A chipmaker’s challenge: building resilience in a fast-changing world
Michael Doherty, infrastructure head at semiconductor giant Renesas, on creating a tech stack to support a global strategy
In recent years the semiconductor industry has experienced a series of seismic shocks, from the massive upswing in demand for GPUs to power GenAI, to sanctions, tariffs and trade wars, to the supply chain snarl-ups caused by Covid. The latter was a huge wake-up call, said Michael Doherty, senior director, head of global infrastructure at Japanese chipmaker Renesas.
“The industry as a whole is still recovering from the shock of Covid.”
As the supply of raw materials dried up and factories closed their doors, those contract manufacturers that were still operating found themselves massively overbooked, causing major backlogs in production. As a result, chipmakers had to rapidly reconsider their strategy, diversifying their supply chains. This policy remains firmly in place to this day, Doherty told Computing.
“Renesas tries to be global in its strategy, rather than having one based in one country. That's been a big, big drive from our chief executive.”
Being global means maintaining production capacity in multiple territories and independence from single suppliers or jurisdictions, he explained.
“You have to have dual sources for whatever you're doing. Sometimes we'll be asked by our customers to make sure that if we're manufacturing something, we can manufacture it in more than one location.”
Growth by acquisition
Renesas must also keep up with the ever-shifting demands of the automotive sector (its biggest customer), as well as developments in AI, mobile, laptops, Wi-Fi, IoT and other markets that it serves. Its strategy is growth by acquisition, snapping up fabless design startups and boutique manufacturers around the world — along with the specialised skills and knowledge they contain — to stay ahead of the curve.
The company is now in the top three semiconductor manufacturers in the world, recently opening a major design centre in current silicon hotspot Vietnam.
Operating in one the globe’s most geopolitically strategic and intensely competitive sectors, the company’s core IP is a fiercely protected secret. Naturally, then, it is kept in house within its multiple datacentres around the world, rather than in the cloud.
It’s also no surprise that with premises all around the world, multiple acquisitions to integrate, and a diverse set of activities, Renesas operates across multiple clouds.
‘It’s quite a drag managing three clouds’
The first cloud it adopted was AWS, back in the days when Amazon was the only serious option. AWS is still the cloud provider most in use at Renesas, particularly for the development of SaaS products for customers. Azure is the cloud of choice for virtual desktops, “so we don't have to give sub-contractors a laptop anymore or worry about VPNs.” Microsoft’s infrastructure also supports the company’s Modern Data Platform, driven from a hosted Databricks data lakehouse. Meanwhile, Oracle Cloud is used mainly for storing manufacturing-related data and some software development. This choice was initially driven by Renesas' use of Oracle Exadata appliances, which were then only available virtually on Oracle's cloud.
“It’s quite a drag managing three clouds, the overhead of it, because they're all just different platforms at the end of the day,” commented Doherty.
The administrative burden could increase still further as Renesas investigates other cloud providers where it has a presence, such as Alibaba in China.
The company is currently looking into Nutanix NC2 to ease the management overhead, reduce the need for cloud-specific skills, and also increase choice of instance configurations. “You can optimise that cloud infrastructure even better if you're using the bare metal services, because you don't have to take the standard shapes that the cloud providers offer.”
As far as protecting the precious IP on premises goes: “That’ll be Nutanix or NetApp, with IPLM [Innovation and Product Lifecycle Management] tools on top - they're enforcing business rules out of that IP to allow where it can be copied to where it can and who can access it.”
Managing that IP in a can be a challenge for chipmakers, in that silicon chips are modular, built up of component parts like Lego, and from time to time individual modules can find themselves subject to trade restrictions, meaning a brick is missing. An important management task, therefore, is ensuring that certain data can only be resident in particular countries while maximising its availability.
A HCI early adopter
Renesas was an early adopter of hyperconverged infrastructure (HCI), in an effort to simplify the tech stack. Its journey began in 2014 with SimpliVity, which was later purchased by HP. “It was a very good product, but it did tie you into hardware,” Doherty noted. “When things failed, it's not off-the-shelf hardware and it took some time to get fixed.
“We knew we liked HCI. It was the ease of implementation, taking us away from the traditional three-tier model. It was great that we could have a team that could do things very quickly, rapidly.”
When the HP contract came to an end the company decided to shift to something hardware agnostic, landing on Nutanix AHV, after first considering VMware. Renesas remains a major VMware customer, but has embarked on an 18-month programme to replace it as contracts end.
“VMware was king for so long,” said Doherty. “But here were lots of modules that you had to licence separately. It was a very confusing model. Licensing decisions, particularly around virtual machine software, have always been a complex subject.”
Indeed, over and above potential cost savings, licensing issues were the main factor that swung the HCI decision in favour of Nutanix.
“Nutanix was a really straightforward. You knew exactly where you stood. You knew if you needed to do something, then you didn't have to go and find a specialist licensing consultant to ensure that you've made all the right decisions. It was just easy to understand.”
AI document search
Renesas is looking to AI to help rationalise aspects of its operations. The company has set up a global AI group as a centre of excellence for AI development. It’s first tool, still in development, is a system to pore over the reams of manufacturing and product documentation spread all over the world to allow people to better understand the company’s vast repositories of knowledge.
“We're creating vector databases and using RAG type algorithms to build some applications to search documentation.” Doherty said. “They're looking at using LLM models that are there today, rather than trying to build something from scratch.”
In the future he sees great potential for AI to identify problems in the semiconductor manufacturing process to increase efficiency and reduce waste.
"You never get 100% yield of what you manufacture, but you want to get that as close to 100% as you can. What we're thinking AI will allow us to identify problems in the manufacturing process... trace it back through the systems and find out, what was it? Was it a person on a day, you know, did something wrong? Was it a machine that was failing?”
Are you a standout in UK IT? Computing’s IT Leaders 100 names, recognises and honours the most influential and inspiring people shaping the sector, and it’s open for entries from all end-user IT leaders now.
Whether you’ve spearheaded major change, tackled complex security threats or built a team that delivers real results, we want to hear your story. Click here to enter before 30th April – for free.