Toshio Nakama and the value of finding out early

A prototype has an awkward job. It must behave enough like the finished thing to reveal what is wrong with it. Nobody buys one for its elegance at a dinner party. Its finest hour may be the discovery that a very expensive idea needs changing. Toshio Nakama has spent much of his career making that discovery easier to arrange.

His subject is the chip before the chip: a design that engineers can run, examine, and connect to software before committing it to manufactured silicon. As a co-founder and longtime chief executive of S2C, Nakama worked in the less photogenic part of the semiconductor business. The finished gadget gets the applause. The equipment that helped its designers catch a mistake gets another assignment.

There is a practical ambition running through his work. More engineers should be able to use a prototype. They should be able to share it, change it, and learn from it while there is still time to act. Following that ambition takes us from a university circuit-design class to San Jose, Shanghai, and the decidedly unromantic business of making difficult tools usable.

A circuit class with consequences

Nakama encountered field-programmable gate arrays, or FPGAs, in a digital circuit design course at Cornell. Their programmability caught his attention. He earned a bachelor's degree in electrical engineering there and later an executive MBA at China Europe International Business School. The combination suits a career spent moving between the behavior of circuits and the behavior of customers.

He began at Altera in 1997 as a field applications engineer, then worked at Aptix between 1998 and 2003 in technical and sales management. A field applications engineer helps customers put technology to work. It is a useful education in the distance between something that can operate and something another person can successfully operate.

In 2003, Nakama co-founded S2C in Silicon Valley. His account of the beginning is refreshingly concrete: former Aptix colleagues and a finance professional pooled their money. In 2004, the company established research, development, and manufacturing teams in Shanghai. Its opportunity lay partly with emerging Asian design businesses and the regional operations of American and European firms.

The board needed a supporting cast

The appeal of FPGA prototyping is that programmable devices can stand in for circuitry still under development. S2C's early task was to make that approach work as a coherent process. A board alone left too much work with the customer. Software, interfaces, and a path back into the wider chip-design process all mattered.

At the Design Automation Conference in 2005, S2C introduced IP Porter. It used four Xilinx FPGAs and supported designs of up to three million gates. Compiler and navigation software accompanied the hardware. The system let a prototype communicate with its host and with simulation tools, giving engineers ways to control and investigate the design.

Early users estimated savings of three to six months in design time. That was a customer estimate for an early product, rather than a promise for every project. The interesting detail was where the savings came from: reducing the time spent making and testing revisions. Finding a problem is only helpful if the next version can arrive before everyone's patience retires.

That distinction helps explain Nakama's sustained attention to the machinery around a prototype. The first successful run matters. The tenth revised run can matter just as much. A tool earns its place by helping engineers move through that cycle repeatedly, with fewer chores between the question and the answer.

One prototype per desk?

By 2013, Nakama was making an argument about access in unusually tangible terms. Ideally, he said, a project manager would give every hardware and software engineer a prototyping system. High-end systems costing $100,000 or more made that difficult. The budget could turn a useful engineering instrument into a scarce office possession.

S2C's K7 TAI Logic Module addressed that problem with a smaller platform. At launch, the company described its cost as less than a tenth of a high-end system's, with capacity up to 4.1 million ASIC gates. Teams could use it for individual blocks, software work, and other tasks that did not require the largest available prototype.

There is a recognizable management instinct here: fit the equipment to the work. Giving a smaller task a suitable platform leaves other resources available. Nakama's argument connected price to scheduling. If more people could work in parallel, hardware verification and software development could advance together instead of queuing politely behind one another.

The prototype leaves the room

The next step enlarged the unit of organization from a desk to an enterprise. In October 2014, S2C announced a $4.6 million Series C financing led by GVT Fund and Industrial Technology Investment Corp. The stated uses included expanding sales and support and developing prototyping as a cloud-based resource.

Cloud Cube followed in May 2015. Its chassis could hold up to 32 FPGAs, with power, cooling, and remote control built in. S2C specified capacity of up to 1.4 billion gates. The system supported multiple users and worked with the company's partitioning software, simulation connections, and interface-card library.

By 2017, Nakama was explaining the organizational reason for shared equipment. Design and verification teams were dispersed. Their members could be working in different offices on different designs. A resource-sharing system therefore had to accommodate several projects, rather than merely offer a remote view of one machine doing one job.

His own working geography reflected that dispersed business: he divided his time between San Jose and Shanghai. The company's product direction and his calendar occupied similar territory. A prototype connected people as well as circuitry, and the people were increasingly unlikely to be sitting beside it.

Growth required someone nearby

Remote access did not remove the need for local help. S2C opened a Beijing office in 2006 to support demand for its TAI Logic Modules. The location, in Zhongguancun, extended the company's presence beyond Shanghai. Nakama linked the decision directly to supporting an expanding customer base.

In 2014, the company opened direct offices in Japan and Korea. Its Japanese operation, led by Tomoyuki Tsuji in Shin-Yokohama, followed work with more than 20 customers through distributors. Nakama emphasized closer collaboration on system-level methods and product support. In Seoul, S2C appointed Suk-Ha Lee, an experienced Korean EDA sales and support executive.

These were decisions about proximity. A complex tool needs explanations, adjustments, and people who understand the customer's work. In both announcements, Nakama spoke about the work engineers were trying to complete. The expanding map mattered because the support had to reach the same places as the equipment.

Making the complicated parts less troublesome

Some of S2C's consequential improvements sound thoroughly administrative. In June 2016, Player Pro 6.3 added partitioning and multi-FPGA debugging capabilities. Partitioning divides a design across programmable devices. Debugging helps engineers inspect what that divided design is doing. Both become central when the prototype no longer fits inside one FPGA.

A May 2017 update introduced a new graphical configuration cockpit and extended support to Arria 10 modules. Guided partitioning became available to Logic Module users without a separate compile license. Nakama kept the promise attached to the user. A modest objective, with a considerable amount of engineering underneath it.

“The updates to our Player Pro system will help our users be more efficient.”

Toshio Nakama · 2017

In September 2022, S2C introduced Neuro, browser-based software for managing prototyping resources. Its functions included user permissions, project allocation, remote hardware controls, and reports. Engineers could check equipment status and run diagnostics. The company was addressing an installed base with more prototypes and more complex arrangements to manage.

The vocabulary had moved from individual circuits to users, groups, permissions, and utilization. That progression is familiar to anyone who has watched a useful tool become shared infrastructure. Once several teams depend on it, knowing who can use it and whether it is working becomes part of the engineering problem.

Asking whether the specification is right

Another thread in Nakama's thinking reaches earlier than implementation. In 2020, S2C and Mirabilis Design announced a collaboration connecting VisualSim architecture exploration with S2C's Prodigy Logic System. An FPGA prototype could serve as a functional part of a larger system model, reusing available design blocks rather than recreating everything in a new model.

Nakama's concern was specification errors: a circuit could be implemented carefully and still answer the wrong requirements. Exploring the system earlier gave designers a chance to examine those decisions. Deepak Shankar, Mirabilis Design's founder, described reuse as a way to reduce the effort of modeling custom blocks. The partnership joined different tools around a shared question.

A conceptual view of the workflow, not a measured schedule.

Twenty years, still moving the test earlier

Looking back on 2023, Nakama singled out S2C's twentieth anniversary. In his February 2024 outlook, he discussed chiplets, AI, RISC-V, and the Arm ecosystem. He also emphasized software development alongside hardware. The applications had changed; the familiar scheduling problem remained.

He described a broader product approach spanning simulation, emulation, and debugging, and continued participation in DAC, DVCon, and RISC-V events. The founder who had helped build a prototyping business was now discussing how several verification methods could work together. Getting an answer earlier still required choosing the right way to ask.

Nakama's career offers an unusually practical view of technological progress. A programmable device needs software. Software needs a usable interface. A shared machine needs allocation. A customer needs someone who can help. Each addition makes the original promise a little more available to the person doing the work.

The finished chip eventually receives its name, its launch, and its place in a product. Long before that comes the less ceremonial moment when an engineer can try something and discover a mistake. Nakama spent years building a business around making room for that moment. It is a sensible place to invest one's attention. Mistakes tend to become more expensive once they acquire a physical address.