Inside Apple’s Spectacular Failure to Build a Key Part for Its New iPhones
The company set out to design a silicon chip that would allow it to cut ties with Qualcomm, a longtime supplier and bitter foe
The company set out to design a silicon chip that would allow it to cut ties with Qualcomm, a longtime supplier and bitter foe
The new iPhone models unveiled last week are missing a proprietary silicon chip that Apple had spent several years and billions of dollars trying to develop in time for the rollout.
The 2018 marching orders from Apple Chief Executive Tim Cook to design and build a modem chip—a part that connects iPhones to wireless carriers—led to the hiring of thousands of engineers. The goal was to sever Apple’s grudging dependence on Qualcomm, a longtime chip supplier that dominates the modem market.
The obstacles to finishing the chip were largely of Apple’s own making, according to former company engineers and executives familiar with the project.
Apple had planned to have its modem chip ready to use in the new iPhone models. But tests late last year found the chip was too slow and prone to overheating. Its circuit board was so big it would take up half an iPhone, making it unusable.
Investors had counted on Apple saving money with an in-house chip to help compensate for weak demand in the larger smartphone market. Apple—which hasn’t publicly acknowledged its modem project, much less its shortcomings—is estimated to have paid more than $7.2 billion to Qualcomm last year for the chips.
Engineering teams working on Apple’s modem chip have been slowed by technical challenges, poor communication and managers split over the wisdom of trying to design the chips rather than buy them, these people said. Teams were siloed in separate groups across the U.S. and abroad without a global leader. Some managers discouraged the airing of bad news from engineers about delays or setbacks, leading to unrealistic goals and blown deadlines.
“Just because Apple builds the best silicon on the planet, it’s ridiculous to think that they could also build a modem,” said former Apple wireless director Jaydeep Ranade, who left the company in 2018, the year the project began.
There were two reasons for the push, said former Apple executives and engineers familiar with the matter: Apple believed it could replicate the success of the microprocessor chips it designed for iPhones. Adoption of those chips fattened profit margins and improved performance for billions of devices. Second, Apple wanted to sever ties with Qualcomm, which it had accused in a 2017 lawsuit of overcharging for its patent royalties.
The companies settled the suit in 2019, and Apple, facing the expiration of its previous Qualcomm agreement, announced a deal last week to continue buying the company’s modem chips through 2026. Apple isn’t expected to produce a comparable chip until late 2025, people familiar with the matter said. There could be further delays, these people said, but the company believes it will eventually succeed.
Apple found that designing a microprocessor, essentially a tiny computer to run software, was easy by comparison. Modem chips, which transmit and receive wireless data, must comply with strict connectivity standards to serve wireless carriers around the world.
“These delays indicate Apple didn’t anticipate the complexity of the effort,” said Serge Willenegger, a former longtime Qualcomm executive who left the company in 2018 and doesn’t know the current state of the Apple chip. “Cellular is a monster.”
Apple’s push to build more of the various semiconductors used in its products stretches back more than a decade. In 2010, the company began using its own processing chips in iPhones and iPads. The chips helped Apple outperform many of its Android rivals, which relied on chips from Qualcomm, Taiwan-based MediaTek and other makers.
The company in 2020 began replacing processor chips from Intel, used for years in Mac computers, with a proprietary chip that allowed its laptops to run faster and generate less heat, improvements that helped boost flagging Mac sales. The Apple chip also saved the company an estimated $75 to $150 on every computer.
Credit for the success of Apple processor chips brought praise and increased authority to Johny Srouji, the company’s chip leader. “After shipping the first iPhone, we decided that the best way to deliver the best experience to our customers is to own and develop and design our silicon in-house,” Srouji said this year at Technion-Israel Institute of Technology, his alma mater.
Apple code-named its modem chip project Sinope, after the nymph in Greek mythology who outsmarted Zeus. It began taking shape in 2018, following the directive of Cook, Srouji, and others for Apple to build its own wireless components, said Chris Deaver, a former Apple human-resources executive and co-founder of BraveCore consultants.
By then, Apple’s relationship with Qualcomm had turned ugly. The companies bickered and swapped accusations of lying, theft and monopolistic practices.
Rubén Caballero, Apple’s longtime head of wireless, supported the Intel chip partnership at the time, while Srouji, senior vice president of hardware technologies, backed the pursuit of a company-built chip, said people involved in the project. Caballero left Apple in 2019.
Many members of Caballero’s team who were versed in wireless chip design were placed under Srouji. Other employees engaged in complementary wireless work, such as antenna design, were split off into the hardware engineering group. One of the top project managers on Srouji’s team had no background in wireless technology, said people who worked on the project.
Apple, which had been poaching engineering talent from Qualcomm for years, stepped up those efforts in March 2019. The company announced a new engineering hub in San Diego, Qualcomm’s hometown, and planned to add around 1,200 local jobs. That summer, Apple announced the acquisition of Intel’s wireless team and a portfolio of wireless patents.
Srouji flew to Munich to greet Apple’s newly acquired Intel wireless employees in December 2019. He told a gathering that the modem-chip project would be a game changer for Apple, the next step in the company’s evolution, said people who watched the meeting. He said the chip would distinguish Apple devices, as Apple’s processors had done.
As Apple filled the project’s ranks with Intel engineers and others hired from Qualcomm, company executives set a goal to have the modem chip ready for fall 2023. It soon became apparent to many of the wireless experts on the project that meeting the goal was impossible.
Apple found that employing the brute force of thousands of engineers, a strategy successful for designing the computer brain of its smartphones and laptops, wasn’t enough to quickly produce a superior modem chip.
Modem chips are trickier to make than processing chips because they must work seamlessly with 5G wireless networks, as well as the 2G, 3G and 4G networks used in countries around the world, each with its own technological quirks. Apple microprocessors run software programs designed solely for its iPhones and laptops.
Apple executives who didn’t have experience with wireless chips set tight timelines that weren’t realistic, former project engineers said. Teams had to build prototype versions of the chips and certify they would work with the many wireless carriers worldwide, a time-consuming job.
Executives better understood the challenge after Apple tested its prototypes late last year. The results weren’t good, according to people familiar with the tests. The chips were essentially three years behind Qualcomm’s best modem chip. Using them threatened to make iPhone wireless speeds slower than its competitors.
The company scratched plans to use the chips in Apple’s 2023 models, and the planned rollout was moved to 2024. Eventually, Apple executives realized the company wouldn’t meet that goal either. Apple instead opened negotiations with Qualcomm to continue supplying the modem chips. Apple’s licensing deal with Qualcomm expires in April 2025, though it can be extended for another two years.
Apple has the cash and the desire to keep pursuing its modem chip, according to people involved with the project.
“Apple isn’t going to give up,” said Edward Snyder, a managing director of Charter Equity Research and a wireless industry expert. “They hate Qualcomm’s living guts.”
Australian actor Chris Hemsworth has joined Archie Rose Distilling Co. as co-owner and strategic business partner as the Sydney spirits company prepares to enter the United States. The partnership brings Hemsworth together with Archie Rose founder Will Edwards, with the pair aiming to build greater international recognition for Australian whisky. Founded in Sydney in 2014, …
Continue reading “Chris Hemsworth Joins Archie Rose as Co-Owner Ahead of Global Expansion”
Porsche has revealed a one-off 911 GT2 RS that brings the extraordinary silhouette of its 935/78 “Moby Dick” racing car onto the road. Created by Porsche Sonderwunsch with Manthey, the commission began with a near-new GT2 RS. The owner requested greater performance, a Slantnose profile and a rear wing influenced by the limited 911 GT3 …
Continue reading “Porsche Builds a Road-Going “Moby Dick” From a 911 GT2 RS”
Nvidia’s earnings will test Wall Street’s confidence in the AI boom.
Chip makers are fighting to assure investors that the artificial-intelligence boom is racing forward. Wall Street might not believe it until Nvidia’s NVDA -0.98%decrease; down pointing triangle Jensen Huang says so.
When Huang steps up to the mic for his company’s earnings call Wednesday, he will have the world’s attention. What he says about Nvidia’s present will preview the future of AI, dictate the path forward for a tech-crazed stock market and influence an American economy increasingly tethered to hopes that the boom won’t go bust.
The $5 trillion chip maker has provided the key building blocks for AI since the launch of ChatGPT in 2022 set off a race for dominance among OpenAI, Anthropic and established Silicon Valley giants. Now, as Nvidia backstops sprawling data-center projects and an exotic money pipeline to boost chip demand, the company’s influence is arguably bigger than ever.
But there are signs of trouble ahead. Political pushback to AI is growing. A bond selloff propelled borrowing costs to their highest levels in years. The hyperscalers that include some of Nvidia’s key customers—once cash-printing machines—are relying more on debt. OpenAI recently told investors its revenue rose by a tepid 18% in the second quarter while its losses deepened.
Nvidia is increasingly stepping in to shore up potential weak points across the market. Earlier this month, the company teamed up with six of Wall Street’s biggest firms on a $500 billion AI-financing plan, pledging to backstop lending to customers that can’t afford its chips otherwise. The chip maker last week also took a stake in Cloverleaf Infrastructure, which arranges power for data centers, and struck a $6 billion deal with startup Poolside aimed at developing a powerful open-weight AI model.
After watching shares in other chip makers and the so-called Magnificent Seven tech companies swing wildly in recent months, Wall Street is hoping Nvidia can beat expectations—again. The countdown is on.
“It’s kind of becoming more and more like the World Cup final than the Super Bowl at this point,” said Brian Mulberry, chief market strategist at Zacks Investment Management. “It’s just gotten to be that big.”
The company has smashed analysts’ earnings estimates for each of the 14 quarters since the AI boom kicked into high gear. Nvidia posted 210% annual growth in net income in its last three-month period, according to FactSet, making Wall Street’s 126% projection look pedestrian.
Expectations for a blowout second quarter have risen rapidly over the course of this year. All Nvidia will have to do to beat this target: outrun 95% annual earnings growth to more than $51.5 billion. Analysts project the chip maker will report record sales of $92 billion for the period, up from a forecast of $78 billion at the start of this year.
In July, big-tech earnings sparked volatility. Concerns about runaway capital spending spread across the sector after Alphabet’s and Tesla’s results, driving a $890 billion wipeout that contributed to the unwind of hedge fund Situational Awareness. Microsoft posted the largest one-day gain in market capitalization by any company, ever, after a quarter proving that it could still show investors the money. SpaceX rocketed higher after a record-breaking initial public offering, only to see $1 trillion in value evaporate.
Surging memory prices and borrowing costs have fueled fears that those and other companies will be unable to keep plowing more money into supplies including Nvidia chips. Shaia Hosseinzadeh, founder of OnyxPoint Global Management, has recently bought dips in AI-infrastructure stocks when Wall Street has strained to absorb massive debt issued by Silicon Valley.
“The macro data is really quite robust,” he said. “Of course, there’s a level at which everything breaks.”
Investors have kept pumping money into the AI trade despite concerns around chip consumers—and to the benefit of chip producers. That is why Nvidia’s outlook for semiconductor demand could send ripples through counterparts such as Micron Technology and Sandisk, developers of the data centers in which their chips reside, and a supply chain of power producers, contractors and other specialists that underpin the globe-spanning AI build-out.
“We joke internally that we’re all Nvidia analysts now,” said David Lefkowitz, head of U.S. equities at UBS Global Wealth Management.
The irony is that investors have tended to sell Nvidia stock immediately after blockbuster earnings, with shares falling each trading session after its four past quarterly reports. Some are betting that will be the case this time around, too.
The options market is pricing in a 5.3% swing, higher or lower, in Nvidia shares during the session following earnings, according to Option Research & Technology Services. That is higher than the 4.8% average move in Nvidia’s stock over the last 12 months after the company reports quarterly results.
In recent days, some of the most actively traded Nvidia options have been put contracts tied to the stock falling from its Friday value of $214.75 to $205 and $210 apiece, according to Cboe Global Markets data. Put options give the right to sell a stock by a set price and typically represent a bearish wager.
Many analysts remain optimistic. Frank Lee, global head of tech hardware and semiconductor research at HSBC Global Investment Research, recently raised his price target for Nvidia shares to $360 from $325, citing, among other things, Nvidia’s strategic partnerships with suppliers and its role as a top contributor to open-source AI.
When the Writers Festival was called off and the skies refused to clear, one weekend away turned into a rare lesson in slowing down, ice baths included.
New research suggests that bonuses make employees feel more like a mere cog in a wheel.