Inside An Apple Lab That Makes Custom Chips For iPhone And Mac

We passed through a small hallway through a simple set of double doors. Welcome to one of our chip labs. There's a simple room filled with a couple hundred machines, flashing lights, a handful of engineers in lab coats, and a bunch of postage-stamp-sized

chips

undergoing rigorous testing. Our goal is to be able to find errors, manufacturing problems and design problems. We want to find them so we can fix and address them before we ship our

chips

into our systems. Apple has enjoyed skyrocketing valuations for years thanks to its advanced consumer products MacBook, iPhone, Apple Watch and AirPods.

But under the hood, it's also designing its own

custom

silicon that powers them all. The transition to Mac is a proud moment. When we started scaling our chips to iPads and watches, proud moment. Building chips for AirPods that couldn't be created otherwise, proud moment. Apple first introduced its own semiconductors in iPhones in 2010. Starting in June of this year, all new Mac computers also run on Apple's own silicon, ending the company's more than 15 years of dependence on Intel. . I think one of the most profound changes, if not the most profound, at Apple, certainly in our products over the last 20 years, is how we now implement many of those technologies internally.

But Apple is not immune to industry risks. All of its most advanced silicon is manufactured by a single player, Taiwan Semiconductor Manufacturing Company. Smartphones are just recovering from a deep decline in sales and competitors like Microsoft are making great strides in generative AI. It's feasible. On last year's Apple chip, this year's M3 chip has even more capability, but the software has to catch up, so developers are taking advantage and writing tomorrow's AI software again on Apple Silicon. CNBC went to the company's headquarters in California, where we were the first journalists allowed to film inside an Apple chip lab and had the rare opportunity to talk to the silicon chief about how he broke into the incredibly expensive and complex processor business, starting a trend of companies that don't make chips like Amazon, Google, Microsoft and Tesla now rushing to do the same.

Apple launched the first iPhone in June 2007 with a 90-nanometer processor inside made by Samsung. In 2008, Johny Srouji joined after working at IBM and Intel. I came to Apple with the purpose of building our own silicon for the iPhone. At the time it was a very small team, between 40 and 50 engineers, and since then we have grown the team enormously. A month after Srouji joined, Apple bought PA Semiconductor, a 150-person chip startup, for $278 million. They are going to start making their own chips. That was the immediate conclusion when they bought PA Semi. I just think knowing Apple's culture, their inherent design approach to controlling the majority of the stack.

In fact, two years later, Apple launched its first

custom

chip, the A4, in the original iPad and iPhone 4. And we created what we call the unified memory architecture that is scalable across products. We built an architecture that starts with the iPhone, but then scaled it to the iPad and then the Watch and finally the Mac. Now, the Apple Silicon team has reached thousands of engineers working in chip labs in places like Israel, Germany , Austria, the United Kingdom, Japan and the United States in Austin, San Diego and Silicon Valley, where we were the first journalists to film the interior. one.

On either side of me, there are about 70 machines testing chips, and these are the M3 series that goes into the new MacBooks, as well as the A series chips that end up in the iPhone 15. And these particular machines are being tested. testing for extreme temperatures, high temperatures, low temperatures, things like that. Our board holders, we have color coded them. And what that does is it

makes

it easier for our engineers and technicians to be able to detect, oh, I'm looking for an M3 validation board. The main type of chips Apple is developing here are known as systems on a chip, or SoC.

It is the silicon and all the blocks that go to that silicon. So there are CPUs, there are GPUs, there are DSPs, there are accelerators. And in the case of Apple there is also an NPU that runs the neural engines. Apple's first SoC was the A series, which went from the A4 in 2010 to the A17 Pro, announced in September. It is the central processor of iPhones, as well as some iPads, Apple TV and HomePod. Its other major SoC is the M series, which now powers all new Macs and the most advanced iPads. First released in 2020, it is now dependent on M3 Macs.

The S series is a smaller system in package, or SIP, for Apple Watch. First released in 2015, H and W are even smaller chips used in AirPods. U chips enable communication between Apple devices and the newest chip, the R1, is for Apple's Vision Pro headphones, processing input and transmitting images to the screen in 12 milliseconds. We can design the chips in advance, working with our software and operating system partners on John's team to accurately and precisely build chips that will be intended for those products and only those products. The H2 inside the second-generation AirPods Pro, for example, allows for better noise cancellation inside the new Apple Watch series 9.

The S9 allows for unusual capabilities like double-tap. On the iPhone side, the A11 Bionic in 2017 was a major milestone because it had the first Apple Neural Engine, a part of the SoC dedicated to performing tasks entirely on the device. That's when we started thinking, wow, how do we incorporate this advanced intelligence into things like our camera processing? That's how things are in iOS 17 today, like being able to lift the subject of a photo. Kaiann Drance, who now leads marketing for the iPhone, has been at Apple for more than 15 years. She says the latest A17 Pro is another big step forward, allowing for big changes to features like computational photography.

You can take seven different shots, all from the same place. This is a macro shot and then you have the 13mm ultra wide angle lens. With the iPhone 15 Pro Max, you can go up to 5x optical zoom. The A17 Pro's new GPU architecture also enables advanced rendering for gaming on the iPhone 15 Pro. It was actually the largest GPU architecture redesign at Apple, in the history of Apple silicon. So for the first time we have hardware accelerated ray tracing and we have mesh shading acceleration, allowing game developers to create some really impressive visual effects. Now, for the first time, some big games are coming out with native iPhone versions: Ubisoft's Assassin's Creed Mirage, The Division Resurgence, and Capcom's Resident Evil 4.

Apple says the A17 Pro is the first 3-nanometer chip to ship in high volume. The reason we use three nanometers is that it gives us the ability to pack more transistors into a given dimension which is important for the product and much better power efficiency. We are leading. Although we are not a chip company, we lead the industry for a reason. Apple's jump to 3 nanometers continued with the M3 chips for Macs announced in October. These are the main chips that CNBC saw tested in Cupertino. We have one of those M3s at this thermal station. We have a socket and then the chip is there, and then the thermal actuator applies heat or cold directly onto the SoC package.

The M3 line allows for new features like a 22-hour battery life and, like the A17 Pro, improved graphics performance. They are the first days. We have a lot of work to do, but I think there are a lot of Macs out there now. Almost all Macs are capable of running Triple-A titles, which is not like five years ago. Now, Apple hardware chief John Ternus has been with the company for 22 years. When I started, the way we tended to make products was that we used technologies from other companies and effectively built the product around that. We've always had an amazing design team and made these beautiful products, but they were limited by what was available.

In an important change for the semi-industry. Apple stopped using Intel PC processors in 2020 and switched to its own M1 chip inside the MacBook Air and other Macs. It was almost as if the laws of physics had changed. Suddenly, we were able to build a MacBook Air. It's incredibly thin and light, it's fanless, it has 18 hours of battery life, and it outperformed the MacBook Pro we just shipped, and we just launched a MacBook Pro with M3 Max that's 11 times faster than the fastest Intel MacBook Pro that we have launched. we were manufacturing and shipping it just two years ago.

Intel processors are based on the x86 architecture, the traditional choice of PC manufacturers because a lot of software is developed for it. But Apple bases its processors on the rival ARM architecture, which is known for consuming less power, helping laptop batteries last longer. The M1 was a proving point for ARM-based processors in high-end PCs, and others like Qualcomm, AMD, and Nvidia are now developing ARM-based processors as well. In September, Apple extended its deal with ARM until at least 2040, when its first custom chip came out. 13 years ago, Apple was the only major non-chip company trying to make it in the cutthroat and prohibitive semiconductor market.

Apple is something of a pioneer. They kind of showed that if you do this, you can try to differentiate your products. And certainly more players have followed that path. In recent years. Amazon, Google, Microsoft and Tesla are among other giants that can afford to try their hand at custom chips. I think what everyone was looking for was: does this really give them a real sustainable advantage? Because that's really the question. It is an expensive task. Designing your own chips is incredibly expensive. Is this CapEx sustainable? We have thousands of engineers. But if you look at the chip portfolio, we are actually very efficient and very efficient.

Again, as we focus on the quality of the engineers we hire, we also focus on how we manage our teams. And since we don't actually sell chips overseas, we focus on the product. And that gives us freedom to optimize. And the scalable architecture allows us to reuse parts between different products. But Apple still doesn't make all the silicon parts for its devices. Modems, for example, are an important component that Apple has yet to conquer on its own. And their processors have been remarkably good. Where they have had problems is on the modem side and on the radio side on the phones.

Modems are difficult. And they've been struggling with that. Apple uses Qualcomm for its modems, but that relationship has become strained after a two-year legal battle over intellectual property. A settlement ended all legal action in 2019, and Apple then bought most of Intel's 5G modem business for $1 billion in a bid to develop its own cellular modem. That hasn't happened yet, and in September, Apple signed with Qualcomm to supply its modems until at least 2026. Qualcomm still

makes

the best modems in the world and, honestly, until Apple can do a job this good, it's hard to see them jumping. completely to that.

Tell me about what's going on with modems and the recent announcement to extend the deal with Qualcomm and why that decision was made. We have a strong team working internally to enable enabling technologies that would improve our products. But I can't really comment much on future technologies and products for obvious reasons. But we care about mobile telephony and we have equipment that allows it. Personally, I would say that the longer it takes for Apple to walk away, the more likely it is that they will never do it because it is a moving target and it is difficult.

That said, Apple has a very good semiconductor business in general, as if nothing is overlooked. Apple is reportedly working on its own Wi-Fi and Bluetooth chip, but for now it has a new multimillion-dollar deal with Broadcom for wireless components and relies on third parties like Samsung and Micron for memory. Does Apple have aspirations to be in charge of every part of the chip, including, say, memory? Our aspiration is the product. We want to build the best products on the planet as technological equipment, which in this case also includes chips. If there is a technology that we can buy off-the-shelf that meets our goal for the product, we will do it because I want to focus the team on what really matters.

Regardless of how much silicon Apple can design in-house, it will still need to manufacture its chips externally in massive manufacturing plants owned by foundry companies like TSMC. He's like any other person. I mean, they need cutting-edge semiconductors. There is a place, mainly a place. You get them right now, which is Taiwan. This leaves Apple, like all advanced chip companies, vulnerable amid growing threats from China to invade Taiwan. I think thatobviously there's a lot of tension, like what would be plan B if that happened? There is no other good option. Hopefully Samsung will also be competitive and Intel will want to be there.

But again, we are not. Right now. It's really all in TSMC. That's why Apple has committed to becoming the largest customer of TSMC's upcoming factory in Arizona. We always want to have a diversified offer. Asia, Europe and the US, so I think TSMC is building factories in Arizona and other founders are doing the same. There are other foundries, Samsung and Intel and others that are building in the US, so I think that's great. But another concern is the shortage of qualified labor to manufacture chips in the United States, where advanced factories have not been built for decades.

TSMC says its Arizona factory has now been delayed until 2025 due to a lack of skilled workers. Some have also wondered whether a fierce talent war for good chip engineers has led to a slowdown in Apple's release of new chips. There have been questions about why the iPhone lineup might have had last year's processor on the lower end and then the newer processor. I think part of that might also have to do with TSMC's capacity. Generations are taking longer because they are becoming more difficult, and the ability to pack more and get energy efficient is also different than ten years ago.

But again, I think this is actually one of the advantages that Apple has because we're not a chip company. So I don't have to worry about where I sell my chips? In 2019, Apple chip architect Gerard Williams suddenly left to lead a data center chip startup called Nuvia, bringing some Apple engineers with him. Apple sued Williams over intellectual property issues and then dropped the case this year. Qualcomm bought Nuvia in 2021, in an attempt to compete in ARM-based processors like Apple's. I can't really speak to legal issues, but we really care about intellectual property protection. When certain people leave for certain reasons, that is their choice.

Alright. And again, as I mentioned, we have thousands of engineers and we have a lot of talent. Another challenge facing the entire industry is that smartphone sales are barely recovering from their lowest levels in years. Smartphones have been horrible, right? And part of the issue is what are smartphones? They all look like this. It's like a black glass rectangle. So the industry is playing with things like they're playing with things like foldables. I think people right now are hoping that maybe AI will drive an update. Thanks to the hype around big language models like ChatGPT, the need for silicon has skyrocketed in recent months, primarily on the graphics processor side, where companies like Nvidia and AMD dominate.

But companies that don't make chips are also joining the trend. Google has designed a tensor processing unit for AI since 2016. Amazon Web Services has had its own AI chips for the data center since 2018, and now Microsoft launched a new chip in November. What does Apple have in the works when it comes to AI chips? An A11 Bionic, which we shipped in 2017 and usually release a chip many years in advance. It tells you how long we started. We actually started. We built our machine learning engine. We call it the Apple Neural Engine and that is a computing element.

We have built-in machine learning accelerators and our CPU. We have our GPU highly optimized for machine learning. Apple's neural engines power what it calls on-device machine learning features, such as Face ID and Animojis. We keep pushing and we keep seeing more and more apps, more features in our operating system that take advantage of this, and that's a really exciting time. In July, Apple reportedly built its own large language model called Ajax and a chatbot called Apple GPT. It has also acquired more than two dozen companies since 2015. Critics have been surprised that Apple appears to be falling behind when it comes to AI.

How do you respond? I don't think we are. Not too worried. Not too worried. And, in fact, we have an advantage. Again, going back to, we own the silicon, the hardware, the software, and the machine learning all on one machine. Let's optimize for that. Everything we've seen about generative AI is starting to trickle down to these devices at the platform level, right? Apple is not there yet. You know, Apple had the opportunity to really do it from day one, but I think everyone is looking forward to it coming next year. Meanwhile, Apple is preparing to ship its all-new line of R1 chips for the Vision Pro early next year, which means thousands of engineers like D'Souza will put them through rigorous testing behind closed doors in Apple's chip labs.

Apple around the world. the world. It's just a source of great pride to say, okay, we worked on this and, you know, here's the final product that we're actually using. It is fascinating.