Your browser is not supported. Please update it.

16 March 2021

Apple funds 1bn chip R&D plant in Germany, a big step to silicon self-sufficiency

Apple’s journey towards self-sufficiency in silicon has been proceeding in fits and starts for 15 years but is accelerating and has taken a major leap forward with the announcement the company is building a €1bn European Center for Chip Design in Munich, Germany over the next three years. This will become Apple’s largest semiconductor facility in Europe’s and the continent’s biggest dedicated to wireless.

Occupying 30,000 square meters and hosting its silicon engineering teams it is the latest in a series of moves geared towards ending Apple’s dependence for mobile chipsets on Qualcomm, with which it has shared a tempestuous and litigious relationship over the last few years. Qualcomm does still hold a few aces in its ownership of key IP relating to LTE and 5G, but Apple seems to think that by investing heavily from its cash mountain, those advantages can be steadily eroded by technological inflation as processes advance.

That is more or less how Apple CEO Tim Cook put it himself when speaking about the new German factory. “I couldn’t be more curious about what our engineering teams in Munich will discover – from exploring new possibilities in 5G technology to a new generation of technologies that will enable even more performance, speed and connectivity,” he said.

Apple, of course, is one of the very few companies with the resources to reach for self-sufficiency across almost its entire ecosystem, and that has really been its ambition ever since the first Mac was launched in 1984. The refusal to allow other manufacturers to provide the competition and choice available within the alternative Windows based PC ecosystem, and then much more recently the Android platform, meant it was always going for the higher end of the market, relying on superlative marketing and branding to ensure that enough consumers would be hooked.

Now in the 5G era it has become clear that having control over the chips and underlying processes can reap great dividends as they become more critical for differentiation, cost and ability to innovate around new use cases. The writing has been on that wall for a while and Apple’s trajectory towards chip self-sufficiency dates back before 5G, naturally enough to the launch of the first iPhone, followed by the iPad.

After that stage, it seems Apple is aiming to cut Qualcomm out of its loop with its homegrown 5G modem that will give it full control over the iPhone’s guts, including the so-called neural engine serving as the basis for AI based innovations. This is Apple’s most significant step in processor design since that launch of the A4 just over a decade ago.

The motivation is primarily control over engineering and innovation, but also commercial for self-sufficiency, although not for direct revenue generation. It does give tighter control over costs but the greater prize is freedom from the IP of others.

The saga has been a long one. In the early days of its mobile devices, Apple aligned behind Intel processors for Mac laptops and ARM design for the iPhones, which it had first adopted back in 1993 for a long-forgotten personal digital assistant called the Newton. A key step came in 2008 when Apple acquired PA Semi, a maker of 64-bit low power processors, for $278m. Apple was less interested in the chips themselves but wanted the skills and 150-strong engineering team around low power mobile processor design with an eye of future iPhones in particular. This led, two years later in 2010, to devices with Apple’s first homegrown chip, the A4, first in an iPad and then in the iPhone 4.

The A4 was the first SoC (system-on-chip) chip designed by Apple inhouse and was based on a custom implementation of an ARM Cortex-8 CPU architecture (Apple is one of the small band of chipmakers that has an architectural licence, which allows it to adapt the standard architecture – it was also an early stakeholder in the company that became ARM in the early days of microcomputers).

The A4 was manufactured by Samsung, Apple’s arch-rival at the device level, in the true spirit of coopetition. Another significant step then came in 2012 with launch of the A6, the first SoC custom-designed by Apple, using the ARMv7 based dual-core CPU called Swift. This was still fabricated by Samsung but Apple then switched to Taiwan’s TSMC for fabrication in 2014 for the A8 64-bit  SoC manufactured with a 20nm process.

Meanwhile, Apple was using Qualcomm chips for the cellular modems, leading to the various legal battles that we covered extensively at the time. This was a factor in Apple’s decision to switch to Intel for these and ultimate acquisition of the latter’s smartphone modem business for $1bn in July 2019. Intel in turn exited from smartphone modems, although retained its other cellular modem business around PCs, IoT devices and autonomous vehicles.

Again, Apple’s main motivation was acquisition of Intel’s 2,200 employees with expertise in cellular modems, along with various equipment, patents and IP. However, this did not immediately wean Apple off Qualcomm completely, because it was still reliant on the chip maker for 5G modems.

Indeed, only in February 2021 Qualcomm announced its latest Snapdragon X65, the world’s first 10 Gigabit 5G modem and antenna system for smartphones. Like the previous Snapdragon X60, the X65 can aggregate data from mmWave and sub-6 GHz bands simultaneously to achieve an optimal combination of high speed and low latency coverage. The latest modem however offers greater mmWave coverage and speed through being paired with Qualcomm’s most advanced mmWave antenna module. Apple therefore has little choice but to use the Snapdragon X65 in its upcoming 5G iPhones for launch in 2022, if it wants these to be cutting edge.

Apple is still dependent on the ARM architecture, but that is more of a common denominator and does not impede design or impose restrictions. Qualcomm itself is after all in the same boat, just as most semiconductor companies now rely on one or two major fabricators such as TSMC. However, the dynamics may change again, should Nvidia’s proposed acquisition of ARM go ahead, putting this vital shared IP into the hands of a single chip vendor.