Apple’s hovering inventory worth over the previous two decades has been pushed by its iconic shopper gadgets. It began with the iPod and that iMac. Then got here the iPhone and iPad. And extra lately, the Apple Watch and AirPods.
But there’s much more to the largest U.S. company by market cap than simply devices. At its Silicon Valley headquarters, in a non-descript room stuffed with a pair hundred buzzing machines and a handful of engineers in lab coats, Apple is designing the customized chips that energy its most common merchandise.
Apple first debuted homegrown semiconductors in the iPhone 4 in 2010. As of this yr, all new Mac computer systems are powered by Apple’s own silicon, ending the company’s 15-plus years of reliance on Intel.
“One of the most, if not the most, profound change at Apple, definitely in our merchandise over the final 20 years, is how we now accomplish that a lot of these applied sciences in-house,” mentioned John Ternus, who runs {hardware} engineering at Apple. “And prime of the listing, after all, is our silicon.”
That change has additionally opened Apple up to a brand new set of dangers. Its most superior silicon is primarily manufactured by one vendor, Taiwan Semiconductor Manufacturing Company. Meanwhile, smartphones are recovering from a deep sales slump, and rivals like Microsoft are making big leaps in artificial intelligence.
In November, CNBC visited Apple’s campus in Cupertino, California, the first journalists allowed to movie inside one in every of the company’s chip labs. We acquired a uncommon probability to discuss with the head of Apple silicon, Johny Srouji, about the company’s push into the complicated enterprise of customized semiconductor growth, which can be being pursued by Amazon, Google, Microsoft and Tesla.
“We have hundreds of engineers,” Srouji mentioned. “But for those who look at the portfolio of chips we do: very lean, truly. Very environment friendly.”
Unlike conventional chipmakers, Apple shouldn’t be making silicon for different firms.
“Because we’re probably not promoting chips exterior, we give attention to the product,” Srouji mentioned. “That offers us freedom to optimize, and the scalable structure lets us reuse items between completely different merchandise.”
Apple’s head of silicon, Johny Srouji, talks to CNBC’s Katie Tarasov at Apple headquarters in Cupertino, California, on November 14, 2023.
Andrew Evers
Powering iPhones since 2010
Srouji got here to Apple in 2008 to lead a small workforce of 40 or 50 engineers designing customized chips for the iPhone. A month after he joined, Apple purchased P.A. Semiconductor, a 150-person startup, for $278 million.
“They’re going to begin doing their very own chips: that was the quick takeaway after they purchased P.A. Semi,” mentioned Ben Bajarin, CEO and principal analyst at Creative Strategies. With its “inherent design focus,” Bajarin mentioned, Apple desires “to management as a lot of the stack” as potential.
Two years after the acquisition, Apple launched its first customized chip, the A4, in the iPhone 4 and authentic iPad.
“We constructed what we name the unified reminiscence structure that’s scalable throughout merchandise,” Srouji mentioned. “We constructed an structure that you just begin with the iPhone, however then we scaled it to the iPad after which to the watch and finally to the Mac.”
Apple’s silicon workforce has grown to hundreds of engineers working throughout labs throughout the world, together with in Israel, Germany, Austria, the U.Ok. and Japan. Within the U.S., the company has amenities in Silicon Valley, San Diego and Austin, Texas.
The major kind of chip Apple is creating is called a system on a chip, or SoC. That brings collectively the central processing unit (CPU), graphics processing unit (GPU) and different elements, Bajarin defined, including that for Apple there’s additionally a neural processing unit (NPU) “that runs the neural engine.”
“It is the silicon and all of the blocks that go on to that silicon,” Bajarin mentioned.
Apple’s first SoC was the A collection, which has superior from the A4 in 2010 to the A17 Pro introduced in September of this yr. It’s the central processor in iPhones, in addition to some iPads, Apple TVs and the HomePod. Apple’s different main SoC is the M collection, first launched in 2020, which now powers all new Macs and extra superior iPads. That product is up to the M3 line.
Launched in 2015, the S collection is a smaller system in package deal, or SiP, for Apple Watch. H and W chips are used in AirPods. U chips permit communication between Apple gadgets. And the latest chip, the R1, is ready to ship early subsequent yr in Apple’s Vision Pro headset. Dedicated to processing enter from the machine’s cameras, sensors and microphones, Apple says it would stream photos to the shows inside 12 milliseconds.
“We get to design the chips forward of time,” Srouji mentioned. He added that his staffers work with Ternus’s workforce “to precisely and exactly construct chips which can be going to be focused for these merchandise, and just for these merchandise.”
The H2 inside the 2nd technology AirPods Pro, for example, allows higher noise cancellation. Inside the new Series 9 Apple Watch, the S9 permits for uncommon capabilities like double tap. In iPhones, the A11 Bionic in 2017 had the first Apple Neural Engine, a devoted a part of the SoC for performing AI duties completely on-device.
The newest A17 Pro introduced in the iPhone 15 Pro and Pro Max in September allows main leaps in options like computational images and superior rendering for gaming.
“It was truly the largest redesign in GPU structure and Apple silicon historical past,” mentioned Kaiann Drance, who leads advertising and marketing for the iPhone. “We have {hardware} accelerated ray tracing for the first time. And we’ve mesh shading acceleration, which permits sport builders to create some actually gorgeous visible results.”
That’s led to the growth of iPhone-native variations from 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 at excessive quantity.
“The cause we use 3-nanometer is it offers us the capacity to pack extra transistors in a given dimension. That is vital for the product and significantly better energy effectivity,” Srouji mentioned. “Even although we’re not a chip company, we’re main the trade for a cause.”
Apple’s first 3-nanometer chip, the A17 Pro, allows ray tracing and different superior graphics rendering for improved gaming on the iPhone 15 Pro and Pro Max, proven right here in Cupertino, California, on September 12, 2023.
Katie Tarasov
Replacing Intel in Macs
Apple’s leap to 3-nanometer continued with the M3 chips for Mac computer systems, introduced in October. Apple says the M3 allows options like 22-hour battery life and, comparable to the A17 Pro, boosted graphics efficiency.
“It’s early days,” mentioned Ternus, who’s been at Apple for 22 years. “We have lots of work to do, however I feel there’s so many Macs now, just about all Macs are able to operating Triple-A titles, which isn’t what it was like 5 years in the past.”
Ternus mentioned that when he began, “the method we tended to make merchandise is we had been utilizing applied sciences from different firms, and we had been successfully constructing the product round that.” Despite a give attention to stunning design, “they had been constrained by what was obtainable,” he mentioned.
In a serious shift for the semiconductor trade, Apple turned away from utilizing Intel’s PC processors in 2020, switching to its personal M1 chip inside the MacE book Air and different Macs.
“It was nearly like the legal guidelines of physics had modified,” Ternus mentioned. “All of a sudden we might construct a MacE book Air that is extremely skinny and light-weight, has no fan, 18 hours of battery life, and outperformed the MacE book Pro that we had simply been delivery.”
He mentioned the latest MacE book Pro with Apple’s most superior chip, the M3 Max, “is 11 instances sooner than the quickest Intel MacE book Pro we had been making. And we had been delivery that simply two years in the past.”
Intel processors are based mostly on x86 structure, the conventional alternative for PC makers, with lots of software program developed for it. Apple bases its processors on rival Arm structure, recognized for utilizing much less energy and serving to laptop computer batteries last more.
Apple’s M1 in 2020 was a proving level for Arm-based processors in high-end computer systems, with different huge names like Qualcomm — and reportedly AMD and Nvidia — additionally creating Arm-based PC processors. In September, Apple extended its deal with Arm by at least 2040.
When its first customized chip got here out 13 years in the past, Apple was uncommon as a non-chip company attempting to make it in the cutthroat, cost-prohibitive semiconductor market. Since then, Amazon, Google, Microsoft and Tesla have tried their hand at customized chips.
“Apple was kind of the trailblazer,” mentioned Stacy Rasgon, managing director and senior analyst at Bernstein Research. “They kind of confirmed that for those who do that, you’ll be able to have a stab at differentiating your merchandise.”
Apple’s senior director of {hardware} validation Godfrey D’Souza exhibits off an M3 SoC in an Apple chip lab in Cupertino, California, on November 14, 2023.
Sydney Boyo
‘Modems are laborious’
Apple is not but making each piece of silicon in its gadgets. Modems, for instance, are one huge part the company has but to conquer by itself.
“The processors have been remarkably good. Where they’ve struggled is on the modem aspect, is on the radio aspect in the telephones,” Rasgon mentioned. “Modems are laborious.”
Apple depends on Qualcomm for its modems, though in 2019, the two firms settled a two-year legal battle over mental property. Soon after, Apple bought the majority of Intel’s 5G modem business for $1 billion, in a probable transfer to develop its personal mobile modem. That hasn’t occurred but, and in September, Apple signed on with Qualcomm to supply its modems through 2026.
“Qualcomm nonetheless makes the greatest modems in the world,” Bajarin mentioned. “Until Apple can do pretty much as good of a job, I’ve a tough time seeing them totally leap to that.”
Apple’s Srouji mentioned he could not touch upon “future applied sciences and merchandise” however mentioned “we care about mobile, and we’ve groups enabling that.”
Apple can be reportedly working on its personal Wi-Fi and Bluetooth chip. For now, it has a fresh multibillion-dollar deal with Broadcom for wi-fi elements. Apple depends on third events like Samsung and Micron for reminiscence.
“Our aspiration is the product,” Srouji mentioned, when requested if Apple will strive to design each a part of its chips. “We need to construct the greatest merchandise on the planet. As a expertise workforce, which additionally contains the chips in this case, we wish to construct the greatest expertise that may allow that imaginative and prescient.”
To ship on that goal, Apple will “purchase off the shelf” if it means the workforce can focus “on what actually, actually issues,” Srouji mentioned.
Regardless of how a lot silicon Apple finally designs, it nonetheless wants to manufacture its chips externally. That requires large fabrication vegetation owned by foundry firms like TSMC.
More than 90% of the world’s superior chips are made by TSMC in Taiwan, which leaves Apple and the remainder of the trade susceptible to the China threat of invasion.
“There is clearly lots of rigidity round, like, what would plan B be if that occurred?” Bajarin mentioned. “There is not one other good choice. You would hope that Samsung can be aggressive and Intel desires to be there. But once more, we’re not proper now. It’s actually all at TSMC.”
Apple is at least wanting to deliver a few of that manufacturing to the U.S. It’s committed to changing into the largest buyer at TSMC’s coming fab in Arizona. And on Thursday Apple introduced will probably be the first and largest buyer of the new $2 billion Amkor manufacturing and packaging facility being constructed in Peoria, Arizona. Amkor will package deal Apple silicon produced at TSMC’s Arizona fab.
“We at all times need to have a diversified provide: Asia, Europe and the U.S., which is why I feel TSMC constructing fabs in Arizona is nice,” Srouji mentioned.
Finding expertise
Another concern is the scarcity of expert chip labor in the U.S., the place superior fabs have not been constructed for decades. TSMC says its Arizona fab is now delayed to 2025 due to an absence of expert staff.
Whether or not it has to do with a scarcity of expertise, Apple has seen a slowdown in the launch of latest chips.
“Generations are taking longer as a result of they’re getting more durable and more durable,” Srouji mentioned. “And the capacity to pack extra and get energy effectivity can be completely different than 10 years in the past.”
Srouji reiterated his view that Apple has a bonus in that regard as a result of “I do not want to fear about the place do I ship my chips, how do I goal a bigger buyer base?”
Still, Apple’s actions underscore the competitiveness in the market. In 2019, Apple chip architect Gerard Williams left to lead an information middle chip startup referred to as Nuvia, bringing some Apple engineers with him. Apple sued Williams over IP considerations, earlier than dropping the case this yr. Qualcomm purchased Nuvia in 2021, in a transfer to compete in Arm-based PC processors like Apple’s.
“I can not actually focus on authorized issues, however we really care about IP safety,” Srouji mentioned. “When sure individuals depart for sure causes, that is their alternative.”
Apple has further macro challenges in its core enterprise as a result of smartphone gross sales are just recovering from their lowest ranges in years.
However, demand for AI workloads is main to a surge in orders for silicon, particularly for GPUs made by firms like Nvidia, whose inventory has jumped greater than 200% this yr tied to the reputation of ChatGPT and different generative AI companies.
Google has designed a tensor processing unit for AI since 2016. Amazon Web Services has had its own AI chips for the knowledge middle since 2018. Microsoft launched its new AI chip in November.
Srouji mentioned his workforce at Apple has been engaged on its machine studying engines, the Apple Neural Engine, since years earlier than it was launched in the A11 Bionic chip in 2017. He additionally pointed to embedded machine studying accelerators in its CPU and “extremely optimized GPU for machine studying.”
Apple’s Neural Engines energy what it calls “on-device machine learning features” like Face ID and Animojis.
In July, Bloomberg reported that Apple built its own large language model referred to as Ajax and a chatbot referred to as Apple GPT. A spokesperson declined to affirm or deny the accuracy of the report.
Apple has additionally acquired more than two dozen AI firms since 2015.
When requested if Apple seems to be falling behind in AI, Srouji mentioned, “I do not consider we’re.”
Bajarin is extra skeptical.
“It’s doable on Apple’s final yr chip, much more succesful on this yr’s chip with M3,” Bajarin mentioned, relating to Apple’s place in AI. “But the software program has acquired to meet up with that, in order that builders take benefit and write tomorrow’s AI software program on Apple Silicon.”
He anticipates enhancements, and shortly.
“Apple had a chance to actually get on that from day one,” Bajarin mentioned. “But I feel everybody expects it is coming in the coming yr.”
Watch the video to study extra.