At Apple's chip lab, the most significant transformation at the company in years is taking place.

Over the past two decades, Apple's soaring stock price has been fueled by its renowned consumer products, including the iPod and iMac, followed by the iPhone and iPad, and more recently, the Apple Watch and AirPods.

Apple, the largest U.S. company by market cap, is more than just gadgets. At its Silicon Valley headquarters, in a nondescript room filled with hundreds of buzzing machines and a few engineers in lab coats, Apple is designing the custom chips that power its most popular products.

In 2010, the iPhone 4 marked Apple's first debut of homegrown semiconductors. Currently, all new Mac computers run on Apple's own silicon, marking the end of the company's 15-plus years of relying on Intel.

Apple's hardware engineering head, John Ternus, stated that one of the most significant changes in the company's products over the past 20 years is the shift towards doing many technologies in-house, particularly in the realm of silicon.

The new development has exposed Apple to new risks, as its most advanced silicon is mainly produced by one vendor, Taiwan Semiconductor Manufacturing Company. At the same time, smartphones are experiencing a sales decline, and competitors like Samsung are making significant advancements in artificial intelligence.

In November, CNBC was granted access to Apple's chip lab in Cupertino, California, where they were able to film inside for the first time. During this visit, we had the opportunity to interview the head of Apple's silicon, Johny Srouji, about the company's efforts in the field of custom semiconductor development, which is also being pursued by companies such as Intel, AMD, Microsoft, and Tesla.

Srouji stated that although we have a large number of engineers, our chip portfolio is highly efficient and streamlined.

Unlike traditional chipmakers, Apple is not making silicon for other companies.

Srouji stated that since we don't sell chips outside, our focus is on the product. This allows us to optimize and the scalable architecture enables us to reuse components between different products.

Powering iPhones since 2010

In 2008, Srouji joined Apple to lead a team of around 40 to 50 engineers in designing custom chips for the iPhone. A month later, Apple acquired P.A. Semiconductor, a startup with 150 employees, for $278 million.

According to Ben Bajarin, CEO and principal analyst at Creative Strategies, when Apple bought P.A. Semi, the immediate takeaway was that they were going to start doing their own chips. Bajarin explained that Apple's "inherent design focus" drives them to control as much of the stack as possible.

The A4 chip, first launched by Apple two years after an acquisition, was introduced in the iPhone 4 and original iPad.

Srouji stated that we constructed a scalable architecture, which we refer to as the unified memory architecture, that can be applied to various products, starting with the iPhone and expanding to the iPad, watch, and eventually the Mac.

Thousands of engineers work across labs worldwide for Apple's silicon team, including in Israel, Germany, Austria, the U.K., and Japan. In the U.S., the company has facilities in Silicon Valley, San Diego, and Austin, Texas.

Apple is developing a system on a chip (SoC) that integrates the central processing unit (CPU), graphics processing unit (GPU), and other components, as well as a neural processing unit (NPU) that powers the neural engine, according to Bajarin.

Bajarin stated that it is the silicon and the blocks that are added to it.

The A series, initially launched as the A4 in 2010 and now up to the A17 Pro, is Apple's first system-on-chip (SoC) and serves as the central processor in iPhones, iPads, Apple TVs, and the HomePod. On the other hand, the M series, first released in 2020 and now up to the M3 line, is Apple's other major SoC and powers all new Macs and advanced iPads.

The S series, launched in 2015, is a smaller system in package (SiP) used in Apple Watch. H and W chips are used in AirPods, while U chips allow communication between Apple devices. The newest chip, R1, is set to ship early next year in Apple's Vision Pro headset, dedicated to processing input from the device's cameras, sensors, and microphones, with a stream time of just 12 milliseconds.

Srouji stated that his team collaborates with Ternus's team to create chips specifically designed for their products.

The H2 in the 2nd generation AirPods Pro improves noise cancellation. The S9 in the new Series 9 Apple Watch offers unique features such as double tap. The A11 Bionic in 2017's iPhones was the first to have the Apple Neural Engine, a dedicated part of the SoC that performs AI tasks entirely on-device.

The iPhone 15 Pro and Pro Max's latest A17 Pro announced in September allows for significant advancements in computational photography and gaming.

Apple's iPhone redesign was the largest in GPU architecture and Apple silicon history, according to Kaiann Drance, who leads marketing for the iPhone. The device features hardware accelerated ray tracing for the first time and mesh shading acceleration, enabling game developers to create visually stunning effects.

The development of iPhone-native versions of Assassin's Creed Mirage, The Division Resurgence, and Resident Evil 4 has been led to.

The A17 Pro is the first 3-nanometer chip to be produced in large quantities, according to Apple.

We use 3-nanometer technology to pack more transistors in a given dimension, which is crucial for our product and power efficiency. Despite not being a chip company, we are industry leaders.

Replacing Intel in Macs

The M3 chips for Mac computers, unveiled in October by Apple, are part of the company's ongoing effort to advance to 3-nanometer technology. The M3 offers features such as a 22-hour battery life and enhanced graphics performance, similar to the A17 Pro.

Ternus, who has been at Apple for 22 years, stated, "It's early days, but I believe there's a lot of work to be done. However, I think that now, almost all Macs are capable of running Triple-A titles, which was not the case five years ago."

Ternus stated that initially, the company relied on technologies from other companies, which limited their ability to create products that were fully customized to their needs. However, despite the emphasis on design, they were still constrained by the available technology.

In a significant change for the semiconductor industry, Apple abandoned Intel's PC processors in 2020 and began using its own M1 chip in the MacBook Air and other Macs.

Ternus remarked, "It was almost as if the laws of physics had shifted," as we suddenly created a MacBook Air that was remarkably slender and lightweight, had no fan, boasted an impressive 18 hours of battery life, and outperformed the MacBook Pro we had just shipped.

The M3 Max chip in the newest MacBook Pro is 11 times faster than the fastest Intel MacBook Pro we were shipping two years ago, according to him.

Processors developed by Apple are based on a different architecture than the traditional choice for PC makers, which is x86. This architecture is known for its power efficiency and helps laptop batteries last longer.

In 2020, Apple's M1 demonstrated the effectiveness of Arm-based processors in high-end computers, prompting other major companies like AMD, NVIDIA, and Qualcomm to also develop Arm-based PC processors. In September, Apple renewed its partnership with Arm, extending it through at least 2040.

Since its first custom chip was released 13 years ago, Apple stood out as a non-chip company attempting to succeed in the highly competitive and expensive semiconductor market. Now, Amazon, Google, Microsoft, and Tesla have also entered the custom chip market.

Stacy Rasgon, managing director and senior analyst at Bernstein Research, stated that Apple was a trailblazer in showing how to differentiate products by doing something unique.

‘Modems are hard’

Apple has not yet developed all the silicon used in its devices, such as modems.

Rasgon stated that the processors have performed exceptionally well, but they have faced challenges on the modem and radio sides in phones. He explained, "Modems are difficult."

Although Apple settled a legal battle with Qualcomm over intellectual property in 2019, the company still relies on Qualcomm for its modems. In 2019, Apple bought the majority of Intel's 5G modem business for $1 billion, but it hasn't developed its own cellular modem yet. In September, Apple signed on with Qualcomm to supply its modems through 2026.

Bajarin stated that Qualcomm's modems are still the best in the world, and he believes Apple must improve their performance before they can fully transition to using them.

Srouji from Apple stated that he couldn't discuss future technologies and products, but emphasized that the company prioritizes cellular and has teams working on it.

Apple is reportedly developing its own Wi-Fi and Bluetooth chip, but for now, it has a multibillion-dollar deal with Broadcom for wireless components. The company currently sources memory from third parties like Samsung and Micron.

According to Srouji, Apple's aspiration is to create the best products on the planet. As a technology team, including the chips, we aim to build the best technology that will enable this vision.

Srouji stated that Apple will "purchase pre-made solutions" to allow the team to concentrate on "critical matters."

Despite the amount of silicon Apple produces, it must still outsource its chip manufacturing to foundry companies such as TSMC, which own large fabrication plants.

TSMC, based in Taiwan, produces more than 90% of the world's advanced chips, leaving Apple and other companies vulnerable to the threat of invasion from China.

Obviously, there is a lot of tension around what would be plan B if that happened, as Bajarin said. There isn't another good option. Samsung and Intel want to be there, but currently, it's all at TSMC.

Apple is set to become the largest customer at TSMC's upcoming fab in Arizona and has announced it will be the first and largest customer of a new $2 billion manufacturing and packaging facility being built in Peoria, Arizona. Amkor will package Apple silicon produced at TSMC's Arizona fab.

Srouji stated that having a diversified supply from Asia, Europe, and the U.S. is always desired, and he believes TSMC constructing fabs in Arizona is excellent.

Finding talent

The shortage of skilled chip labor in the U.S. has caused delays in the construction of advanced fabs, with TSMC's Arizona fab now scheduled to open in 2025.

Apple has experienced a decrease in the speed of new chip releases.

Srouji stated that generations are becoming increasingly difficult to produce, and the power efficiency and packing capabilities have changed significantly over the past decade.

Srouji emphasized that Apple has an advantage because it doesn't need to worry about where to send its chips or how to target a larger customer base.

Apple's actions demonstrate the intense competition in the market. In 2019, Gerard Williams, Apple's chip architect, left to lead a data center chip startup called Nuvia, taking some Apple engineers with him. Apple sued Williams over IP concerns before dropping the case this year. Qualcomm later bought Nuvia in 2021 as a move to compete with Apple in the Arm-based PC processor market.

Although I am unable to discuss legal matters, our company places a strong emphasis on IP protection," Srouji stated. "People have the right to make their own choices when they choose to leave for various reasons.

Smartphone sales are still recovering from their lowest levels in years, posing additional macro challenges for Apple's core business.

The increasing demand for AI workloads is driving up orders for silicon, particularly GPUs produced by companies such as , whose stock has skyrocketed over 200% this year due to the popularity of ChatGPT and other generative AI services.

Since 2016, Google has created a tensor processing unit for AI, while Amazon Web Services has had its own AI chips in data centers since 2018. Microsoft recently released a new AI chip in November.

Srouji revealed that his team at Apple has been developing the Apple Neural Engine, its machine learning engines, for several years before its launch in the A11 Bionic chip in 2017. He also highlighted the embedded machine learning accelerators in the CPU and the highly optimized GPU for machine learning.

The "on-device machine learning features" of Apple, such as Face ID and Animojis, are powered by its Neural Engines.

In July, a report by Bloomberg stated that Apple developed a large language model named Ajax and a chatbot called Apple GPT. However, a spokesperson refused to verify or refute the report's accuracy.

Apple has also acquired more than two dozen AI companies since 2015.

Srouji stated that he believes Apple is not falling behind in AI.

Bajarin is more skeptical.

According to Bajarin, Apple's AI capabilities are impressive, but the software needs to improve to fully utilize the hardware's potential. This will enable developers to create the next generation of AI software on Apple Silicon.

He anticipates improvements, and soon.

Bajarin stated that Apple had the chance to excel from the beginning, but everyone anticipates its success to occur in the upcoming year.

Watch the video to learn more.