What Happens When 15,000 Desktop 3D Printers Become a Factory?

When I first travelled to Formnext Shenzhen in 2023 I saw a familiar pattern, the show tells you what China’s additive-manufacturing industry wants to sell. Go outside the show and you can start finding out what is actually being made. At Huafast’s Shenzhen operation, 5,600 Bambu Lab machines form part of a 15,000-printer production network. But buying the machines is the easy part. The harder problem is managing the labour, materials, orders and commercial pressures that come with manufacturing at scale. The printers at Huafast are arranged in rows, but the business is organised around deadlines. During a visit to its Shenzhen operation in August, I toured the familiar equipment of desktop 3D printing alongside the less photogenic requirements of manufacturing: finance, administration, processing, packaging and a warehouse. The company had moved into the Guangming site in May. Some office decoration remained unfinished. Its founder, Sven, poured tea from his wife’s hometown before showing me around. His office contained a camp bed, a guitar and boxes of material, as well as a tea table with its own heating and drainage. Guangming is also the district where Bambu Lab will build its 3 million 3D printer factory. Sven put the Shenzhen fleet at 5,600 printers, part of approximately 15,000 machines purchased by the parent company across several locations in China. All were Bambu Lab machines, he said. The figures are the company’s, but the production operation surrounding the printers was there to see. The factory raises a question that equipment classifications handle poorly. A desktop printer can produce a prototype, a fixture or a product for sale. Its price does not determine its purpose. At Huafast, the interesting distinction lies in the organisation around the machine: how an order is allocated, supplied with material, checked and delivered, and whether enough money remains afterwards to justify doing it again. Seven days to catch a market Sven described a toy order exceeding 50,000 pieces, completed in seven days. Huafast allocated roughly 1,000–2,000 printers to the work while other machines continued serving existing customers. Using more of the fleet, he estimated, could have brought the production time down to three days. But the machines were not all available, and customers already in the queue could not simply be displaced. That qualification matters. A printer count measures potential capacity. A delivery promise depends on how much of that capacity can be made available for a particular job. Many of Huafast’s customers are traders selling through their own websites, Amazon and TikTok shops, including to overseas buyers. For products that spread through social media, demand can arrive faster than conventional tooling. Sven described a short window in which a new design attracts both orders and an attractive price. Within another week or fortnight, competitors may have entered and prices begun to fall. Meme products are where the economics of a printer farm become especially obvious. Demand can appear almost overnight and disappear just as quickly. Google Trends data for so-bad-its-good animated film “Niu Lai,” or The Cow is Coming, shows virtually no search interest before a sharp August spike, peaking on 21 August before falling away through September. Huafast’s response is to treat production capacity as something that can be redirected just as quickly: once a design starts moving online, files can be prepared and distributed across hundreds or thousands of printers without waiting for tooling. Sven said this speed from idea to customer is central to the business model. “You have to catch the market,” he said. In that setting, the comparison with injection moulding extends beyond the eventual cost per piece. A cheaper manufacturing process is less useful if production arrives after the opportunity. Huafast can put many printers on the same order without commissioning a dedicated mould, then allocate those machines to something else when the work is finished. The products are not uniformly short-lived. An articulated dragon Sven showed me had been selling for three years. The famous dragons may have become a cliché of desktop 3D printing, but clichés occasionally persist because customers continue buying them. Critics dismiss this churn of meme merchandise as “slop” or future “landfill”: cheap objects manufactured at internet speed for trends that may vanish almost as quickly as they appeared. The hostility can become surreal; Sven said some commenters had even accused him of being AI-generated. A large toy order is still a manufacturing job. It has a specification, a quantity, a deadline and a customer expecting delivery. Testing demand without buying a mould Printing also changes how much a customer must decide before discovering what sells. Sven described customers making 300–500 units of each of roughly 20 designs, then releasing them to see which attracts demand. Instead of committing to tooling for a particular product, they can spend money on a range of saleable objects and use the response to guide the next order. It resembles A/B testing translated into physical goods. Instead of using a forecast to decide which product deserves tooling, the customer can use actual sales. The benefit is a smaller commitment to a particular design and an easier route to changing it. Successful products may eventually justify moulding; others may continue to sell in quantities for which printing remains suitable. Another group of customers has already tested the manufacturing process. Sven said some companies own one or two Bambu printers and have established that the output meets their needs. Their difficulty begins when they require 500 or 1,000 parts within a few days. Huafast sells them the capacity to make that jump without building a farm themselves. He put typical industrial orders at around 500-1,000 pieces and the largest at approximately 100,000. Applications include electronics cases and covers, using materials such as PETG, ABS, ASA and carbon-fibre-filled filament. He estimated that industrial work had risen from roughly 10% of the business three years ago to about 30%. His definition is broad, covering the non-toy work essentially. The proportions are an approximate description of the business mix, rather than a detailed revenue breakdown or a measure of qualified industrial applications. Huafast gave another recent industrial order as an example of what the farm can do when delivery time matters. According to Sven, the company produced 60,000 small industrial PLA parts in seven days, allocating 100 H2S printers to the job. That is roughly 600 parts per machine over the week. The customer’s reason for choosing additive manufacturing was straightforward: speed. The example also illustrates where the economics of a printer farm become interesting. Sixty thousand identical parts is already a quantity at which injection moulding would normally deserve serious consideration. But moulding requires tooling before the first production part exists. A farm can instead throw another hundred machines at an order almost immediately. The advantage is therefore not necessarily the lowest theoretical unit cost; it is the ability to turn available machine capacity into tens of thousands of parts before a conventional production line may even be ready. Questioned on where moulding becomes more economical, Sven resisted a single threshold. For one example, he suggested above 10,000 units, but stressed the design and whether it would continue selling. A forecast of sustained demand can justify tooling that a brief rush of orders cannot. The work between prints The flexibility offered to customers creates work inside the factory. According to Sven, the Shenzhen operation employs 42 people, including production management and more than 30 production workers. How many printers one person can manage depends heavily on what those printers are making. In a written follow-up, Sven described an order passing through sales and technical staff before reaching production, quality control, packing, the warehouse, and shipping. The printers occupy one stage in a sequence of responsibilities extending from the customer to dispatch. An operator responsible for approximately 300 machines producing the same item has a different job from someone overseeing 100 machines running a mixture of files and colours. Sven described 20 or more files as an example of the complexity that brings the manageable number down. Changing materials and keeping different orders straight consumes time that a uniform run does not. That is a more useful account of labour productivity than dividing the printer count by the workforce. The product variety that makes printing attractive to customers also increases the coordination required to deliver it. Much of the online discussion about automating printer farms focuses on getting the finished object off the build plate. YouTube demonstrations use conveyor beds, robots or even the print head itself to eject completed parts and begin another job. At Huafast’s scale, however, part removal is only one piece of the automation problem. His account of operators’ work was practical: running printers, inserting filament and removing finished products. Filament changes, he said, took a substantial amount of their time. Huafast was testing automatic build-plate collection on two machines, but his priorities placed material handling alongside collection of finished parts. Both would reduce the manual work required between print jobs. Controlling the fleet introduces another difficulty. Sven reported Wi-Fi interference when trying to manage thousands of printers equipped with 2.4 GHz connectivity. Some groups were controlled wirelessly, while much of the operation relied on manual control. His account describes the problems of this installation, rather than a universal limit on the number of printers a network can support. Software suppliers have approached Huafast with management tools, he said. Yet the economics of developing them are awkward. An operator with hundreds of machines may consider its existing arrangements sufficient; relatively few businesses have several thousand printers and a reason to pay for more elaborate control. Huafast’s needs are substantial, but they do not necessarily represent a large market of paying customers. Asked in the written follow-up about operator interventions and failed or interrupted prints, Sven estimated 2.5 to 10%, depending on the file. He did not distinguish between those categories, so the figure cannot yet be treated as a reject rate. Scale also changes the consequences of mistakes. During the interview, Sven described incorrect files, unchanged settings and the wrong filament colour. A mistake in preparation can be reproduced across hundreds of machines. The printers may execute their instructions successfully and still produce an unacceptable order. “It’s a very big loss,” he said. This is a different problem from an individual worn component. One of the older machines we examined had accumulated roughly 7,800 printing hours. Sven said some had passed 10,000 hours and identified nozzles, extruder gears and belts as common replacement items. Quieter periods provided an opportunity to maintain machines and return them to service. A large fleet offers flexibility when one machine is unavailable. It offers much less protection against a wrong instruction sent to many machines at once. Expanding the fleet therefore does not remove the need to maintain equipment and control production. Cheap machines still need orders Huafast began on the other side of the transaction. Sven said he founded the company in 2018 and initially made 3D printers. In 2023 it stopped manufacturing machines and moved into printing services. An acquisition by a listed company in 2025, he said, provided capital for expansion from nearly 1,000 printers to more than 5,000. Affordable machines make capacity easier to add in increments. Thousands of them still require capital, premises and staff. The customer’s ability to avoid a mould does not remove those commitments from the business producing the parts. Sven estimated that, on average, around 50 to 60% of the fleet was operating, with peaks of 80 to 90%. During the tour he put the current figure at roughly half. He repeatedly returned to the need for more orders to keep a larger proportion working. A lower utilisation rate after rapid expansion does not necessarily mean less work. Half of a much larger fleet can represent substantially more activity than a heavily occupied smaller one. But it does mean the installed machine count is an incomplete guide to commercial performance. Huafast has to find demand for the capacity it has bought. Asked how long a printer typically takes to repay its investment, Sven said: “Average, I think six months or eight months.” A strong case could be around three months, he estimated, while less attractive order pricing could extend that to ten months or a year. These are the operator’s estimates of machine payback, not audited returns on the entire factory. Sven declined to disclose revenue in the written follow-up, and the reporting did not establish a full cost basis covering premises, labour, rejects, finance and other overheads. His account of the market contained a more immediate warning. Orders were growing, he said. “Just the price is dropping.” The same accessible equipment that enabled Huafast’s expansion also allows other operators to compete. Sven described smaller farms run from home with help from relatives. Their cash overheads can be lower than those of a company paying for a larger organisation. Huafast cannot always match their prices. Its advantage is the ability to take larger orders, meet short deadlines and organise the work through established processes. The business depends on customers valuing those things enough to pay for them. More printers do not automatically mean a lower price or a better margin. Selling printers to the competition Huafast’s relationship with smaller farms is consequently more complicated than a contest for orders. The company also resells Bambu printers in China. Sven said smaller farms are among its equipment customers, and that Huafast helps them establish operations and shares experience of managing files, materials and workers. Those operators can then become suppliers. When Huafast’s machines are heavily occupied and another order arrives, Sven said it can pass work to other farms. It can also direct work to a smaller producer when the customer’s required price is below its own cost base. A competitor can therefore be a buyer of machines, a recipient of technical advice and a subcontractor. Huafast benefits from relationships with capacity beyond the equipment its parent company has purchased. That partner capacity should be distinguished from the reported 15,000-machine group total. “If you have big orders, find us,” Sven said. It is a straightforward statement of what the larger operation offers. Owning a printer gives a customer the ability to make an object. Owning a hundred gives considerably more capacity, along with more work to organise. At larger volumes and tighter deadlines, the customer may prefer to buy the result from someone already managing those problems. The material operation downstairs The wider production system became more visible during a tour of the materials businesses. In one building there were bulk bags of polymer granules and two production lines. On the ground floor of the main building I was shown a filament operation described as having 35 lines. Sven explained that material was produced both for internal use and for other brands, including overseas customers. He also described an affiliated company making filament-production equipment. The printing operation was connected to businesses manufacturing both the consumable and the machinery used to produce it. This was more than a stockroom supplying the printers upstairs. White-label production put the materials business into other companies’ supply chains as well. Sven described buying raw polymers from international and Chinese suppliers, then processing them into filament. Customers can specify particular filament brands, he said. Where they do not, the farm can use material from the affiliated operation. Control over production capacity and access to material supply therefore sit close together, even though the business still depends on outside inputs and customer requirements. The printer rows were the most obvious image of scale. The material lines showed another part of the organisation supporting them. Bambu’s machines, Huafast’s factory Huafast’s choice of equipment followed experience with Bambu machines from 2023, Sven said. He cited their quality and stability, described visits by Bambu employees to discuss operating problems, and said the parent company had a strategic agreement with the manufacturer. The details of that agreement remain to be established. The relationship gives Bambu access to a customer confronting problems that a household or small workshop rarely encounters. It also demonstrates how desktop equipment can become the basis of a substantial production business without changing its basic identity as a machine. That does not make every application interchangeable with those served by high-end industrial systems. A customer still needs the appropriate material, dimensional accuracy, mechanical performance and production controls. A printer’s affordability cannot answer those questions, any more than an expensive machine can guarantee a profitable order book. Huafast brings the distinction into focus. Bambu supplies the machines. The operator must assemble the workforce, material supply, production methods and customer relationships that turn their output into a business. As Sven’s utilisation figures and comments on pricing suggest, acquiring capacity is only the beginning. The harder task is keeping it profitably occupied. The answer to What Happens When 15,000 Desktop Printers Become a Factory? isn’t they become industrial printers. Nothing magical happens to the printers. Everything around them has to become industrial. Read more on the 3D printing industry in Shenzhen. Register now for AMA: Software 2026. Join 3D Printing Industry on 22 October for expert presentations, panel discussions and live audience Q&A exploring the software shaping additive manufacturing. Register here. To stay up to date with the latest 3D printing news, don’t forget to subscribe to the 3D Printing Industry newsletter or follow us on LinkedIn. Explore the full Future of 3D Printing and Executive Survey series from 3D Printing Industry, featuring perspectives from CEOs, engineers, and industry leaders on the industrialization of additive manufacturing, 3D printing industry trends 2026, qualification, supply chains, and additive manufacturing industry analysis.