Data Dividends In An Era Of Hyperscale Computing

By Russ Banham | Research by Luc Alper-Leroux


Five disruptive technologies are poised to reshape industries as diverse as aerospace and defense, manufacturing, healthcare and automotive—promising cutting-edge products that have the potential to make an impact on our daily lives. Each of these technologies requires semiconductors to power and process increasingly gigantic volumes of data.

The five transformational technologies are hyperscale computing, 5G communications, the industrial internet of things (IIoT), artificial intelligence/machine learning (AI/ML) and autonomous vehicles. And all of them are progressing simultaneously, requiring higher-performing semiconductors with greater capacity and speed.

For one thing, the volume of data is increasing faster than predicted. Experts report that 90% of the world’s data was generated in just the last two years. The bulk was produced by social media and streaming content, smart home and wearable devices, digital photos and videos, business information and online shopping.

According to a Forbes article using insights from International Data Corporation (IDC), from 2010 to 2020, the amount of data produced across the world increased from 1.2 trillion gigabytes to 59 trillion gigabytes—a near 5,000% growth rate. What used to be called “big data” is being redefined. Thanks to hyperscale computing and the other four disruptive technologies, today’s data mountain may seem like a foothill in just three years, when more data will be created than has been consumed over the past 30 years, according to IDC.

“Just because you have all this data doesn’t mean it automatically leads to intelligence,” said Anirudh Devgan, president of Cadence, a global leader in electronic design of semiconductor chips and systems. “You have to capture the right data and use sophisticated analytical platforms to make these five technologies perform optimally. And to do that, you need to design smaller semiconductors embedded with billions and billions of transistors.”

The Move Toward Application-Specific Semiconductors

Today, Cadence is using advanced computing software, known as computational software, to manage and ultimately guide the design of semiconductors. That’s good news for the giant cloud computing companies that are networking tens of thousands of servers together to run their businesses and support the massive data and analytics needs of customers. 

These scalable cloud computing infrastructures need extremely high-performing semiconductor chips at a time when they’ve never been in greater demand. In addition to traditional markets like smartphone makers, vertical industries like automotive, healthcare and industrial, as well as horizontal markets like AI/ML and hyperscale computing, are all demanding more innovative and powerful chips. There’s so much demand that there are chip shortages in some cases, notably in the automotive industry.

To fulfill these needs and competitively differentiate products and services, some tech giants are starting to manufacture their own semiconductors to develop self-driving electric cars, smartphones, connected factories and supersized data centers. 

It’s a big trend, said Devgan, with “companies wanting chips customized to their particular applications and solutions. Other factors include economic considerations like cost and availability. But the biggest factor is performance—blazingly fast semiconductors drive power-hungry technologies like autonomous vehicles, 5G, hyperscale computing and AI/ML.”

Cadence’s value proposition is to assist these aims, leveraging the company’s expertise in computational software to design next-generation chips, fulfilling each market’s unique application and solution needs. “The demand for what we do is huge and growing,” said Devgan.

This steady growth is in contrast with the semiconductor industry’s past “boom and bust” cycle, said Devgan. “The five disruptive technologies have changed the dynamic to where we will continue to see robust demand for chips through the foreseeable future.”

Computational Software Basics

To satisfy the demand, smaller and higher-performing chips using less power must be designed. In this work, Cadence’s engineers develop advanced computational software, which uses a superset of algorithms to understand the physical characteristics of semiconductors needed to power a variety of solutions. In short, the algorithms enable chip designs that meet the requirements of the semiconductor manufacturer.

“We’re advancing the state of the art of electronic design of our partnering organizations to five to 10 times faster than what was previously available,” Devgan said. “We’re fortunate to have both the computational software and skilled Ph.D.-level mathematicians trained to develop this software, helping us meet the growing demand for increased excellence in semiconductor and system development.”

As this takes shape, the five disruptive technologies will inch closer toward their full potential, ideally making businesses more efficient and manageable and lives more pleasurable and safer. Scientists are expected to leverage ML-based deep neural networks to create lifesaving drugs, smart factory equipment may begin to produce goods based on real-time demand and everyday people could even stream movies as their autonomous vehicles drive them to work. As Devgan put it, “These are the dividends of computational software.”

Russ Banham is a Pulitzer-nominated financial journalist and bestselling author.

Leave a Reply