In the space of five years, it looks like 450mm manufacturing has become surplus to current requirements

1 min read

Five years ago, the semiconductor industry was getting quite excited about what looked like the next step in its evolution. That step was to embrace manufacturing chips on 450mm diameter wafers – and the plans called for 450mm manufacturing to be mainstream today.

There were a couple of drivers for the migration, including the ability to increase throughput and a potential reduction in cost per die. Intel was one of the companies leading the campaign for 450mm manufacturing, with support from TSMC and, less enthusiastic, Samsung. But there was also interest in 450mm from the European Commission, which commissioned a report into whether it might be feasible to build such a fab in Europe.

One of the authors of that report was Malcolm Penn, chairman of industry watcher Future Horizons. Who better to ask about where 450mm is today?

“The world has changed,” he said. “At the time, Intel believed it couldn’t build enough fabs, but its market has changed and so has industry’s desire to go to 450mm.”

Nevertheless, Intel did some 450mm exploratory work in Oregon, while TSMC built a pilot plant in Taiwan.

In Penn’s view, equipment manufacturers didn’t want to do 450mm because they didn’t see any reason, but agreed to investigate, so long as it was a coordinated project with a single agenda. “However,” he said, “R&D work has carried on and equipment manufacturers have used that to improve 300mm manufacturing – something not intended.”

One factor which influenced progress was the economics. “It doesn’t bring a universal cost per die improvement,” Penn pointed out. “It depends on the process being used and the more lithography there is, the less the cost benefits because it’s a stepper process. But etch and deposition are per wafer processes and do bring benefits.”

But one technology might rescue 450mm manufacturing – 3D flash. “There’s a lot of etch and deposition involved in 3D flash,” Penn said, “perhaps 128 steps, and that could bring a serious cost advantage over 300mm manufacturing. But the process isn’t working well yet.

“If Samsung – and maybe Intel – can get 3D flash sorted, there will be the need for 450mm and there will be a baseline for it to expand into logic.”

Meanwhile, Europe’s appetite for 450mm has disappeared. The idea was part of Neelie Kroes’ 10:100:20 project – where €10billion of EU funding and €100bn from industry would capture 20% of the global chip business – but, said Penn, European chip companies have ‘vetoed’ the idea. “The world has changed in three or four years,” he concluded.