Message from The Chairman
Research and Development at RaaS
We provide semiconductors
as a service, rather than
sell them as a product.
Prof. Tadahiro Kuroda
Director of Systems Design Lab
Graduate School of Engineering
University of Tokyo
A game-changing movement is sweeping the semiconductor industry – the world has embarked on the development of in-house specialized chips. It is time to revisit your semiconductor strategy.
Until now, general-purpose chips have dominated the semiconductor industry. That was the consequence of adopting the von Neumann computer architecture. As part of the growth strategy for computers, general-purpose processors and memory chips are manufactured in high volume in order to encourage widespread adoption. The adaptation for various applications is achieved through software.
A general-purpose chips business cycle is started by device innovation, with the eventual winning supplier decided through capital investment competition. While Japan won in device innovation, it lost the competition in capital investment.
Specialized chips have had its successful run too in the form of ASICs. In the 1980’s, with the creation of CAD technology, the design tools industry was born. In addition, a semi-custom manufacturing process was developed where a semi-finished chip was first made like a standard product, then tailored towards different applications by customizing the interconnect layers.
Using these design methodology innovations, chip development efficiency was improved by three orders-of-magnitude in total, making ASIC a profitable business.
Nevertheless, with Moore’s Law increasing integration density by three orders-of-magnitude in 15 years, even with computer-aided design, it took more man-hours than ever to develop specialized chips. This contributed to the profit erosion of the ASIC business which eventually resulted in its demise.
This illustrates how a specialized chips cycle is started by design methodology innovations and ended by Moore’s Law.
Today, specialized chips are experiencing a resurgence, driven by the unique “energy crisis” of a data society. The energy crisis is being accelerated by the explosive growth in data volume coupled with the increasing sophistication of AI processing.
At the current pace, by 2030 IT equipment alone will consume about 2 times the total power consumed today; by 2050 the ratio will jump to 200 times. We have no possibility of achieving a sustainable future if digital transformation consumes so much energy as to destroy earth’s environment.
Under this current situation, only those who can improve energy efficiency 10-fold can achieve 10-fold increase in computing performance, or 10-fold increase in smartphone battery life. Compared to general-purpose chips which can be used to perform any function, specialized chips can achieve a more than 10-fold increase in energy efficiency by eliminating unnecessary circuits. That is why specialized chips are being sought after.
In addition, since AI processing is adopting neural networks which process data in a parallel fashion, it is difficult to achieve high performance by implementing it in a von Neumann architecture that is built for sequential processing. As a result, AI accelerators in the form of specialized chips are being developed all over the world. Furthermore, the slowing of Moore’s Law is providing a tailwind for the resurgence of specialized chips.
Unfortunately, specialized chips development is not easy and hence not for everyone. The number of transistors integrated on a chip is exceeding the population of the world. Development costs have been rising rapidly and are approaching $100 million. Even with a team of several hundred designers, development can take several years. That is unacceptable given how fast technology advances nowadays.
With software, even if there are bugs, they can be patched post-delivery. By contrast, hardware must be perfect before it is shipped. As a result, hardware design is more difficult and carries higher risk than software. It is indeed “hard.”
Software development is facilitated by the use of compilers. If similar technology - namely, silicon compiler technology - is available for chip development, one can expect both the costs of hardware development and its risk to drop. This will in turn enable more hardware designers. If the culture of open source takes roots in the hardware design community, and its ecosystem hierarchy develops and expands, mass collaboration will become possible. When that happens, we will be able to create chips like the way we write software.
As Alan Kay pointed out, “People who are really serious about software should make their own hardware.” System development requires both hardware and software development.
Our mission at RaaS is to democratize access to silicon technology. We aim to create a silicon compiler to enable designing chips like writing software, and construct a development platform to implement agile authentic prototypes that drastically cuts prototyping time.
Our technology goal is a 10-fold increase in both development and energy efficiencies. To improve development efficiency, we will create an agile design platform and deploy open architecture. To improve energy efficiency, we will fabricate our chips using advanced CMOS processes and integrate them in 3D.
We aim to provide semiconductors as a service, rather than sell them as a product. At RaaS, we are developing the enabling technologies.
Overview of RaaS
Research Association for Advanced Systems
Our technology goal is a 10-fold increase in both development and energy efficiencies.
To improve development efficiency, we will create an agile design platform and deploy open architecture.
To improve energy efficiency, we will fabricate our chips using advanced CMOS processes and integrate them in 3D.
The users will be able to design their own specialized chips the same way they write software and use them to realize digital transformation.
- Democratization of Access to Silicon Technology
- Quick-turn Prototyping (Mantra: Agile Authentic Prototyping)
- Silicon Compiler (Disruptive Technology to Enable Designing Chips Like Writing Software)