Additive manufacturing for production (metal and polymer 3D printing)

Scope: what “production AM” means

“Additive manufacturing for production” is not a single technology. It spans at least seven process families: vat photopolymerization, material jetting, binder jetting, material extrusion, powder bed fusion, directed energy deposition, and sheet lamination. The production-relevant split for this note is:

• Metal AM: laser powder bed fusion (LPBF), electron beam melting (EBM), directed energy deposition (DED), and metal binder jetting.
• Polymer AM: vat photopolymerization, polymer powder bed fusion/sintering, material extrusion, and material jetting.

The technical review literature treats “production” as the transition from prototype/serial low-volume parts to qualified, repeatable industrial manufacturing. Tofail et al. identify the central barriers as process repeatability, material qualification, post-processing cost, and certification burden rather than raw printer speed alone [S2, DOI 10.1016/j.mattod.2017.07.001, Tier 1]. Frazier’s metal AM review describes the same production challenge at the process-physics level: porosity, residual stress, surface finish, and build-to-build variability are the limiting factors for structural aerospace and medical parts [S3, DOI 10.1007/s11665-014-0958-z, Tier 1]. Polymer AM’s production bottleneck is different: it is often not the print step but the material anisotropy, cure/sintering behavior, and post-print mechanical stability [S24, DOI 10.1021/acs.chemrev.7b00074, Tier 1].

Production AM therefore combines hardware, process parameters, feedstock chemistry, thermal post-processing, nondestructive inspection, and qualification data. A claim that a machine is “production-grade” is a system-level claim, not a single-machine claim.

Metal AM technical deep-dive

2.1 Powder bed fusion

LPBF and EBM selectively melt powder layers. The key process variables are laser or electron-beam power, scan speed, hatch spacing, layer thickness, shield-gas flow, and scan strategy. Frazier’s review reports that LPBF is widely used for titanium alloys such as Ti-6Al-4V, nickel superalloys such as Inconel 718, stainless steels, cobalt-chrome, and some aluminum alloys; EBM is used especially for titanium and cobalt-chrome in orthopedic implants because of the vacuum environment and high build temperatures [S3, DOI 10.1007/s11665-014-0958-z, Tier 1]. The production limitations are not only material availability: LPBF generates residual stresses from rapid solidification, requiring stress relief and, for many load-bearing parts, hot isostatic pressing or machining of critical surfaces [S3, Tier 1].

The NIST AM Bench program exists because process variability and part-to-part scatter are serious enough that NIST publishes benchmark measurement series for AM materials and processes S22, NIST AM Bench, [nist.gov , Tier 1]. The existence of a dedicated NIST benchmark series is itself evidence that “production” claims require measured, repeatable data rather than vendor certificates alone.

2.2 Directed energy deposition and binder jetting

DED deposits metal powder or wire through a focused energy source, typically for repair, large near-net shapes, or adding features to forgings. Frazier describes DED as lower resolution than powder bed methods and as a candidate where build size or deposition rate matters more than fine feature resolution [S3, Tier 1]. This is materially different from LPBF economics: DED trades surface finish and feature resolution for high deposition rates on large parts.

Metal binder jetting prints a binder into a metal powder bed and then sinters the “green” part. It is often marketed for higher throughput than LPBF because the print step is faster and the thermal step is a furnace sinter run. The technical trade is shrinkage control and final density; sintering can produce distortion if part geometry, binder content, and furnace profile are not tightly controlled [S2, Tier 1]. This creates a process-control burden that moves AM cost from the printer to furnace qualification and post-sinter inspection.

2.3 Qualification burden

Production metal AM is not only a deposition technology; it is a qualification technology. The same review literature notes that aerospace and medical buyers demand statistical process control, witness coupons, powder-lot traceability, and nondestructive evaluation [S2, Tier 1]. That is why AM adoption tends to appear first in lower-volume, high-value parts: aerospace engine components, orthopedic implants, tooling, and defense platforms.

Polymer AM technical deep-dive

Ligon et al. provide the strongest peer-reviewed polymer AM taxonomy. Vat photopolymerization converts liquid photopolymers into solids by light-initiated crosslinking; material jetting deposits photopolymer droplets; powder-based polymer methods sinter or fuse powder; material extrusion deposits thermoplastic filament [S24, DOI 10.1021/acs.chemrev.7b00074, Tier 1].

The production constraint in polymer AM is material property anisotropy. Layer-by-layer fabrication tends to produce direction-dependent mechanical properties, and the degree of anisotropy depends on the process: extrusion parts are often weaker in the build direction, while vat photopolymerization can produce more isotropic parts but with long-term brittleness or water-uptake concerns [S24, Tier 1]. For production parts, therefore, the polymer class and process are chosen together: a tough functional bracket, a clear aligner mold, and a high-temperature aerospace duct are very different material problems.

Continuous-fiber-reinforced polymer printing is a specific technology claimed in Markforged’s granted patent US 9,126,367 B2, “Three dimensional printing,” granted 2015-09-08 [S1, Tier 1]. The patent describes co-extrusion of a matrix material with a continuous reinforcement strand to produce reinforced parts. It is a granted patent, meaning the USPTO examined and issued claims; it is not proof of part performance, flight qualification, or commercial viability. The polymer review literature independently recognizes fiber reinforcement as a route to improve mechanical properties of printed parts [S24, Tier 1]. That gives two distinct source lineages for the technical concept: the patent record and the peer-reviewed materials-science literature.

Patent/IP analysis: what a granted patent does and does not prove

Granted patent used in this note:

US 9,126,367 B2, “Three dimensional printing,” Markforged, granted 2015-09-08. Patent URL: patents.google.com [S1, Tier 1].

This patent claims methods and apparatus for depositing continuous fiber reinforcement within a printed part. It is a primary IP source and evidence of recorded claims. It is not an independent benchmark. A valid patent claim can be granted even if the commercial process has poor yield, high cost, or no qualified production application. The same caveat applies to every AM patent portfolio: patents protect a mechanism, but market moats also require material qualification data, machine reliability, and customer switching costs.

Distinguishing granted patents from applications: this note cites a granted patent only. Applications are excluded from load-bearing technical claims unless explicitly validated against USPTO/EPO/WIPO records.

Market structure and ticker connections

The value chain is broad:

• Hardware/system vendors: 3D Systems, Stratasys, Desktop Metal, Markforged, Velo3D, HP, Nano Dimension.
• Materials/feedstock: Carpenter Technology, ATI, Kennametal.
• Software/simulation: Autodesk, PTC.
• Service bureaus: Proto Labs, Materialise.
• Key buyers/users: GE Aerospace, RTX, Boeing, Lockheed Martin, Honeywell, Align Technology, Stryker.

Ticker connections

• DDD — 3D Systems sells metal and polymer AM systems, materials, and software, and reports AM revenue in its SEC filings; S4.
• SSYS — Stratasys sells polymer AM systems and materials and is a direct hardware/materials participant; S8.
• DM — Desktop Metal sells metal binder jet and polymer AM systems and markets for production applications; S5.
• MKFG — Markforged sells metal and continuous-fiber composite systems and is the assignee of US 9,126,367 B2; S7, S1.
• NNDM — Nano Dimension sells electronics/polymer AM and has been the counterparty in the proposed Desktop Metal combination; S24.
• VLD — Velo3D sells LPBF systems aimed at production metal parts; listing status needs primary source validation; S25.
• PRLB — Proto Labs operates digital manufacturing services including polymer AM for production-scale custom parts; S10.
• MTLS — Materialise provides AM software and medical/industrial AM services; S11.
• HPQ — HP sells Multi Jet Fusion polymer systems and Metal Jet systems, with supplies/services attach; S12.
• GE — GE Aerospace uses metal AM in engine components and is a major production-adoption reference for LPBF; S13.
• RTX — Raytheon/Pratt uses AM in aerospace propulsion/tooling applications; S22.
• BA — Boeing uses AM for aircraft parts/tooling and has longstanding AM R&D programs; S20.
• LMT — Lockheed Martin uses AM in defense platforms and supply-chain qualification efforts; S21.
• HON — Honeywell uses AM in aerospace and energy components; S23.
• ALGN — Align Technology uses polymer AM in clear aligner workflows, one of the largest polymer-AM production ecosystems by volume; S19.
• SYK — Stryker markets metal AM–enabled orthopedic implants; regulatory and market claims need FDA-clearance validation; S25.
• CRS — Carpenter Technology produces metal powders/wire for AM; S14.
• ATI — ATI produces titanium and nickel-based powders/wire used in metal AM; S15.
• KMT — Kennametal produces AM metal powders and printed wear parts; S16.
• ADSK — Autodesk sells design, generative, and build-prep software for AM; S17.
• PTC — PTC sells CAD/PLM software with AM design and print-prep capabilities; S18.

A few smaller public AM names are omitted because current ticker/listing status could not be verified under the no-invention rule.

Unit economics / business model

The AM business model is not a single model. It is at least four models:

1. Hardware plus recurring materials/consumables. System vendors sell or lease machines and then sell proprietary or qualified powders, resins, build plates, filters, and service contracts. The “razor-and-blade” economics are real but the material lock-in is not total; many metal systems can run third-party powders if users accept qualification risk [S3, Tier 1]. Gross margins are typically higher on materials than on hardware, but exact percentages are company-specific and need primary source validation [S4, S5, S8].
2. Service bureau / per-part manufacturing. Proto Labs and Materialise monetize AM by selling parts and engineering services. They capture value from capacity utilization, process expertise, and software, not from selling a printer [S10, S11].
3. Software/licenses. Autodesk and PTC sell design, simulation, and build-prep software as subscription licenses [S17, S18].
4. Feedstock producers. Carpenter, ATI, and Kennametal sell metal powder and wire; they monetize material volume and chemistry/quality control [S14–S16].

Who captures value depends on where the bottleneck is. In production metal AM, the bottleneck is often qualification and repeatability, which means the material supplier and the machine vendor with validated process recipes capture more value than a commoditized printer manufacturer. In polymer AM, the bottleneck is often material property control and post-processing, which favors resin/powder producers and workflow software.

Margin claims: public AM hardware companies have historically reported losses and going-concern or liquidity risk language in their 10-Ks. Exact gross-margin and operating-margin figures must be pulled from EDGAR; this note does not state them as verified. The correct label for unsourced margin percentages is needs primary source validation [S4, S5, S7, S8].

Worked cost model: LPBF titanium bracket

This is an illustrative cost model, not a company-disclosed price. It uses assumptions derived from cost-driver categories discussed in the peer-reviewed metal AM literature [S3, Tier 1] and should not be read as a market quote.

7.1 Formula

For a build containing \(N\) identical parts:

\[

C_{\text{part}} = \frac{1}{N}\left[

\left(\frac{P_{\text{system}}}{L \times H}\right) T_{\text{build}}

+ N\left(\frac{m_{\text{part}} P_{\text{powder}}}{\eta_{\text{powder}}}\right)

+ R_{\text{labor}} T_{\text{labor}}

+ R_{\text{post}} T_{\text{post}}

+ P_{\text{avg}} E_{\text{price}} T_{\text{build}}

\right]

\]

Where:

• \(P_{\text{system}}\) = machine purchase price, USD
• \(L\) = life, years
• \(H\) = annual productive hours, h/year
• \(T_{\text{build}}\) = total build time, h
• \(m_{\text{part}}\) = final part mass, kg
• \(P_{\text{powder}}\) = powder price, USD/kg
• \(\eta_{\text{powder}}\) = powder-to-final-part utilization fraction
• \(R_{\text{labor}}\) = loaded labor rate, USD/h
• \(T_{\text{labor}}\) = operator setup/removal labor, h/build
• \(R_{\text{post}}\) = post-processing rate, USD/h
• \(T_{\text{post}}\) = post-processing hours, h/build
• \(P_{\text{avg}}\) = average machine electrical draw, kW
• \(E_{\text{price}}\) = electricity price, USD/kWh

7.2 Inputs and result

Assumed build: \(N=4\) titanium brackets, 0.5 kg finished mass each, 40 h build.

Machine rate:

\[

\frac{800{,}000}{7 \times 5{,}000} = \$22.86/\text{h}

\]

Machine cost for 40 h = \(40 \times 22.86 = \$914.4\); per part = \(\$914.4/4 = \$228.6\).

Powder cost per part:

\[

0.5 \times \frac{300}{0.5} = \$300

\]

Labor per part: \(8 \times 75 / 4 = \$150\).

Post-processing per part: \(10 \times 100 / 4 = \$250\).

Energy per part: \(3 \times 40 \times 0.10 / 4 = \$3\).

Total:

\[

C_{\text{part}} = 228.6 + 300 + 150 + 250 + 3 = \$931.6

\]

7.3 Sensitivity

• Improve powder utilization \(\eta_{\text{powder}}\) from 0.5 to 0.8: powder cost falls from $300 to $187.5/part, total falls by $112.5/part.
• Increase annual utilization from 5,000 to 7,000 h/year: machine rate falls from $22.86/h to $16.33/h, saving $65.3/part for this 40 h/4-part build.
• Halve post-processing from 10 h to 5 h/build at constant rate: post cost falls from $250/part to $125/part.
• If powder price rises from $300/kg to $400/kg, total increases by $100/part.

The dominant cost levers in this model are powder cost/utilization and post-processing, not electricity. That matches the review literature’s emphasis on post-process burden and material efficiency [S2, S3].

7.4 Parameter table

Cost-model caveat: the formula is derived from standard cost accounting; the parameter values are not company-disclosed or independently benchmarked. They are assumptions to show sensitivity, not a market price.

Cross-read: marketing claims vs. SEC risk disclosure

The user required a cross-check between marketing claims and public filings. This cross-read is based on investor-facing SEC-linked pages, not on direct EDGAR XML retrieval; exact risk-factor language and pagination need primary source validation.

• Desktop Metal markets production-oriented binder-jet systems and has described high-throughput production economics; its investor page links to SEC filings that include risk factors such as continuing losses, customer-adoption uncertainty, and the need for additional capital. Those two statements are not contradictory—they describe the same business from two legal postures—but they imply different confidence levels. Marketing material is forward-looking promotional language; 10-K risk factors are backward-looking loss disclosures [S5].
• Markforged markets production-grade metal and composite systems; its filings include risk language about commercialization and operating losses. The granted patent US 9,126,367 B2 supports an IP claim about fiber co-extrusion, but the 10-K does not certify that the patent portfolio creates a durable economic moat [S1, S7].
• 3D Systems and Stratasys both market production applications; their periodic SEC disclosures include demand cyclicality, customer consolidation, and competition risks [S4, S8].
• HP markets Multi Jet Fusion and Metal Jet production systems; HP’s SEC filings are issuer-wide and do not break out AM gross margin in the investor page. Specific AM segment margin claims need primary source validation [S12].

The discrepancy is generic but material: marketing pages describe “production” as a machine capability; SEC risk factors describe “production” as subject to qualification, capital-constrained demand, and operating losses. Investors should not substitute a machine’s technical capability for a company’s financial viability.

Disconfirming Evidence

The positive thesis for this technology is that AM is moving from prototyping into qualified production across aerospace, medical, and industrial markets. Evidence that would disconfirm or materially weaken that thesis includes:

1. Persistent losses and cash burn at public AM hardware vendors. If 10-Ks continue to show negative operating margins and repeated capital raises, “production adoption” has not translated into profitable business models [S4, S5, S7, S8].
2. Low serial-production part counts. If major aerospace/medical users continue to qualify AM parts only for legacy or low-volume applications and do not expand part numbers, the production ramp is not occurring [S2, S3].
3. High qualification cost per part. If NIST benchmark data show wide part-to-part or machine-to-machine variability, buyers must over-test and over-inspect, destroying the economics [S22].
4. Powder and post-processing cost resistance. The cost model above shows that powder utilization and post-processing dominate LPBF costs; if those parameters do not improve, metal AM remains a niche for high-value parts only [S2, S3, cost model].
5. Customer concentration. If the leading aerospace/medical buyers cut capital spending or re-insource conventional machining, AM system revenue will fall faster than machine-unit marketing suggests [S5, S7].
6. Patent-portfolio strength is not performance. A granted patent may protect a mechanism, but if no independent performance data show qualified serial parts, the patent does not support a production moat [S1].