IBM recently announced a partnership with Rapidus, an evolving Japanese microprocessor manufacturer, to jointly develop two nanometer gate-all-around nanosheet technology (Rapidus will jointly develop this technology and take it to manufacturing). This news is important for several reasons:
- The global semiconductor industry needs to diversify. (Chip shortages over the past few years have caused manufacturing problems that affect global supply chains). The world needs more sources of supply – and this IBM/Rapidus partnership recognizes the need for global expansion and will help address future supply problems;
- In the future, two nanometer technology will become the basis for a multitude of devices (computer systems, appliances, automobiles, etc.) – in other words, two nanometer technology – that now offers 45% better performance using 75% less power as compared to five nanometer technology – will become a high-growth market that will need to be served by a large number of specialized chips suppliers;
- Geopolitical concerns have jeopardized microprocessor manufacture, causing nations concern about the stability of their current sources of microprocessor supply. Accordingly, several countries are moving to bring microprocessor manufacture closer to home.
A closer look
IBM has invested billions of dollars in microprocessor research over the past decade, and delivered new processing technologies (the new gate-all-around nanosheets, silicon on insulators; etc.); and new functions (such as artificial intelligence technology and security at the chip level). That said, it is not in IBM’s best interest to corner the market on the manufacture and distribution of its microprocessors (other companies may be able to deliver chips less expensively; may create value-add extensions; and can broaden IBM’s market exposure through other channels). Hence, partnerships such as this Rapidus relationship, where Rapidus takes on the manufacturing (foundry) process, and adds value through customization, works well for both parties.
It is important to note that as part of IBM’s investment in microprocessor research, the company has:
Committed to spend more than $2 billion dollars to grow its high-tech footprint in New York state (including the establishment of the IBM Research AI Hardware Center at the Albany NanoTech Complex for artificial intelligence-focused microprocessor research) and will develop protypes, conduct testing, and perform simulation on its microprocessors. The facility spans five buildings and includes more than 100,000 square feet of semiconductor fabrication space
For its part, Rapidus is expected to spend hundreds of millions of dollars building its pilot and production foundry facilities. The capital for this investment will come from Japan’s METI (Ministry of Economy, Trade and Industry), as well as from eight partner industries (including Toyota, Sony, Denso, MUFG, KIOXIA, NEC, Softbank Group and NTT Group).
What this means to future semiconductor markets
In short, this AI Hardware Center will become the nucleus for AI-focused microprocessor development. IBM has already partnered with over a dozen companies to extend its base microprocessor technologies. Expect them to partner with more firms and to help broaden the market for IBM microprocessor solutions. It should be noted that, domestically, IBM is working with the American Semiconductor Innovation Coalition (a group of 160+ organizations dedicated to strengthening American leadership in semiconductor innovation and manufacturing) under the auspices of the National Semiconductor Technology Center (NSTC).
Summary observations
Over the course of my research analyst career (now 28 years) my company’s name has been regularly associated with computer hardware. My firm still writes about microprocessor architectures (link, link), storage hardware (link, link, link), servers (link, link), and hybrid cloud architecture (link, link, link). The funny thing about this “hardware association” is that many vendors and buyers consider hardware to be inconsequential. And many vendors tend to redirect conversations and briefings toward software solutions (look what we can do in system management; look at our financial/banking/insurance, healthcare solutions; …). And to a certain degree vendors are correct in redirecting the conversation toward solutions because buyers buy solutions – hardware is just a platform to deliver those solutions. But, to a certain extent, vendors (and buyers) who don’t understand the differences that microprocessor technologies make in terms of performance, execution, efficiency and cost are missing the boat. Microprocessors are the heart of a given solution. Failing to pay attention to microprocessor characteristics can result in greatly increased computing expense (by choosing the wrong microprocessor for the job); energy waste (power consumption, cooling, etc.); and increased deployment and management complexity.
As for this IBM/Rapidus partnership, expect that by helping to bring advanced microprocessor manufacture to Japan, geopolitical concerns will be mitigated. Expect this partnership to yield more functionality, better performance and more power efficiency. And expect that with many sources of supply, microprocessor supply chain shortages will be addressed.