3D Printed Shoe Startup Hilos: What the $3 Million Round Proves
Hilos raised $3 million to push additive manufacturing into footwear. This page explains the mechanism behind that model: what MJF does well, where the material ceiling sits, and when a machined mold or a CNC-cut tool still makes more sense.
Why a 3D Printed Shoe Startup Attracted $3 Million
In March 2023 a 3D printed shoe startup called Hilos raised $3 million. The company is based in Portland, Oregon and builds footwear on HP Multi Jet Fusion (MJF) equipment instead of traditional lasts and presses. That single equipment choice explains most of the funding story.
MJF is a powder bed fusion process. A recoating blade spreads a thin layer of nylon powder, typically 80 μm thick, then a fusing agent is jetted where the part should be solid and a detailing agent is jetted at the edges. Infrared lamps sweep the bed and the treated powder melts. Layer by layer, the part grows inside a cake of loose powder.
The economic point is what the process does not need. No steel mold, no aluminum tool, no mold release agent. A design change is a file change. For a footwear company that wants to sell colorways, widths, and stiffness variants, that is a large structural advantage, and it is why investors looked at Hilos as a platform rather than a single shoe.
The catch is that MJF parts come out of the build chamber as semi-finished goods. They are grey, slightly porous, and dimensionally soft at the edges. Post-processing, dyeing, sealing, and sometimes CNC trimming decide whether the part is a product or a prototype.
How MJF Lattice Midsoles Carry Load
A running shoe midsole has to do two opposite jobs. It has to compress under a heel strike and recover, and it has to stay laterally stiff so the foot does not roll. Foam gets this from chemistry. A printed lattice gets it from geometry.
The common structure is a strut lattice, a pattern of thin diagonal members that intersect at nodes. Under compression the struts bend rather than buckle. Change the strut diameter and you change stiffness. Change the cell angle and you change how the part shears. Designers tune these two numbers along the length of the sole.
This is why the 3D printed shoe startup model scales in a way foam does not. A single MJF build can hold several stiffness zones because geometry, not material, sets the spring rate. Foam blocks have to be glued or molded in segments to reach the same result.
The limit is fatigue. Nylon 12 lattice struts accumulate damage at the nodes where stress concentrates. A solid foam midsole can survive millions of cycles; a thin lattice can crack at a node well before that. Designers compensate by thickening the struts near the heel and accepting a heavier part.
Where Additive Stops and Subtractive Starts
Additive wins when the part is geometrically complex, low in volume, or still changing. It loses when the part is simple, high in volume, or needs a surface finish that only a cutting tool can produce. That split holds for footwear and for industrial hardware alike.
Take a shoe last. It is a smooth, mostly convex form with no internal channels. Printing one wastes machine time on a shape a 5-axis mill can cut from aluminum in a fraction of the cost, and the aluminum last survives thousands of cycles. If your product is 10,000 pairs a month, the mold or the last is the right answer.
Now take a lattice midsole with 40,000 struts and three stiffness zones. No mold can produce that geometry in one shot, and no cutting tool can reach the internal nodes. Additive is the only route. The same logic applies to heat sinks with internal fins, robot grippers with compliant hinges, and brackets with organic load paths.
There is a third case that gets missed. Hybrid parts. A printed lattice core bonded into a machined aluminum frame uses each process where it is strongest. We build these regularly for robotics and medical fixtures, and they usually beat a single-process design on both weight and stiffness.
Material Ceiling of Printed Footwear
MJF runs on polyamide powders, mostly PA12 and PA11, sometimes with glass or carbon fill. PA12 has a tensile strength around 48 MPa and a flexural modulus near 1,700 MPa. Those numbers are fine for a midsole. They are not fine for a cleat plate that has to resist 2,000 N of bending.
PA11 is the better choice when toughness matters. It comes from castor oil, has slightly lower stiffness, and absorbs impact better than PA12. For a midsole that flexes a million times, that difference shows up in testing.
When a part needs metal-level stiffness, additive in nylon is the wrong tool. Options are CNC-machined 7075 aluminum, titanium Ti-6Al-4V, or a carbon fiber layup. We machine all three. A 7075 plate at 4 mm thickness is stiffer than a 40 mm nylon lattice and weighs less.
There is also the surface question. MJF parts sit around Ra 8–12 μm as built. That is rough enough to trap dirt and to feel unfinished in the hand. Bead blasting brings it to a matte finish; dyeing adds color; sealing closes the porosity. None of these reach the Ra 0.8–1.6 μm that a machined and anodized aluminum face delivers.
What the Hilos Round Says About Tooling-Free Runs
The $3 million did not buy machines. It bought the ability to iterate without committing to steel. That is the real lesson for anyone building a physical product, whether it is a shoe, a robot, or a medical device.
A traditional footwear program spends weeks and tens of thousands of dollars on a mold set before the first sellable pair exists. If the design changes after that, the mold is scrap. MJF removes that commitment. You can print 20 pairs, test them with runners, change the lattice, and print again the next week.
The trade is unit cost. A printed midsole costs more per piece than a molded one at volume, and the gap grows as volume rises. The break-even sits somewhere between a few hundred and a few thousand units, depending on part size and geometry. Below it, printing wins. Above it, tooling wins.
This is the same decision we walk through with clients every week. A prototype bracket gets printed or machined from stock. At 5,000 units, the same bracket moves to die casting or injection molding. The part does not change. The process does.
Printed Lattice vs Molded Foam vs CNC Metal
Use this to pick a process before you commit to a design.
| Factor | MJF lattice | Molded foam | CNC metal part |
|---|---|---|---|
| Tooling cost | None | High, weeks of lead time | None for cutting |
| Best volume band | 1 to a few thousand | 10,000 and up | 1 to 10,000+ |
| Geometry freedom | Internal lattice, hollow channels | Simple, draft-angled shapes | External forms, no internal voids |
| Stiffness range | Tuned by strut diameter | Set by foam density | Highest, alloy dependent |
| Surface finish | Ra 8–12 μm as built | Mold texture | Ra 0.8–1.6 μm typical |
| Fatigue life | Node-limited | Millions of cycles | Very high with proper alloy |
| Design change cost | One file edit | New mold | New toolpath |
When to print, when to machine, when to mold
If the part is geometrically complex and the volume is under a few thousand, print it. If the volume is above 10,000 and the shape is simple, cut a mold. If the part has to carry real load or hold a tight tolerance, machine it from aluminum or titanium and skip the debate.
Questions engineers ask after reading this
Can I print a lattice part and then machine it?
Yes, and it is often the right call. Print the lattice to near-net shape, then machine the mating faces, bores, and any sealing surfaces on a 3-axis or 5-axis mill.
The printed body gives you the internal geometry. The machined faces give you the tolerance. Expect the machined surfaces to hold ±0.005 mm while the printed surfaces stay at the process tolerance of the printer.
What tolerance should I expect from MJF?
MJF holds roughly ±0.3 mm on a 100 mm part, and the tolerance drifts with part orientation in the build chamber. Long thin features warp more than compact blocks.
If your design needs ±0.005 mm, that feature has to be machined after printing. We plan the print so there is stock on the machined faces.
Is nylon 12 stiff enough for a structural bracket?
For light loads, yes. PA12 sits near 1,700 MPa flexural modulus, which is roughly one twenty-fifth of 6061 aluminum.
If the bracket carries more than a few hundred newtons, move to machined 6061-T6 or 7075. The weight penalty is small and the stiffness gain is large.
How do I choose between SLS and MJF?
MJF gives better surface detail and slightly better mechanical consistency because of the detailing agent. SLS is more widely available and often cheaper for one-off parts.
For a production run of functional parts, MJF is usually the better pick. For a single visual model, either works.
Can you match a printed part to a machined housing?
Yes. We machine the housing first, measure the actual mating features, then adjust the print file to match. That avoids stacking two sets of tolerances.
Send both parts in one RFQ and we will sequence the operations so the fit is checked before shipment.
What about post-processing on printed parts?
Dyeing, bead blasting, tumbling, and sealing are all standard. Laser marking works on printed nylon if the character height is at least 1.5 mm.
If the part needs a conductive or hard-wearing surface, plating on nylon is difficult. In that case, machine the part from aluminum and anodize it.
Send us the geometry and we will tell you which process fits
Upload a STEP file and get a quotation plus a free DFM analysis within 12 hours. No minimum order quantity, from one prototype to 10,000+ part runs.
12-hour quote100% inspectionNDA on request
More process notes from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
