The fabless revolution

There was once a saying in Silicon Valley: "Real Men Have Fabs" - that is, real semiconductor companies are those that have their own fabrication ('fab') plants or factories. But in the late 1980s, this idea changed completely with the help of a new generation. These entrepreneurs used to design chips themselves, but outsource the manufacturing work, whose main partners were companies like TSMC. This newly born business model is called the 'fabless model'.

In 1984, Gordon Campbell and Dado Banatao founded 'Chips and Technologies', which is considered the first fabless company. Many people hesitated to call it a semiconductor company at the time, because they did not have their own fab. However, the graphics chips designed by this company gained great popularity in the market, which even started to compete with the products of large companies. Later, Intel bought the company. But by then they had proven that a good business idea and a few million dollars in capital were needed to succeed in the fabless model.

At that time, computer graphics had become an attractive field for semiconductor startups. Because while Intel and AMD dominated the microprocessor market, the world of graphics chips was not under anyone's control. With the emergence of new foundries and the reduction in startup costs, ordinary entrepreneurs were also getting a chance to enter the competition.

It was at this crucial time that three young engineers - Chris Malachowski, Curtis Priem, and Jensen Huang - gave birth to a company called Nvidia in 1993. Among them, Jensen Huang later became the most recognizable face and took over the responsibility of the company's CEO. Born in Taiwan, Huang moved to America as a child. He always wore black jeans, a black shirt, and a leather jacket and was known as a visionary about future technology, like Steve Jobs.

Nvidia started out making chips for video and gaming companies. At that time, the PC world was almost entirely two-dimensional (2D). Nvidia realized that the future of graphics would be three-dimensional (3D). They created the Graphics Processing Unit (GPU)—a type of chip that could run thousands of calculations in parallel and was suitable for creating images. Their GPUs began to be used not only for graphics, but also for parallel calculations. In this way, Nvidia discovered a huge new market, and this GPU quickly became the basis of AI or artificial intelligence.

Today, Nvidia's chips are used in almost all large data centers, and TSMC produces them. If Nvidia had to build its own fab, it would not have been possible to move so quickly, because it costs hundreds of millions of dollars to build a fab.

Meanwhile, another inventor, Erwin Jacobs, was thinking that the future of microprocessors was now in mobile phones. He founded Qualcomm in 1985, which took its name from Quality Communications. At that time, mobile phones meant big, heavy, car-mounted sets. Everyone was working on 2G technology.

But Jacobs proposed a different technology, CDMA, where call data would be sent repeatedly over different frequencies, making it possible to make many calls at once. Many thought it was practically impossible. Jacobs went out into the field to prove it, setting up a small network with a few cell towers and showing that it worked. This is how CDMA technology was born which later spread around the world.

Qualcomm has brought new innovations to each generation of mobile technology, creating chips that make more efficient use of the radio spectrum, without which modern smartphones are impossible. They design their own chips, but do not manufacture them - that work is done by foundries like TSMC or Samsung. Qualcomm's patents are so important that no phone can be made without their technology.

Many say that the move of semiconductor production offshore is a loss for America. But companies like Qualcomm or Nvidia wouldn’t have survived if they had to spend billions of dollars building fabs. They were able to focus on their core competencies – design, algorithms, and innovation – because the manufacturing was being done by others.

The ‘fabless model’ was not limited to these two companies. Many other US companies also relied on foundries. Ultimately, this model didn’t just change manufacturing, it changed the entire world of computing. Mobile phones, advanced graphics, and artificial intelligence – all of this is the fabless revolution, where talent and innovation were at the center, and factories were in the hands of others.

(Adapted and abridged from Chris Miller’s groundbreaking book ‘Chip Wars’, Chapter 36, “The Fabless Revolution”)

Author

Mahmud Hossain, a BUET graduate, has over three decades of leadership experience in Bangladesh’s telecom and ICT sectors. He played a key role in introducing mobile technologies in the country. He now serves as a Commissioner at BTRC, following senior leadership roles in several national and multinational industry-leading companies.