Rapidus, a semiconductor consortium backed by the Japanese government and conglomerates, plans to start mass-producing chips on a 2nm manufacturing process in 2027. To ensure full utilization of its production capacity, Rapidus hopes to win orders from at least one global company that effectively competes with TSMC and other foundries. But at the same time, Rapidus does not plan to become like TSMC.
Developing the latest manufacturing technology costs billions of dollars, and while IBM, which is backing Rapidus with its 2nm manufacturing process, has done a lot of the heavy lifting around materials research and other time-consuming tasks, Rapidus will still have a lot to do. Rapidus estimates that it will need about $35 billion to start manufacturing 2nm chips in 2027, as it costs a lot to install top-notch equipment in factories.
To recoup the R&D and manufacturing costs of production nodes would require producing large numbers of chips at cutting-edge nodes, but Japanese companies may not create significant demand for such parts. Therefore, Rapidus needs him to win orders from multinational companies like Apple and AWS.
“We are looking for a US partner and have started talks with some GAFAM.” [Google, Apple, Facebook, Amazon and Microsoft] said Atsuyoshi Koike, CEO of Rapidus, in an interview. Nikkei newspaper. “There is a particular need for [for chips] from the data center [and] At the moment, TSMC is the only company that can manufacture the semiconductors they envision. That’s where Rapidus comes in. ”
In fact, recently there has been a surge in the number of companies developing custom chips for data centers. Amazon and Google have developed a number of their own designs to serve their specific needs, and Apple, of course, has dozens of his SoCs and chips powering their own consumer electronics. All of these companies use TSMC because it has the most competitive process technology and can leverage their IP across various product lines.
Apparently, Rapidus would like to get an order from at least one of these companies, but it’s not an impossible endeavor. In addition to conglomerates like Apple and Google, there are other hyperscalers who need custom silicon, and the former may choose Rapidus over TSMC or Samsung Foundry if it can offer competitive production technology, good yields and competitive pricing.
Interestingly, Rapidus isn’t looking to serve dozens, it’s looking to serve 5-10 customers.
“Our business model is not TSMC’s business model of manufacturing for every customer,” Koike said. “We will start with five companies at most, and eventually increase to 10 companies, and we will consider whether to increase it further.”
It remains to be seen whether 5-10 companies will generate enough demand to recoup the tens of billions of dollars Rapidus will need to invest to start 2nm production in 2027. On the other hand, the number of companies willing to invest in designs done at leading-edge nodes is fairly limited, so in 2027 he will have a pretty tough time getting even five customers with high volume orders for 2nm.
Again, from the Japanese government’s perspective, Rapidus aims to revitalize the country’s cutting-edge semiconductor supply chain. So even if the company’s 2nm node isn’t a triumph, it will pave the way for successors and open new doors for local chip designers.
