The fab bottleneck isn't the lithography tool — it's the OHT cars on the ceiling
What really caps wafer fab capacity isn't just lithography tools, but the OHT cars running back and forth on the ceiling. The 300mm transition made them mandatory, and today the market is a Daifuku–Muratec duopoly.
The argument · tap a timestamp to hear it
The fab's transport problem is set by how bays are laid out
A fab is divided into processing zones, and within each zone tools are arranged into bays. Moving wafers inside one bay is Intrabay; moving between bays is Interbay. Each bay gets stockers — essentially automated parking towers that hold roughly 150-250 wafer boxes. Stockers replaced open shelving, saving cleanroom floor space, helping track work-in-progress, and serving as buffers and zone entry/exit points. Every problem in the transport system grows out of this layout.
Cleanroom floor is too expensive, so transport has to go up to the ceiling
The moment a wafer leaves a tool it must be inside a FOUP, because falling particles will destroy it. Early on, people pushed FOUPs between bays by hand, but a single box of work-in-progress could be worth $100,000, so automation began in the 1980s. First came floor-running AGVs, limited by safety rules to about 1 foot per second; then rail-guided RGV systems ran faster but consumed floor space and needed isolation. PRI's Aerotrak put a monorail on the ceiling, and by 1988 this kind of interbay AMHS was common. Cleanroom floor is expensive, so this move upward was inevitable.
Once interbay was automated, people became the new bottleneck
After interbay automation, the bottleneck became the manual hauling operators did between stockers and tools. A skilled operator took 30-90 seconds per load change, but a bay has many tools. The author heard constant beeping in a DRAM fab and was told it was tools paging people to change loads. Intrabay automation has to sense tool processing state and load/unload automatically; get the alignment or timing off and the tool goes down. In the early years there was no industry standard, so each company invented its own. Hot lots and SUPER hot lots, because they were a pain to program, were often still grabbed and run over by hand.
300mm wasn't an upgrade, it was automation's final ultimatum
The shift to 300mm wafers in the late 1990s made the old manual handling unsustainable. A SEMATECH study found 15-20% of tool time in older fabs was wasted waiting on operators or FOUPs; tools were also more expensive, heavier and more floor-hungry. A fully loaded 300mm FOUP weighs about 9 kg, above the ergonomic "maximum acceptable weight" for repeated lifting. A survey of Taiwanese 200mm fabs: 40-60% of workers reported shoulder discomfort, 30-50% back problems. A high-traffic 300mm bay needs 300+ moves per hour, versus 125-175 in older fabs, so automation was the only option.
OHT won by not taking up floor space, not by being fast
The core of second-generation AMHS is OHT: vehicles drive along ceiling rails to directly above a tool and use a hoist to pick up or drop a FOUP at the tool's front-end interface in a few seconds, at up to 60 meters per minute. It takes no floor space and doesn't compete with people for aisles, but installation is expensive, routing is inflexible, and hanging above tools carries particle risk. For this, SEMI set common standards for loadport configuration and for optical signal exchange between vehicle and tool, ensuring handoffs succeed and saving cost. OHT became the only choice for a unified system.
A system costs over $100 million, so most fabs kept using people
TSMC was one of the first companies to open a fully ceiling-mounted AMHS 300mm fab: Hsinchu Fab 12, which ran 150nm first and then moved to 130nm, with 2,000 vehicles inside moving 600,000 trips per day. But the scale and cost of the whole system are daunting — a complete 300mm interbay/intrabay system runs $50 million to $100 million, and installation takes 2 years. So for many years afterward, many fabs still relied on human labor to move wafers. Automation was not a one-time switch.
There is no optimal AMHS schedule, only trade-offs
Scheduling is done centrally by the fab's control center: assigning vehicles, planning routes, handling congestion, and avoiding deadlocks and down tools. A fab won't keep extra vehicles for busy periods; it can only use the ones it has, and hot lots and rework have to be written as special rules. The author offers one judgment: there is no optimal algorithm. So operators use simulation to optimize objectives such as average delivery time, on-time delivery rate, cycle time or throughput. It's like designing a traffic system for the entire fab.
The 300mm transition reshuffled the OHT supplier landscape
Daifuku is the largest OHT supplier; it started out making steel forging machinery, changed its name after WWII to avoid the zaibatsu breakup, and took its name from the first characters of its Osaka and Fukuchiyama plants. The turning point was licensing Jervis B Webb's chain conveyors and supplying Toyota's Motomachi plant in 1959, growing along with the rise of Japanese automakers. The 300mm transition reset the market: PRI got into trouble over Aeroloader development costs and sold to Brooks Automation for $500 million; in the end Daifuku and Muratec both consolidated the industry and have monopolized it ever since.
In their own words · checked verbatim
The first things you will notice upon stepping inside a fab clean room are the equipment. Massive, inscrutable, expensive. But the second things you will notice are the little vehicles puttering around on the ceiling above your head.
I was told that it was because the tool wants to let the operator know that it needs its diaper, sorry wafers, changed.
They even have SUPER hot lots which represent the absolute first priority for every tool and operator on the line.
A study done by SEMATECH in the mid-1990s discovered that in older fabs, 15-20% of tool time is wasted waiting for an operator or FOUP to arrive.
The OHT vehicle aligns itself above the tool. And then it lowers a hoist to either pick up or drop off the FOUP right from the tool's frontend interface. It is magnificent to watch and just takes a few seconds.
There is no optimal algorithm.
In the end, Daifuku and Muratec consolidated the industry between themselves and they dominate it to this day.
Figures
| Distance a 200mm wafer traveled inside a fab in the 1990s | 8-10 miles, touching up to 250 tools | 2:08 |
| Cost per square foot of an older 200mm cleanroom | $350-700 | 5:11 |
| Cost per ping of cleanroom at leading fabs | $89,000-265,000 | 5:11 |
| SEMATECH's figure for tool time wasted in older fabs | 15-20% of tool time waiting on operators or FOUPs | 9:18 |
| Weight of a fully loaded 300mm FOUP | 19.8 lbs / 9 kg | 10:21 |
| Share of workers reporting physical discomfort at Taiwanese 200mm fabs | 40-60% shoulders, 30-50% back | 10:21 |
| Moves per hour in a high-traffic 300mm bay | 300+, versus 125-175 in older fabs | 11:28 |
| Top OHT travel speed | 60 meters per minute | 12:29 |
Glossary
- AMHS / Automated Material Handling System
- The collective term for the automated equipment that moves FOUPs inside a fab, covering AGVs, OHT and stockers.
- FOUP / Front Opening Unified Pod
- The standard wafer container, typically holding 25 wafers, used to move them safely between tools.
- OHT / Overhead Hoist Transport
- A vehicle that runs on ceiling rails and uses a hoist to pick up and drop FOUPs at a tool's front-end interface.
- Stocker
- A wafer storage unit like an automated parking tower, used as a buffer and as a zone entry/exit point.
- Interbay / Intrabay
- Interbay is transport between bays; Intrabay is transport between tools within the same bay.
How to listen
Fab manufacturing and automation engineers, semiconductor equipment investors, and anyone trying to understand the hidden costs inside the advanced-node capacity story.
You can fast-forward through Daifuku's pre-WWII founding history and start at the industry consolidation conclusion at 18:38.