preg carbon laminate, so the tool is what defines the aerodynamic surface. One documented example ran roughly 900 mm across, printed in polyetherimide (the SABIC ULTEM 1010 grade) on a large frame industrial extrusion system, and the team evaluated three separate design iterations inside about a three day window. An equivalent aluminum tool is a multi week job with real capital attached.
The engineering question is not whether the polymer melts. It is whether the tool holds geometry while it is simultaneously hot and under pressure.[2] Three things set that limit:
- Deflection under load at temperature, which is a stiffness and creep problem, not a melting point problem.
- Glass transition margin. The tool needs headroom above the cure dwell temperature, not parity with it.
- Thermal expansion mismatch against the laminate, plus vacuum integrity. An as printed wall is not gas tight, so printed tooling generally needs sealing or a surface coat before it will hold a bag.
| Designation | What it governs |
|---|---|
| ASTM D648 and ISO 75-2 | Deflection temperature under flexural load. Always state the load case, because the value at 0.45 MPa and the value at 1.82 MPa are very different numbers for the same polymer, and only one of them is relevant to a loaded tool. |
| ASTM D7028 | Glass transition temperature of polymer matrix composites by dynamic mechanical analysis. This is the defensible way to state the real service ceiling of the tool. |
| ASTM E1640 | Assignment of the glass transition temperature by DMA for the neat polymer. |
| AMS 2750 | Pyrometry practice covering thermocouples, instrumentation, and temperature uniformity surveys. If you are claiming a cure temperature, this is the class of practice that makes the claim auditable. |
| ASTM D5687 | Guide for preparation of flat composite panels, useful as the reference frame for lay up and debulk practice. |
| UL 94 | Flammability classification. Relevant if the tooling polymer will also be considered for on vehicle or enclosure use. |
Print the tool, not the part. That is where the schedule actually collapses.
How do you make a hollow carbon fiber duct without a metal mandrel?
Brake cooling ducts are hollow carbon structures with internal geometry that no two piece steel tool can produce. The answer is a sacrificial core. You print a form, lay the pre-preg over it, cure the assembly, then remove the core through the open end. The printed object never sees a race lap. It exists to be destroyed.
Removal is the design driver. Water soluble cores are the cleanest route because you wash the core out rather than pulling it, which means you can produce undercuts and internal returns that would trap a rigid mandrel. Breakout cores and low melt cores are the alternatives. Whichever you choose, the core has to be stiff enough to resist debulk and autoclave pressure without collapsing, dimensionally stable at cure temperature, and then still be removable afterwards. A core that survives too well is scrap, and the part goes with it.
The thermal environment is what makes the duct necessary in the first place. Carbon brake discs reach extreme peak temperatures during heavy braking, and managing that heat is an aerodynamic job as much as a friction one.[3] The duct is shaping a gas flow path around that heat source, and its own thermal requirement is far lower than the disc peak.
This is the most underused capability in polymer additive manufacturing. It is not a way to make parts. It is a way to make geometry that was previously unmanufacturable, and the same approach shows up across aerospace ducting, hydroforming, and composite pressure structures.
| Designation | What it governs |
|---|---|
| ASTM E1933 | Practice for determining emissivity in situ using infrared imaging. This is what turns a thermal camera reading of a brake disc into a defensible temperature. |
| ASTM E2758 | Guide for selection, calibration, and use of infrared thermometers, covering the error sources that dominate high temperature non contact readings. |
| FIA Formula One Technical Regulations | Permitted brake duct geometry, cooling arrangements, and bodywork boundaries. Duct design is a legality exercise before it is a thermal one, and the allowable envelope is defined there rather than by convention. |
How fast can a team turn a driver complaint into a fitted part?
A new radio and data system went onto a 2017 Formula 1 car and the harness fouled the driver in the cockpit. The fix was a flexible rubber like boot to join the harness wires. Three design iterations went through in a single day, the final part printed in under two hours, and it was on the car for the opening race of the season.
The part itself is trivial. What is not trivial is that the loop closed without a supplier quote, a tooling budget approval, or a minimum order quantity. When the cost of trying a design approaches zero, you stop arguing about which of three concepts is best and simply build all three. That change in behavior is worth more than any individual part, and it is the same reason three tool iterations were evaluated for the wing flap instead of one.
| Designation | What it governs |
|---|---|
| ASTM D412 | Tensile properties of vulcanized rubber and thermoplastic elastomers, the right method for a flexible boot rather than the rigid plastics tensile method. |
| ASTM D624 | Tear strength. Tear, not tensile, is usually what kills a printed elastomeric part in service. |
| ASTM D2240 | Durometer hardness, the property most often quoted for these materials and the one most often quoted without a scale. |
Why put a printer in the garage rather than the factory?
Because at a race weekend the dominant term in lead time is freight, not manufacturing. Teams have been running printers at test sessions and circuits since at least 2017 to produce parts and tooling on demand rather than waiting on a shipment from the factory. It converts a logistics problem into a scheduling problem. The constraints are unglamorous: machine transport, power, calibration after transit, ambient humidity control, and feedstock drying discipline. And the output still has to satisfy scrutineering, so what can legally change between sessions is bounded by the parc fermé provisions of the FIA Formula One Sporting Regulations.
What does “9,000 printed parts a year” actually count?
By 2022 the reported output of a leading team’s additive lab was up to 9,000 parts per year, against a fleet now reported at more than 20 machines spanning vat photopolymerization, material extrusion, and material jetting. That number gets repeated as though 9,000 components were bolted to race cars. They were not.[4]
The large majority of that count is wind tunnel model hardware built by vat photopolymerization on large frame stereolithography systems. High temperature material extrusion in polyetherimide, PEEK, and carbon filled polyamide is the smaller share by count, and that is the stream producing on car parts and the tooling that makes on car parts.
The lab is running at production cadence rather than prototype cadence. Two different machine populations serve two different jobs. One builds geometry for measurement, at scale and at speed. The other builds functional polymer parts and tools that have to survive heat, load, and fluid contact.
Why do wind tunnel restrictions make print speed a competitive advantage?
This is the part of the story that turns a manufacturing capability into a championship variable, and it is entirely a consequence of the rulebook.
Aerodynamic development in Formula 1 is rationed. The allocation of wind tunnel runs and CFD resource is set on a sliding scale tied to championship position, so the team leading the constructors’ standings receives the smallest allocation and the team at the back receives the largest. Model scale and tunnel wind speed are capped as well.[5]
| Designation | What it governs |
|---|---|
| FIA F1 Sporting Regulations, Aerodynamic Testing Restrictions | Restricted wind tunnel testing and restricted CFD simulation, the accounting periods, and the sliding scale that ties allocation to championship position. Cite the article and the season, since the coefficients change. |
| FIA F1 Technical and Sporting Regulations, model scale provisions | Maximum model scale and maximum permitted wind speed. Quote the clause rather than the number in isolation. |
Now consider what that does to the value of a fast model shop. If the number of runs is fixed by regulation, the only free variable left is how much information you extract per run and how quickly you can build the next configuration. Model build lead time becomes the binding constraint on the entire aerodynamic program. Stereolithography earns its place here because a printed model can carry embedded pressure tap channels and sensor routing that would be painful to machine, and because a whole family of geometric variants can be produced in parallel rather than in series.
When runs are capped by rule, the fastest iterator wins the development war by default.
Did 3D printing win the championship?
The results are on the record. McLaren took the Constructors’ Championship in 2024, its first since 1998, and repeated in 2025 with Lando Norris taking the Drivers’ Championship alongside it. The constructors’ title is the engineering trophy, awarded on the performance of the Formula 1 car the team designed and developed.[6]
The defensible claim is narrower than the headline and still useful. Additive manufacturing shortened the interval between having an idea and measuring it, and over a season that interval compounds into more evaluated concepts than a slower rival can run.
What transfers from this to a normal engineering shop?
Every material in this story is commercially available outside motorsport: polyetherimide grades, PEEK, and carbon fiber reinforced polyamide all run on open high temperature extrusion platforms. The transferable insight is not the material list. It is the selection criterion.
- Target the parts where lead time is the constraint, not strength. If a part is holding up an assembly because it is queued at a vendor, that is the candidate. If it is holding up an assembly because it is highly loaded, it probably still wants metal.
- Go after tooling, fixtures, jigs, and cores before production parts. Tooling has looser certification exposure and delivers the same schedule saving.
- Qualify the process, not just the part. ISO/ASTM 52920 covers requirements for industrial additive manufacturing processes and ISO/ASTM 52930 covers installation and operation qualification. ISO/ASTM 52902 defines the geometric capability test artifacts you use to prove a machine is holding what you think it is holding.
- Control the two variables that cause most unexplained failures: moisture in hygroscopic feedstock, and undocumented build orientation.
- Compare against a real baseline. If the machined alternative would have been drawn to ISO 2768 general tolerances, state the printed part’s capability in the same language rather than in vendor accuracy claims.
- Budget post processing honestly. Support removal, sealing for vacuum integrity, and inspection are where the printed schedule quietly gives time back.
The compounding argument is the whole point. One part saved a week is a curiosity. A few thousand parts a year, each saving days, changes how many ideas an engineering organization can afford to test. That is the actual competitive mechanism, and it does not require a race car to work.
Precision notes
Figures in this article that are directionally correct but numerically loose, undefined, or unverified, with what to check before you repeat them.
01
Two weeks versus four hours is not a like for like comparison
The four hour figure is build time for one part. It excludes support removal, drying, thermal conditioning, dimensional verification, and the CAD work upstream. The two week figure is a full external procurement cycle. Both are plausible and the direction is right, but treat the ratio as a schedule comparison rather than a process time comparison.
Return to text
02
Autoclave cure temperatures are quoted loosely
The figure usually cited for this class of cure is somewhere above 175 degrees Celsius. Real pre-preg cure schedules are a ramp, dwell, and cool profile specific to the resin system, and both dwell temperature and pressure vary widely between systems. Specify the tool against the actual cure schedule and against a stated number of cycles, because printed tools drift dimensionally with repeated thermal exposure long before they visibly fail.
Return to text
03
The 1,000 degree brake disc figure describes a transient surface peak
Peak carbon disc temperatures near 1,000 degrees Celsius get repeated constantly. It is directionally correct for a heavy braking event, but it describes a transient surface peak on the disc rather than a steady state, and not the temperature the duct sees. Non contact measurement of a glowing carbon surface is strongly emissivity dependent, so the number moves with the instrument and the assumed emissivity.
Return to text
04
The 9,000 parts figure is team and supplier reported
“Parts” is undefined in the published figure. Whether a multi piece wind tunnel model assembly counts as one part or forty changes the headline by an order of magnitude. The figure is also several seasons old and was published in a supplier context, so treat it as an order of magnitude indicator of capacity rather than a measured production statistic.
Return to text
05
The 60 percent model scale limit is season specific
Sixty percent is the commonly cited maximum wind tunnel model scale, but that limit and the associated wind speed cap have been revised across regulation cycles, as have the sliding scale coefficients governing testing allocation. Verify against the current edition of the FIA Formula One Sporting and Technical Regulations rather than repeating the number.
Return to text
06
Championship results are correlation, not attribution
No manufacturing process wins a championship. Power unit competitiveness, tyre operating window, driver performance, pit execution, aerodynamic concept, and the timing of the regulation cycle all move the result far more than any single production capability.
Return to text