How Intel forgot innovation?

For decades, Intel was the undisputed leader in the semiconductor world, a symbol of American technological prowess. The microprocessors they invented, along with the famous x86 architecture, were the lifeblood of everything from PCs to data centers.

By 2010, even though the PC market had slowed down, the business was an endless gold mine for Intel. The company was making billions of dollars in profit every year. They spent more than $10 billion a year on research, which was four times what TSMC did.

It was at a time when the semiconductor industry entered a new era of EUV (Extreme Ultraviolet) technology. Interestingly, Intel founder Andy Grove invested $200 million in EUV in the 1990s. Later, a large part of the investment in implementing this technology came from Intel. But unfortunately, Intel could not utilize that technological advantage.

Intel’s decline is due to three main factors: first, its failure to identify and adapt to the AI revolution; second, its failure to adapt to new chip architectures; and third, a series of setbacks in its own manufacturing process.

Intel remains one of America’s most modern and profitable chipmakers. But its future is more uncertain than ever. This instability is not just a question of a company’s, but also of America’s technological security.

In the 2010s, when rivals AMD and others were selling off their factories and focusing solely on chip design, Intel clung to its old “integrated model,” where design and manufacturing were done by the same company. Intel claimed that this created a perfect match between design and manufacturing.

But this model also had a major weakness. TSMC, which made chips for many companies, had about three times the manufacturing capacity of Intel. More production meant more experience, faster learning opportunities, and more refined technology. While Intel considered new startups a competitor, TSMC saw them as potential customers. TSMC's goal was to 'make the world's best chips at the lowest cost'. On the other hand, Intel fell behind on both fronts when it came to handling the two difficult tasks of design and manufacturing.

Intel's biggest strategic mistake was not realizing in time that artificial intelligence (AI) would completely change the world of computing. In early 2010, Intel had a monopoly on the data center server market. The clouds of tech giants like Amazon, Google, and Microsoft were dependent on Intel's CPUs.

But just as Intel was happy to capture this market, the demand for computing started to grow. In the case of AI, there is a lot of data and the same calculations have to be done repeatedly. But Intel's main product CPU was not suitable for this task at all. CPUs work in a ‘serial processing’ mode, which is slow and expensive to train AI models.

On the other hand, Nvidia’s GPUs, which are designed for video graphics, were used by Stanford researchers for AI training. Since GPUs could perform many calculations at the same time through parallel processing, the speed of AI work increased dramatically. Nvidia quickly transformed into an AI-centric company. They left TSMC to make chips and focused only on design and software development, and very quickly became the most valuable chip company in America.

Meanwhile, Google, Amazon, and Microsoft - all began to bypass Intel and start making their own custom AI chips (such as Google’s TPU). This change ended Intel’s data-center dominance.

In mid-2010, Intel also tried to get into the foundry business (making chips for other companies) by imitating TSMC. But Intel’s institutional culture was not ready for it. TSMC was open and customer-friendly; Intel wanted to force customers to play by its own rules. TSMC did not compete with customers; Intel competed with almost everyone. Naturally, Intel’s foundry business failed to gain customer trust. As a result, the venture was shut down within a few years.

When Intel turned 50 in 2018, the company was very weak inside - innovation had declined, bureaucracy had increased, and its share price was falling. The biggest failure was its consistent delay in introducing advanced manufacturing technology. Intel was the last to use the EUV technology that Intel had funded for decades. By 2020, TSMC owned half of the world’s EUV machines, and Intel had yet to launch EUV.

By the end of the decade, the only countries with the capacity to make the most advanced chips in the world were Taiwan (TSMC) and South Korea (Samsung), which are geographically very close to China. As a result, the United States has become dependent on foreign countries for the production of its most advanced chips, which is a huge risk to the country's technological security.

(Adapted and abridged from Chris Miller's groundbreaking book, "Chip Wars," Chapter 41, "How Intel Forgot Innovation")

Author: Mahmud Hossain, a graduate of BUET, has led Bangladesh’s telecom and ICT sector for more than three decades. He played a vital role during the introduction of mobile technology in the country. Currently, he serves as a Commissioner at BTRC, having previously held senior positions in several leading local and international organizations.