The short answer
Vacuum casting — also called urethane casting — is usually **better than CNC** when you need a short run of **plastic-like parts that should look and feel molded**, and you are not ready to cut injection-mold tooling. The usual stack is: finish one master (often SLA or CNC), pull a silicone mold from that master, then pour polyurethane copies under vacuum so trapped air stays down. That path shines for appearance models, sales samples, and pilot or bridge batches on the order of **several to a few dozen** matching parts.
Choose **CNC machining** instead when the prototype must behave like a machined part: a real engineering plastic or a metal, tight fits, sealing lands held to the print, cut threads that must gauge, or a production path that will stay CNC anyway. 3D printing sits between the two for one-off geometry checks — see SLA vs SLS vs MJF and CNC vs 3D printing for that fork.
This article is typical buyer guidance — no invented scrap rates, tool-life guarantees, or price tables. If you already know you want cast copies, read the vacuum casting service page and still upload STEP plus a readable PDF on quote-new.php. If the job may be either cast or machined, write the decision rule (cosmetics versus fits) so the preliminary min–mid–max band matches the process you intend to buy.
What vacuum casting is — and what it is not
For buyers, vacuum casting means four steps:
1. **A master pattern** that already carries the cosmetic surfaces you care about — texture, fillet language, logo deboss, gloss versus matte. Masters are commonly high-resolution 3D prints or CNC-machined plastics.
2. **A silicone tool** pulled from that master. Soft tooling picks up fine texture and can tolerate some undercuts that would be painful in a rigid steel injection cavity, at the cost of limited tool life.
3. **Urethane (polyurethane) copies** poured in the closed tool, usually under vacuum to reduce bubbles. Rigid grades often stand in for ABS-like or PP-like covers; elastomeric grades stand in for overmolds, boots, and gaskets.
4. **Optional secondary work**: trimming gates and flash, installing inserts, painting or tinting, and machining a bore or sealing face that the cast skin cannot hold.
It is **not** metal die casting, and it is **not** injection molding: you do not get a hardened steel cavity or the full mechanical property card of a production resin. Think of it as **bridge production** between printed or machined singles and a future molded tool. This piece is only about **when the silicone-tool step beats machining every copy**.
Silicone tools typically support on the order of **a dozen to a few dozen** shots before tear-out, dulling, or flash grow unacceptable. Treat tool life as a planning band you confirm on the RFQ, not as a catalog rating.
When vacuum casting wins
Vacuum casting is the better prototype path when most of these are true:
- You need **matching copies**, not one hero print — photo units, field samples, a trade-show set, or a small pilot build.
- The parts should read as **molded**: relatively uniform walls, rib-and-boss language, and a consistent texture, color, or transparency that layer lines or cutter witness will not sell.
- Quantity sits in the common **short-run** band: more than a couple of cosmetic units, fewer than the volume that justifies injection-mold tooling.
- Material intent is **urethane as a stand-in** — rigid, clear, tinted, or rubber-like — rather than a named machinable grade such as POM or PEEK.
- You want to test **assembly and handling** of a future molded design before paying for steel.
- Some **undercuts** are easier to pull from a flexible tool than to mill, provided you accept cast — not machined — accuracy.
When CNC machining is the better prototype
Stay with CNC — or machine the copies after a simpler print — when:
- **Fits and datums matter.** Bearing bores, dowel patterns, sealing glands, and ISO-fit shafts belong on a machined print. Cast urethane can be close; it is not a substitute for a reamed or bored size held to the drawing.
- You need a **real engineering plastic or a metal**. POM vs PEEK for CNC functional prototypes and choosing engineering plastics (PEEK, POM, nylon) are questions vacuum casting does not answer.
- Quantity is **one or two** and the geometry is prismatic — a plate, a manifold face, a simple housing. Cutting those directly is usually clearer than funding a master and a tool.
- **Threads must assemble as cut** (see designing threads and tapped holes). You can cast in inserts, or machine and tap after cast, but “the print says 2B and we poured it” is how quote bands widen.
- Production will **remain CNC**. Prototype vs production CNC quotes reminds buyers not to validate a molded process if the released part will be machined from billet.
- Heat, chemicals, or wear make a look-alike urethane the wrong thing to learn from.
If you only needed CNC, send a machining RFQ: name the real alloy or plastic, the cosmetic faces, and the tolerances that actually matter.
Where 3D printing sits between the two
Use printing when you need **one or a few geometry or fit checks quickly**, especially with organic surfaces, internal channels, or lattices that neither a mill nor a silicone tool wants to copy. SLA vs SLS vs MJF walks through surface and toughness trade-offs: SLS/MJF nylon-like parts are often better functional singles than a first urethane shot, while SLA is a frequent **master** for a later silicone tool because it can carry fine cosmetic detail. The common hybrid is **print or machine a master, cast the copies, optionally machine critical faces**.
Design notes that change the cast-versus-CNC call
Write these on the PDF, not only in chat:
- **Walls and ribs.** Vacuum-cast parts prefer relatively uniform walls, like molded covers. Paper-thin skins and very thick unsupported chunks both cause trouble (fill, sink, or heat in the resin).
- **Draft and parting.** A small draft (often around **1°** as a starting conversation, not a universal rule) still helps a silicone tool release cleanly and live longer.
- **Texture and gloss.** The master surface largely transfers. If you need a molded texture rather than as-machined witness, say so.
- **Color and clarity.** Call a target (a RAL or Pantone reference, or plain language such as “neutral opaque,” “smoke tint,” or “water-clear”) as intent, not as a laboratory color certificate unless you actually require one.
- **Inserts and after-cast machining.** List brass or helical inserts and any bores that must be machined after cast.
- **Undercuts.** Possible, not free. Label them so the tool split is planned, not guessed.
Typical size control on cast copies is **looser than CNC** of the same nominal. Do not paste an H7/g6 table onto a urethane housing and hope. If a feature must hold a machined tolerance, call it out as a **secondary machine operation** after cast, or make that component CNC and only cast the cosmetic skin.
How the choice shows up on an RFQ
You feel missing process notes as wider min–mid–max bands and more manual review when the files say “ABS housing, 25 pieces, tight bores.” Reading a preliminary online quote (min / mid / max) is easier when the PDF states one of:
1. **Vacuum cast** these copies (master already exists / please make a master). Include color, texture, rigid versus rubber-like intent, insert list, cosmetic faces, quantity, and any after-cast machining.
2. **CNC machine** these copies in a named grade. Mark cosmetic versus functional faces and the finish state.
3. **Quote both** a cast path and a CNC path, and tell the shop which rule wins (for example: “cast if we can keep a molded look at about 20 pieces; machine if the critical bore cannot be a secondary operation”).
Upload **STEP plus PDF** on quote-new.php with that paragraph visible in the notes or title block. A cast path includes master finishing, tool cure, and first-shot review; a CNC path includes setups and any finishing stack.
RFQ checklist (cast vs CNC prototypes)
Before you submit on quote-new.php, confirm:
- Process intent: vacuum cast, CNC, 3D print, or “either, with this decision rule”
- Quantity now, and whether a later lot should reuse the same silicone tool
- Whether a master already exists, and in what material and finish
- Color / texture / clear-versus-opaque / rigid-versus-elastomeric intent
- Insert list and after-cast machine features (bores, threads, sealing lands)
- Cosmetic faces versus hidden faces (where gates and parting are allowed)
- Real CNC material grade if the fallback is machining — do not write “PEEK properties, urethane process”
- Which tolerances may be cast-typical and which must be machined
- STEP and PDF uploaded together with the process paragraph visible on the PDF
Common buyer mistakes
- Treating vacuum casting as **cheap injection molding** with unlimited tool life
- Treating it as **CNC with a nicer surface** and leaving fine fits on the cast skin
- Confusing urethane vacuum casting with **metal die casting**
- Sending a sales-sample kit to be **printed raw** or **machined from plate** without saying the goal is a molded look
- Omitting whether the **master already exists**
- Calling PEEK, POM, or 6061 on a cast-urethane RFQ (or the reverse)
- Ignoring gate and parting lines on faces that will be photographed
- Putting the process note only in email — not on the drawing uploaded to quote-new.php
Bottom line
Vacuum casting is better than CNC for **soft, short-run, molded-look prototypes** — several to a few dozen urethane copies from a silicone tool — when appearance, matching sets, and a bridge toward later injection molding matter more than machined-size control or a real engineering-plastic data sheet. CNC remains the right prototype when fits, metals, POM/PEEK-class materials, cut threads, or a CNC production path dominate. 3D printing covers singles and often supplies the master. State the process (or the rule for choosing), list inserts and after-cast operations, and send STEP plus PDF on quote-new.php so the quote band matches the parts you actually want. Ready to quote a short-run housing either as cast copies or as machined plastics? Start on quote-new.php.