Industries / Industrial equipment and services / 3D printing
Plastics
Plastics printing makes parts from polymers in one of four main ways: it hardens liquid resin with light, lays down melted plastic from a moving nozzle, fuses plastic powder with a laser or heat, or jets tiny droplets that harden, building the part one thin layer at a time.
7
Milestones achieved, from 1986 to 2022.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1986
Light hardens resin into parts
In 1983 Chuck Hull built a system in which an ultraviolet (UV) laser traces each cross-section of a part in a vat of liquid photopolymer (a plastic that hardens under light), while a platform lowers the part one layer at a time. The process is called stereolithography. Hull filed for a patent on August 8, 1984, received US patent 4,575,330 in 1986, and co-founded 3D Systems that year to sell machines. The company's SLA-1 of 1987 is, by ASME's account, the first 3D printer made for commercial sale, and The Henry Ford museum holds the first one the company shipped. ASME's timeline also credits Hideo Kodama of the Nagoya Municipal Industrial Research Institute in Japan with the first patent application for this kind of layered prototyping, filed in 1980, after he made plastic parts by hardening polymer with UV light.
Credit
Chuck Hull and 3D Systems; Hideo Kodama for the earlier Japanese application
1989
A laser fuses plastic powder
Carl Deckard, a mechanical engineering student at the University of Texas at Austin, had the idea by the end of his senior year in 1984: spread a thin layer of polymer powder, melt together (sinter) only the part's cross-section with a laser, and repeat. Working under his professor Joe Beaman, he built a first machine around a 100-watt laser controlled by a Commodore 64 home computer. The university's patent (US 4,863,538, filed in October 1986) was granted in September 1989, the year selective laser sintering (SLS) was first shown in public, at the Autofact trade fair. The university licensed the process in 1988 to the start-up that became DTM, and the first machine was sold to Sandia National Laboratories, which used it to make casting patterns.
LessonBeaman recalls the team expecting to print cars within two years, while the first buyers wanted prototypes and casting patterns.
Credit
Carl Deckard and Joe Beaman, University of Texas at Austin; DTM
1992
Melted plastic from a nozzle
S. Scott Crump's patent, filed in October 1989 and granted to Stratasys in June 1992 (US 5,121,329), describes a computer-driven head that lays down a material heated until it flows, in thin layers that harden almost at once on the layer below; it cites wax guns and glue guns as earlier hand tools. ASME's timeline says Stratasys produced the world's first fused deposition modeling (FDM) machine in 1991. Hackaday wrote in 2013 that the current crop of 3D printers used this "squirting melted plastic" method, and dated the key patent's expiry to 2009.
Credit
S. Scott Crump, Stratasys
2008
A printer prints its parts
Adrian Bowyer at the University of Bath in England started RepRap, a desktop printer that extrudes plastic and is built largely from parts it can print itself, and released the designs free under the GNU General Public License. The project dates its start to February 2004, while London's Science Museum says 2005. By the project's count, commercial machines then cost about €30,000, while a RepRap's materials cost about €350. On May 29, 2008, at Bath, Bowyer and Vik Olliver showed the first complete working "child" machine made by a parent RepRap; the first RepRap, the Darwin built by Bowyer and Ed Sells, is now in the Science Museum's collection. Hackaday links the 2009 expiry of the FDM patent to the rise of 3D printing that followed, with the RepRap Mendel and the founding of MakerBot.
LessonFree designs and an expired patent together let hobbyists and small firms build printers for a small fraction of the old price.
Credit
Adrian Bowyer, Vik Olliver and Ed Sells, University of Bath
2012
Crowdfunded resin printer, then lawsuit
Formlabs, then a three-person start-up from the MIT Media Lab, raised close to $3 million on Kickstarter in 2012 for its first stereolithography printer, the Form 1, and shipped more than 1,000 printers. In November 2012, 3D Systems sued, claiming the Form 1 infringed a stereolithography patent granted to it in 1997 (US 5,597,520, which ran until 2017); the two companies settled in December 2014 on terms that were not made public. By CONTEXT's count, Formlabs had the largest share of global shipments in the professional price class in 2025, at 38%, and resin (vat photopolymer) printers made up 71% of that class's units in the last quarter of 2025.
LessonPatents on later improvements can still block a newcomer after the first patent on a process has run out.
Credit
Formlabs; 3D Systems as plaintiff
2014
A rushed extruder backfires
MakerBot, owned by Stratasys, unveiled its fifth-generation printers at CES in 2014 with a "Smart Extruder", a swappable unit that melts the plastic filament and pushes it out. Owners reported frequent clogs; Hackaday put the part's price at $175 and its life at 200 to 500 hours, and noted that Stratasys shares fell from $130.83 in early 2014 to $31.88 by July 2015. Investors sued in July 2015, alleging the printers were rushed into production without a standard testing program. A federal court in Minnesota dismissed the suit with prejudice in 2016, ruling that the quality claims were "puffery" (sales talk that cannot be checked), and MakerBot later fixed the extruder.
LessonTest reliability before shipping, and judge printers by measured numbers, because the court held that words such as "unmatched" quality cannot be checked.
Credit
MakerBot and its parent company, Stratasys
2022
Printers that set themselves up
Bambu Lab, then a little-known Chinese company, put its X1 printer on Kickstarter in May 2022 and raised over $7 million from 5,575 backers, by its own count. The X1 came assembled and was sold on speeds up to 500 mm/s, automatic calibration using lidar (a laser distance sensor), a camera watched by an AI chip inside the printer that pauses failed prints, and a four-spool unit for multi-color printing. By CONTEXT's count, Shenzhen-based Bambu Lab shipped 37% of the world's entry-level printers (under $2,500) in 2025, and Chinese makers shipped over 90% of them.
LessonCheap machines that calibrate themselves pulled buyers away from pricier extrusion printers, and CONTEXT puts most of the 15% fall in professional-class shipments in 2025 down to Bambu Lab.
blog.bambulab.comcnx-software.comwiki.bambulab.comtctmagazine.comtomshardware.com
Credit
Bambu Lab
Watch for
Nine tools, six-second swaps
Bambu Lab says its H3C desktop printer carries nine print heads that it swaps between (a toolchanger), each with its own filament path and heater, and changes material in under 6 seconds; with an older AMS unit attached, one print can use up to 24 materials. The company also claims moves up to 1,400 mm/s, prints that finish on average more than 10% faster than on its previous flagship, and flexible TPU plastic at up to three times the flow of a conventional printer, with some features due by the first quarter of 2027. The US price is $2,499, and the machine was shown at Formnext in Frankfurt on November 17 to 20, 2026.
Achieved whenindependent reviewers time tool changes and whole prints on shipped units and confirm the speed gain.
Claimed by
Bambu Lab, November 2026
Toolchanger claims less waste
Snapmaker says its U1, a four-head toolchanger, swaps heads in 5 seconds and, in its own lab tests, used up to 85% less filament than printers that switch filaments through one shared nozzle, for the same multi-color prints. Single-nozzle multi-color printers push out and discard plastic at each color change (purge waste), which Snapmaker says can exceed the plastic in the model; it also says its head-locking mechanism ran over 1,000,000 swaps without a mechanical failure. The U1's Kickstarter raised about $20 million from about 20,000 backers, which All3DP reported on October 1, 2025 as a record for 3D printing hardware. One outside figure exists: Tom's Hardware printed the same multi-color batch on a pre-release U1 and on a Bambu Lab P1P in August 2025 and reported 4.4 g of waste on the U1 against 25 g more on the P1P, without saying whether the waste was weighed or taken from the slicer's estimate.
Achieved whenan independent tester weighs, on a scale, the waste from several multi-color models of low to high color-change counts printed on a shipping U1 and on shared-nozzle printers, and publishes the raw weights.
Claimed by
Snapmaker, on its U1 product page, as read in November 2026
Desktop full-color material jetting
HeyGears, a dental 3D printing company from Guangzhou, China, says its G1X will print full-color parts on a desk by material jetting (an inkjet head lays down tiny droplets that harden, layer by layer), using an Epson i3200 head with 3,200 nozzles, 3.9-picoliter droplets, layers 10 to 20 µm thick and water-soluble supports. Its Kickstarter ran from July 23 to September 11, 2026 and raised $17,685,524 from 3,981 backers, with campaign units due to ship in December 2026; Makers101 notes that HeyGears has published no time for printing a full build plate. Market analyst CONTEXT expected full-color material jetting products to come to market in 2026.
Achieved whenbuyers have shipped units and an independent reviewer publishes print times, color accuracy and part strength.
Claimed by
HeyGears, July 2026
Open
01
A labeled print-failure benchmark
Add outcome labels to a public set of whole-print videos from desktop printers: whether the print failed, the kind of failure (part knocked loose and "spaghetti" tangle, clogged nozzle, warping), and the first frame where the failure shows. Publish a scoring script that rates a detector the way a printer uses it: would it have stopped each bad print, how many minutes after the failure began, and how many good prints it would have stopped.
The briefCredit · unclaimed02
Head-to-head failure detector test
A filmed, repeatable test of the failure detectors that buyers rely on, run on the same failures: Bambu Lab's built-in detection, Obico, and an open YOLO11 model. Cause failures on purpose with edited print files, record them with a separate time-stamped camera, and count catches, misses, minutes to pause and false pauses on light, colored and black filament.
The briefCredit · unclaimed
Open 01 · the brief
A labeled print-failure benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Add outcome labels to a public set of whole-print videos from desktop printers: whether the print failed, the kind of failure (part knocked loose and "spaghetti" tangle, clogged nozzle, warping), and the first frame where the failure shows. Publish a scoring script that rates a detector the way a printer uses it: would it have stopped each bad print, how many minutes after the failure began, and how many good prints it would have stopped.
Why it is still open
Searches of Hugging Face, Roboflow, Kaggle, Zenodo, IEEE DataPort and university repositories in November 2026 found print-failure datasets but no benchmark of this kind. The closest work falls short in three ways. The community Bambu Timelapse Dataset on Hugging Face (CC BY 4.0, last updated April 2025) holds several hundred time-lapses but states that "No print-outcome labels are provided yet". Cambridge's CAXTON dataset has 1,272,273 images from 8 printers labeled with print settings, not outcomes, and its filtered version removes complete failures. A 2025 University of Twente thesis proposed a "Print Failure Stopping Metric" and found a YOLO11 model beat Obico's older YOLOv2-based detector, but on its own recordings, with no shared test set.
Built with what exists
The Bambu Timelapse Dataset as a starting set; Label Studio or CVAT (open-source labeling tools); the open-source Obico server model and Ultralytics YOLO11 as baselines; Python for the scoring script; Hugging Face for hosting. The advance that makes this practical is fast object-detection models such as YOLO11, and the AI chips now built into some printers, which run these models on every camera frame.
First step
Download the Bambu Timelapse Dataset, write a one-page labeling guide, and label the outcome and failure start for the first 100 time-lapses.
Done looks like
At least 300 whole-print videos labeled, at least 100 of them failures, released under CC BY 4.0 with a fixed test split and scoring script, and published scores for at least two detectors (Obico's model and a YOLO11 baseline).
Open 02 · the brief
Head-to-head failure detector test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A filmed, repeatable test of the failure detectors that buyers rely on, run on the same failures: Bambu Lab's built-in detection, Obico, and an open YOLO11 model. Cause failures on purpose with edited print files, record them with a separate time-stamped camera, and count catches, misses, minutes to pause and false pauses on light, colored and black filament.
Why it is still open
Searches in November 2026 for independent tests of Bambu Lab's detection against Obico or Prusa found only vendor pages and forum anecdotes, including reports of missed failures on black filament. Bambu's own wiki says accuracy is "still not high enough", that false alarms are reported, and that black or dark filament is harder, but gives no rates. Obico's 2026 comparison lists features and admits its AI makes false positives and misses, without a controlled test. Bart Leenheer's University of Twente thesis compared YOLO11 with Obico's model on recorded images, not on live failures and not with the detectors built into printers.
Built with what exists
A Bambu Lab X1C (built-in detection with a chamber camera), with Obico watching the same printer (Obico supports Bambu Lab printers); a USB webcam with a clock overlay recorded through OBS Studio or ffmpeg as ground truth; G-code edits that raise the nozzle several millimeters mid-print so it extrudes into air, which makes the spaghetti tangle the detectors look for; Ultralytics YOLO11 run on the webcam feed as an open baseline. Stop any print by hand after a fixed time if no detector does, to keep the tangle from wrapping the hot end.
First step
Write one G-code edit that reliably makes the nozzle print into air, run it five times with Bambu's detection and Obico both watching, and post the clips and timings.
Done looks like
For each detector, 20 induced failures on each of three filament colors (white, red and black), plus at least 100 hours of normal prints, with the videos and a public tally of the share caught, median minutes to pause and false pauses per 100 hours.