Connect with us

Published

on

The Huawei booth at the Mobile World Congress in Barcelona, 2025.

Arjun Kharpal | CNBC

Despite being beaten down by years of U.S. trade restrictions, China’s telecom giant Huawei has quietly emerged as one of the country’s fiercest competitors across the entire AI landscape.  

Not only does the Shenzhen-based firm appear to represent Beijing’s answer to American AI chip darling Nvidia, but it has also been an early adopter of monetizing artificial intelligence models in industrial applications. 

“Huawei has been forced to shift and expand its core business focus over the past decade… due to a variety of external pressures on the company,” said Paul Triolo, partner and senior vice president for China at advisory firm DGA-Albright Stonebridge Group.

This expansion has seen the company get involved in everything from smart cars and operating systems to the technologies needed for the AI boom, such as advanced semiconductors, data centers, chips and large language models. 

“No other technology company has been able to be competent in so many different sectors with high levels of complexity and barriers to entry,” Triolo said. 

This year, Nvidia CEO Jensen Huang has become increasingly vocal in calling Huawei “one of the most formidable technology companies in the world.” He has also warned that Huawei will replace Nvidia in China if Washington continues to restrict U.S. chip firms’ exports to the Asian country.

Nvidia surpassed $4 trillion in market capitalization last week to become the world’s most valuable company. Its cutting-edge processors and a related “CUDA” computing system remain the industry standard for training generative AI models and applications. 

But that moat may be narrowing, as Huawei proves that it not only does it all, it does it well. While challenging American AI stalwarts like Nvidia is a tall order, the company’s history shows why it can’t be counted out.

Nvidia CEO Jensen Huang calls Huawei a formidable competitor

Telephone switches to national champion

Huawei, which now employs more than 208,000 people across over 170 markets, came from humble beginnings. Founded by ambitious entrepreneur Ren Zhengfei in 1987 out of an apartment in Shenzhen, the firm started as a small telephone switch distributor.

As it grew into a telecoms player, it gained traction by targeting less developed markets such as Africa, the Middle East, Russia and South America, before eventually expanding to places like Europe.

By 2019, Huawei would be well-positioned to capitalize on the global 5G rollout, becoming a leader in the market. Around this time, it had also blossomed into one of the world’s largest smartphone manufacturers and was even designing smartphone chips through its chip design subsidiary, HiSilicon. 

But Huawei’s success also attracted increasing scrutiny from governments outside China, particularly the U.S., which has frequently accused Huawei’s technology of posing a national security threat. The Chinese company has refuted such risks

The export controls have ironically pushed Huawei into the arms of the Chinese government in a way that CEO Ren Zhengfei always resisted.

Paul Triolo

partner and senior vice president for China at DGA-Albright Stonebridge Group

Huawei’s business suffered a major setback in 2019 when it was placed on a U.S. trade blacklist, preventing American companies from doing business with it. 

As the impact of the sanctions kicked in, Huawei’s consumer business – once the company’s largest by revenue – halved to about $34 billion in 2021 from the year before.

The company still managed a breakthrough on AI chips, and pressed ahead despite additional U.S. restrictions in 2020 that cut the company off from chipmaker Taiwan Semiconductor Manufacturing Co. A year earlier, Huawei officially launched its Ascend 910 AI processing chip as part of a strategy to build a “full-stack, all-scenario AI portfolio” and to become a provider of AI computing power.

But the U.S. targeting of Huawei also had the effect of turning the company into a martyr-like figure in China, building upon attention it received in 2018 when Meng Wanzhou, Huawei’s CFO and daughter of Ren, was arrested in Canada for alleged violations of Iran sanctions.

As the U.S.-China tech war continued to expand and broad advanced chip restrictions were placed on China, Huawei was an obvious choice to become a national champion in the race, with more impetus and state backing for its AI plans. 

“The export controls have ironically pushed Huawei into the arms of the Chinese government in a way that CEO Ren Zhengfei always resisted,” Triolo said. In this way, the restrictions also became “the steroids” for Huawei’s AI hardware and software stack.

The comeback 

After another year of declining sales in the consumer segment, the unit started to turn around in 2023 with the release of a smartphone that analysts said contained an advanced chip made in China. 

The 5G chip came as a shock to many in the U.S., who didn’t expect Huawei to reach that level of advancement so quickly without TSMC. Instead, Huawei was reportedly working with Chinese chipmaker SMIC, a company that has also been blacklisted by the U.S.

While semiconductor analysts said the scale that Huawei and SMIC could produce these chips was severely limited, Huawei nonetheless had proved it was back in the advanced chip game. 

It was also around this time that reports began surfacing about Huawei’s new AI processor chip, the Ascend 910B, with the company looking to seize upon gaps left by export controls on Nvidia’s most advanced chips. Mass production of the next-generation 910C is reportedly already on the way. 

Weekly analysis and insights from Asia’s largest economy in your inbox
Subscribe now

To fill the void left by Nvidia, Huawei “has been making big strides in replicating the performance of high-end GPUs using combinations of lower chips,” said Jeffrey Towson, managing partner at TechMoat Consulting.

In April, Huawei unveiled its “AI CloudMatrix 384”, a system that links 384 Ascend 910C chips in a cluster within data centers. Analysts have said CloudMatrix is able to outperform Nvidia’s system, the GB200 NVL72, on some metrics.

Huawei isn’t just catching up, “it’s redefining how AI infrastructure works,” Forrester analysts said in a report last month about CloudMatrix.

Meanwhile, Huawei has also developed its own “CANN” software system that acts as an alternative to Nvidia’s CUDA

“Winning the AI race isn’t just about faster chips. It also includes delivering the tools developers need to build and deploy large-scale models,” Forrester’s report said, though authors noted that Huawei’s products are still not integrated enough with other commonly used tools for developers to switch over quickly from Nvidia.

The ‘Ascend Ecosystem Strategy’

Jensen Huang: China is not behind the U.S. in AI development

These data centers, in turn, have provided the training capabilities and computing power used by Huawei’s suite of AI models under its Pangu series. 

Unlike other general-purpose AI models like OpenAI’s GPT-4 or Google’s Gemini Ultra 1.0, Huawei’s Pangu model is designed to support more industry-specific applications across the medical, finance, government, industrial and automotive sectors. Pangu has already been applied in more than 20 industries over the last year, the company said last month

Rolling out such AI applications often involves having Huawei tech staff working for months at the project site, even if it’s in a remote coal mine, Jack Chen, vice president of the marketing department for Huawei’s oil, gas and mining business unit, which provides digital and intelligent solutions to transform these industries, told CNBC.

That research enabled the company in May to deploy more 100 electric-powered trucks that can autonomously transport dirt or coal using the telecom company’s 5G network, AI and cloud computing services.

And it’s not limited to China. The technology can “be replicated on a large scale in Central Asia, Latin America, Africa, and the Asia-Pacific,” Chen said.

Huawei has also open-sourced the Pangu models, in a move it said would help it expand overseas and further its “Ascend ecosystem strategy,” which refers to its AI products built around its Ascend chips.

Speaking to CNBC’s “Squawk Box Asia” on Thursday, Patrick Moorhead of Moor Insights & Strategy said he expected Huawei to push Ascend in countries part of China’s Belt and Road Initiative — an investment and development project aimed at emerging markets. 

Over a period of five to 10 years, the company could begin to build serious market share in these countries, in the same way it once did with its telecommunications business, he added.

Continue Reading

Technology

Apple-supplier Luxshare shares pop 10% on report of OpenAI hardware deal

Published

on

By

Apple-supplier Luxshare shares pop 10% on report of OpenAI hardware deal

In this photo illustration, the Luxshare Precision company logo is seen displayed on a smartphone screen.

Sopa Images | Lightrocket | Getty Images

Shenzhen-listed Luxshare saw its shares jump about 10% on Monday, following a report that the Chinese device assembler had signed a deal with OpenAI to produce a consumer AI device. 

The company, which is also a supplier for Apple, is already developing a prototype of the device using ChatGPT large language models, The Information reported Friday, citing people familiar with the matter.

One source said that one of the products OpenAI has talked to suppliers about making would resemble a smart speaker without a display, which could put it in competition with Apple devices using Siri, with the company targeting late 2026 or early 2027 for its first device launches.

Luxshare’s 10% jump brought its year-to-date gains to about 50%. Most stocks on the Shenzhen Stock Exchange aren’t allowed to trade up or down more than 10% in a single day, relative to the previous day’s closing price. The company is also reportedly considering a secondary listing in Hong Kong this year. 

Luxshare and OpenAI did not immediately respond to CNBC’s requests for comment. 

OpenAI CEO Sam Altman on path to profitability: Willing to run at a loss to focus on growth

OpenAI has long been signaling a desire to push into dedicated AI devices, presenting a potential challenge to Apple and its iPhone. 

As part of these plans, The Information reported that OpenAI has been poaching staff from Apple to join its hardware division, now led by ex-Apple executive Tang Tan.

Earlier this year, the AI company, headed by Sam Altman, partnered with former Apple designer Jony Ive after buying his hardware startup, io Products, in a $6.4 billion deal. Ive is responsible for designing many of Apple’s most iconic products, including the iPhone, iPad and MacBook Air.

An OpenAI partnership with Luxshare would bring Apple’s longtime supplier — responsible for assembling items like the AirPods and Vision Pro — into the AI comapny’s orbit.

The AI company has also approached China-based Goertek, which also assembles AirPods, HomePods and Apple Watches, to provide components, including speaker modules, for its planned devices, according to the Information report.

OpenAI has been pushing deeper into hardware, hiring the former head of Meta’s Orion augmented reality glasses initiative in November to lead its robotics and consumer hardware efforts.

In the same month, OpenAI invested in robot startup Physical Intelligence, which focuses on “bringing general-purpose AI into the physical world,” according to its website.

Continue Reading

Technology

Apple takes control of all core chips in iPhone Air with new architecture to prioritize AI

Published

on

By

Apple takes control of all core chips in iPhone Air with new architecture to prioritize AI

iPhone Air is the big newcomer among Apple‘s latest lineup that went on sale Friday, but inside the slim phone’s raised plateau is another new piece of hardware that signals a renewed focus on artificial intelligence. 

Apple’s custom A19 Pro chip introduces a major architecture change, with neural accelerators added to each GPU core to increase compute power. Apple also debuted its first ever wireless chip for iPhone, the N1, and a second generation of its iPhone modem, the C1X. It’s a move analysts say gives Apple control of all the core chips in its phones.

“That’s where the magic is. When we have control, we are able to do things beyond what we can do by buying a merchant silicon part,” said Tim Millet, Apple vice president of platform architecture. He sat down with CNBC at Apple Park in September for the first U.S. interview about the new chips.

Until now, Broadcom was the main provider of wireless and bluetooth chips for iPhones, although Apple has made networking chips for the AirPods and Apple Watch for nearly a decade. Apple’s N1 is in the entire iPhone 17 lineup and the iPhone Air.

Arun Mathias, Apple vice president of wireless software technologies and ecosystems, gave CNBC an example of the N1’s improved Wi-Fi functionality. 

“One of the things people may not realize is that your Wi-Fi access points actually contribute to your device’s awareness of location, so you don’t need to use GPS, which actually costs more from a power perspective,” Mathias said. “By being able to do this more seamlessly in the background, not needing to wake up the application processor as much, we can do that significantly more efficiently.”

Apple’s new custom SoC for iPhone, A19 Pro, has neural accelerators added to the GPU cores to prioritize AI workloads

Emily Park

For iPhone modems, Qualcomm has been the sole provider since 2020. That changed in February when Apple unveiled the C1 in the iPhone 16e. It’s a plan first set in motion in 2019, with Apple’s purchase of Intel’s modem business for $1 billion. Qualcomm has long warned investors of the coming change. 

Qualcomm modems remain in the iPhone 17, 17 Pro and 17 Pro Max, but Apple’s C1X is in the iPhone Air. 

“It may not be as good as Qualcomm’s yet, in terms of just overall throughput and performance, but they can control it and they can make it run at lower power. So you’re going to get better battery life,” said Ben Bajarin, CEO of Creative Strategies, a technology research and consulting firm. He expects Apple to “completely phase out” Qualcomm in the “next couple of years.”

Apple’s Mathias said the C1X is “up to twice as fast” as the C1 and “uses 30% less energy” than the Qualcomm modem in the iPhone 16 Pro.

Neither Qualcomm or Broadcom saw much market impact following Apple’s announcement, and both companies will maintain licensing deals with Apple for certain core technologies.

AI accelerators on A19 Pro

Apple’s three new chips come amid increasing pressure from Wall Street about the company’s AI strategy.

“They probably won’t ever have their own Apple model like Google or OpenAI,” Bajarin said. “They’re still going to run those services on iPhone, right? They want the iPhone to be the best place for developers to run their AI.”

Apple has been making its own system on a chip, or SoC, since the A series launched with the iPhone 4 in 2010. The latest generation A19 Pro has a new chip architecture that prioritizes AI workloads, adding neural accelerators to the GPU cores.

“We are building the best on-device AI capability that anyone else has,” Millet told CNBC. “Right now we are focused on making sure that these phones that we’re shipping today, or shipping soon, will be capable of all the important on-device AI workloads that are coming.”

Privacy is a major reason Apple is prioritizing on-device AI, but Millet said there’s another reason, too. 

“It is efficient for us. It is responsive. We know that we are much more in control over the experience,” he said. 

One “built-in AI” feature Millet highlighted is the new front camera that uses AI to detect a new face and automatically switches to taking a horizontal photo. “It’s leveraging a full complement of almost all the capabilities in the A19 Pro,” Millet said.

Apple’s original AI hardware, its Neural Engine, was first unveiled back in 2017. It was barely mentioned at the launch. Instead, it’s all about adding compute power to the GPUs. 

“The integration of the neural processing is reaching MacBook Pro class performance inside an iPhone,” Millet said. “It’s a big, big step forward in ML compute. And so when you look inside the Neural Engine, for example, you have a lot of dense matrix math. We didn’t have that capability in our GPU. But now we do with A19 Pro.”

Bajarin told CNBC that Apple’s neural accelerators may work similarly to the tensor cores on Nvidia‘s AI chips, such as the H100.

“We’re integrating neural processing in a way that allows someone who’s writing a program to one of those small processors, extending the instruction set so they have a new class of computer that they have access to right there, and they can switch back and forth between 3D-rendering instructions and neural-processing instructions, all seamlessly inside the same microprogram,” Millet said.

Apple’s previous generation A19 SoC is in the base model iPhone 17, while the A19 Pro is in the iPhone Air, iPhone 17 and 17 Pro Max.

Apple’s iPhone 17 Pro shown on September 9, 2025 at Apple Park in California has enhanced 3D-rendering capabilities powered by Apple’s custom chip, A19 Pro, with neural accelerators added to the 6 GPU cores.

Katie Tarasov

Following overheating issues in the iPhone 15, a new “vapor chamber” in the Pro models keeps the custom chips cool.

“It’s actually positioned in concert with where the system on a chip, the A19 Pro is positioned,” said Kaiann Drance, Apple’s vice president of worldwide iPhone product marketing. “We think about how that all goes together, including with that forged unibody aluminum design, which is incredibly thermally conductive so that we can effectively dissipate heat with the vapor chamber, with where it’s positioned with our chip. And it’s even laser welded into it, which creates a metallic bond which also helps dissipate heat.”

More chips, more U.S. manufacturing

Apple still relies on others for smaller components, like Samsung for memory and Texas Instruments for analog chips. All bigger core chips, however, may be Apple-designed in every iPhone as soon as next year, according to Bajarin.

“We expect that there would be modems coming to Mac. We would expect there’s modems coming to iPad. There’s probably N variants of the networking chip coming to Mac,” Bajarin said. “I think over the course of the next few years, it will be on all of the portfolio.”

When CNBC asked Apple’s Millet if neural accelerators will be in the GPU cores of M5, the next anticipated SoC for Mac, he said, “We have a unified approach to architecture.”

The iPhone maker plans to manufacture at least some of its custom chips in the U.S., at facilities like Taiwan Semiconductor Manufacturing Company‘s new campus in Arizona, where CNBC got a tour of the first completed fab.

Apple’s A19 Pro is made at the leading edge of TSMC’s 3-nanometer node. While TSMC is working toward 3nm production in Arizona by 2028, it’s not there yet.

“If you need to be on the leading edge, it’s going to be Taiwan for the time being,” Bajarin said. 

In August, Trump announced a 100% tariff on chips from companies not making domestically. That same day, Apple increased its U.S. spending commitment to $600 billion over the next four years. CEO Tim Cook said part of that will go toward creating an “end-to-end silicon supply chain right here in America.”

“There’s really a question of what part of tariffs impact the silicon supply chain,” Bajarin said. “This is obviously why Apple and Tim Cook are on their mission and out there talking about investing in America.”

As part of that plan, Bajain said Apple could give struggling U.S. chipmaker Intel “serious consideration if 14A really does deliver on all of its promises.” Although, he added, it’s “going to be awhile” before Intel “becomes a viable option.”

For now, Apple is committed to making chips at TSMC Arizona.

“We are super excited about TSMC’s push into U.S. manufacturing. Obviously it will help us from a time zone perspective, and we also appreciate that the diversity of the supply is also really important,” Millet said.

When asked if he knows how much of Apple’s $600 billion U.S. spend will go toward custom silicon, Millet said, “I hope it’s a lot.”

Watch the video to see a behind-the-scenes look at Apple’s latest custom silicon.

Kif Leswing contributed to this report.

Continue Reading

Technology

Hands-on with the Meta Ray-Ban Display glasses

Published

on

By

Hands-on with the Meta Ray-Ban Display glasses

Mark Zuckerberg, chief executive officer of Meta Platforms Inc., wears a pair of Meta Ray-Ban Display AI glasses during the Meta Connect event in Menlo Park, California, US, on Wednesday, Sept. 17, 2025.

David Paul Morris | Bloomberg | Getty Images

When it comes to the new $799 Meta Ray-Ban Display glasses, it’s the device’s accompanying fuzzy, gray wristband that truly dazzles.

I was able to try out Meta’s next-generation smart glasses that the social media company announced Wednesday at its annual Connect event. These are the first glasses that Meta sells to consumers with a built-in display, marking an important step for the company as it works toward CEO Mark Zuckerberg’s vision of having headsets and glasses overtake smartphones as people’s preferred form of computing.

The display on the new glasses, though, is still quite simplistic. Last year at Connect, Meta unveiled its Orion glasses, which are a prototype capable of overlaying complex 3D visuals onto the physical world. Those glasses were thick, required a computing puck and were built for demo purposes only.

The Meta Ray-Ban Display, however, is going on sale to the public, starting in the U.S. on Sept. 30.

Though the new glasses include just a small digital display in their right lens, that screen enables unique visual functions, like reading messages, seeing photo previews and reading live captions while having a conversation with someone.

Controlling the device requires putting on its EMG sensor wristband that detects the electrical signals generated by a person’s body so they can control the glasses via hand gestures. Putting it on was just like strapping on a watch, except for the small, electric jolt I felt when it activated. It wasn’t as much of a shock as you feel taking clothes out of the dryer, but it was noticeable.

Donning the new glasses was less shocking, until I had them on and saw the little display emerge, just below my right cheek. The display is like a miniaturized smartphone screen but translucent so as to not obscure real-world objects.

Despite being a high-resolution display, the icons weren’t always clear when contrasted with my real-world field of view, causing the letters to appear a bit murky. These visuals aren’t meant to wrap around your head in crystal-clear fidelity, but are there for you to perform simple actions, like activating the glasses’ camera and glancing at the songs on Spotify. It’s more utility than entertainment.

The Meta Ray-Ban Display AI glasses with the Meta Neural Band wristband at Meta headquarters in Menlo Park, California, US, on Tuesday, Sept. 16, 2025.

David Paul Morris | Bloomberg | Getty Images

I had the most fun trying to perform hand gestures to navigate the display and open apps. By clenching my fist and swiping my thumb on the surface of my pointer finger, I was able to scroll through the apps like I was using a touchpad.

It took me several attempts at first to open the camera app through pinching my index finger and thumb together, and when the app wouldn’t activate I would find myself pinching twice, mimicking the double clicking of a mouse on a computer. But whereas using a mouse is second nature to me, I learned I have subpar pinching skills that lack the correct cadence and timing required to consistently open the app.

It was a bit strange and amusing to see people in front of me while I continuously pinched my fingers to interact with the screen. I felt like I was reenacting an infamous comedy scene from the TV show “The Kids in The Hall” in which a misanthrope watches people from afar while pinching his fingers and saying, “I’m crushing your head, I’m crushing your head!”

With the camera app finally opened, the display showed what I was looking at in front of me, giving me a preview of how my photos and videos would turn out. It was like having my own personal picture-in-picture feature like you would on a TV.

I found myself experiencing some cognitive dissonance at times as my eyes were constantly figuring out what to focus on due to the display always sitting just outside the center of my field of view. If you’ve ever taken a vision test that involves identifying when you see squiggly lines appearing in your periphery, you have a sense of what I was feeling.

Besides pinching, the Meta Ray-Ban Display glasses can also be controlled using the Meta AI voice assistant, just as users can with the device’s predecessors.

When I took a photo of some of the paintings decorating the demo room’s halls, I was told by support staff to ask Meta AI to explain to me what I was looking at. Presumably, Meta AI would have told me I was looking at various paintings from the Bauhaus art movement, but the digital assistant never activated correctly before I was escorted to another part of the demo.

I could see the Meta Ray-Ban Display’s live captions feature being helpful in noisy situations, as it successfully picked up the voice of the demo’s tour guide while dance music from the Connect event blared in the background. When he said “Let’s all head to the next room,” I saw his words appear in the display like closed-captions on a TV show.

But ultimately, I was most drawn to the wristband, particularly when I listened to some music with the glasses via Spotify. By rotating my thumb and index finger as if I was turning an invisible stereo knob,
I was able to adjust the volume, an expectedly delightful experience.

It was this neural wristband that really drilled into my brain how much cutting-edge technology has been crammed into the new Meta Ray-Ban Display glasses. And while the device’s high price may turn off consumers, the glasses are novel enough to potentially attract developers seeking more computing platforms to build apps for.

WATCH: Next important wearable tech will be glasses, says Meta’s chief product officer.

Meta's chief product officer on its latest AI smart glasses

Continue Reading

Trending