Enthusiast builds his own RAM in garden shed cleanroom — fledgling array of memory cells groundwork for much larger future project
One small step, but the good doctor asks us to ‘stay tuned’ for a PC scale result.
Dr. Semiconductor is back in his shed, and this time he’s checking whether Joe Public can DIY themselves out of the DRAM crisis. In the video embedded below, you can see the good doctor go through the semiconductor process steps required to make an array of memory cells in a backyard shed cleanroom. This is the “first time ever RAM has been made at home,” boasts our hero.
RAMageddon isn’t the only issue affecting PC DIYers and the industry in general. Dr. Semiconductor mentions the AI-industry-fueled RAM price disruption being driven by the big three players (but there are others) not being able to keep up with demand. We are seeing similar effects on storage, GPUs, and some reckon CPU supplies will also begin to be impacted.
With the existing industry incapable of addressing consumer RAM needs at attractive prices in 2026, the TechTuber asks, “I turned a shed in my back yard into a class 100 semiconductor cleanroom… but the question is, can I make my own RAM?”
Article continues belowAfter the intro, Dr. Semiconductor gives a brief description of how computer memory works, and how it is largely made up of huge arrays featuring oodles of capacitors and transistors.
Moving on to the practicalities of the job at hand, the good doctor begins by snipping a couple of silicon chips from a large sheet. This is the beginning of the preparation and cleaning stage of the chipmaking process.
Next up, we move to the initial patterning stage. A layer of oxide is built on the surface of the silicon in a high-temperature furnace. It is estimated that this layer is 330nm thick. On top of this layer, an adhesive layer and photoresist film are applied. UV exposure projects a design mask onto this newly created surface, which allows a developer solution to wash away the areas that have been hit by the light rays.
The source and drain of the transistors in the design are formed in the following steps. This involves more layer etching, doping exposed silicon to make it highly conductive, then annealing the chips to push the doping agent deeper.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Several more carefully targeted deposition and erosion steps later, we are ready for the metallization of the silicon chip. A tiny stencil is used to accurately spray the sample with aluminum, the excess stripped away, and the fully layered and formed chip is at last ready for some tests!
Checking the results
The freshly fabricated DRAM cells are so small that wires can’t be used to hook them up to test machinery by the DIYers. Micromanipulator probes are precisely positioned instead. The good news is that Dr. Semiconductor was pleased with his finished DRAM chips, as cells were measured to offer a hobbyist sweet spot of 12pF capacitance.
At the end of the video, the doc teases that he’s going to build on this significant, though admittedly small-scale, achievement. He’s looking to prepare a much larger array of finished memory cells and says that they will be prepared to “hook up to a PC.” Stay tuned for the PC-scale practical implementation, folks.
We reported on Dr. Semiconductor's creation of this garden shed-based cleanroom back in March.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Mark Tyson is a news editor at Tom's Hardware. He enjoys covering the full breadth of PC tech; from business and semiconductor design to products approaching the edge of reason.
-
bit_user It's cool that someone managed to do DIY photo-lithography, but this is obviously not going to provide RAM with generally usable capacities. I'd be surprised if he could achieve even the density of mass-produced RAM from 30 years ago.Reply -
Pierce2623 I get tired of people saying RAM manufacturers can’t keep up with demand. They could if they wanted to. They already cut production way back from COVID levels and won’t increase it again because they’re a cartel wanting prices to stay stupid.Reply -
bit_user Reply
No, they literally can't. It takes several years to build additional fab capacity. The current level of demand was completely unanticipated.Pierce2623 said:I get tired of people saying RAM manufacturers can’t keep up with demand. They could if they wanted to. -
Paul Dodd ReplyPierce2623 said:I get tired of people saying RAM manufacturers can’t keep up with demand. They could if they wanted to. They already cut production way back from COVID levels and won’t increase it again because they’re a cartel wanting prices to stay stupid. -
Paul Dodd Now we need something like a laser 3D printer for chips. You can feed it with a stack of silicon wafers, various materials and after some hours, the finished chips are in the hopper. As there will be several stages, it could be say 3 times the size.Reply -
bit_user Reply
IIRC, Japan's Rapidus has some sort of printing-based technology. Possibly developed by Canon.Paul Dodd said:Now we need something like a laser 3D printer for chips. You can feed it with a stack of silicon wafers, various materials and after some hours, the finished chips are in the hopper. As there will be several stages, it could be say 3 times the size. -
omnichad Reply
And if it's not dense enough, it also won't run cool enough to hit DDR4 or DDR5 clock speeds.bit_user said:It's cool that someone managed to do DIY photo-lithography, but this is obviously not going to provide RAM with generally usable capacities. I'd be surprised if he could achieve even the density of mass-produced RAM from 30 years ago. -
bit_user Reply
It won't be, but you're right that power efficiency is also something that comes with smaller node sizes. Higher clock frequencies, as well.omnichad said:And if it's not dense enough, it also won't run cool enough to hit DDR4 or DDR5 clock speeds. -
AlexAlexAlexApex Replybit_user said:It won't be, but you're right that power efficiency is also something that comes with smaller node sizes. Higher clock frequencies, as well.
I don't follow.Pierce2623 said:I get tired of people saying RAM manufacturers can’t keep up with demand. They could if they wanted to. They already cut production way back from COVID levels and won’t increase it again because they’re a cartel wanting prices to stay stupid.
In a free market, yes - low supply will increase prices.
But if (as you suggest) the manufacturers were in a cartel, they could each make as much ram as possible, and secretly agree to still sell it at a high price, i.e. rather than undercut one another as they would in a free market. That's the whole point of a cartel (and why they're not allowed).