Chinese Alibaba affiliate company, Ant Group, is exploring new ways to train LLMs and reduce dependency on advanced foreign semiconductors.

According to people familiar with the matter, the company has been using domestically-made chips – including those supplied by Alibaba and Huawei – to support the development of cost-efficient AI models through a method known as Mixture of Experts (MoE).

The results have reportedly been on par with models trained using Nvidia’s H800 GPUs, which are among the more powerful chips currently restricted from export to China. While Ant continues to use Nvidia hardware for certain AI tasks, sources said the company is shifting toward other options, like processors from AMD and Chinese alternatives, for its latest development work.

The strategy reflects a broader trend among Chinese firms looking to adapt to ongoing export controls by optimising performance with locally available technology.

The MoE approach has grown in popularity in the industry, particularly for its ability to scale AI models more efficiently. Rather than processing all data through a single large model, MoE structures divide tasks into smaller segments handled by different specialised “experts.” The division helps reduce the computing load and allows for better resource management.

Google and China-based startup DeepSeek have also applied the method, seeing similar gains in training speed and cost-efficiency.

Ant’s latest research paper, published this month, outlines how the company has been working to lower training expenses by not relying on high-end GPUs. The paper claims the optimised method can reduce the cost of training 1 trillion tokens from around 6.35 million yuan (approximately $880,000) using high-performance chips to 5.1 million yuan, using less advanced, more readily-available hardware. Tokens represent pieces of information that AI models process during training to learn patterns, in order to generate text, or complete tasks.

According to the paper, Ant has developed two new models – Ling-Plus and Ling-Lite – which it now plans to offer in various industrial sectors, including finance and healthcare. The company recently acquired Haodf.com, an online medical services platform, as part of its broader push for AI-driven healthcare services. It also runs the AI life assistant app Zhixiaobao and a financial advisory platform known as Maxiaocai.

Ling-Plus and Ling-Lite have been open-sourced, with the former consisting of 290 billion parameters and the latter 16.8 billion. Parameters in AI are tunable elements that influence a model’s performance and output. While these numbers are smaller than the parameter count anticipated for advanced models like OpenAI’s GPT-4.5 (around 1.8 trillion), Ant’s offerings are nonetheless regarded as sizeable by industry standards.

For comparison, DeepSeek-R1, a competing model also developed in China, contains 671 billion parameters.

In benchmark tests, Ant’s models were said to perform competitively. Ling-Lite outpaced a version of Meta’s Llama model in English-language understanding, while both Ling models outperformed DeepSeek’s offerings on Chinese-language evaluations. The claims, however, have not been independently verified.

The paper also highlighted some technical challenges the organisation faced during model training. Even minor adjustments to the hardware or model architecture resulted in instability, including sharp increases in error rates. These issues illustrate the difficulty of maintaining model performance while shifting away from high-end GPUs that have become the standard in large-scale AI development.

Ant’s research indicates a rise in effort among Chinese companies to achieve more technological self-reliance. With US export limitations limiting access to Nvidia’s most advanced chips, companies like Ant are seeking ways to build competitive AI tools using alternative resources.

Although Nvidia’s H800 chip is not the most powerful in its lineup, it remains one of the most capable processors available to Chinese buyers. Ant’s ability to train models of comparable quality without such hardware signals a potential path forward for companies affected by trade controls.

At the same time, the broader industry dynamics continue to evolve. Nvidia CEO Jensen Huang has said that increasing computational needs will drive demand for more powerful chips, even as efficiency-focused models gain traction. Despite alternative strategies like those explored by Ant, his view suggests that advanced GPU development will continue to be prioritised.

Ant’s effort to reduce costs and rely on domestic chips could influence how other firms approach AI training – especially in markets facing similar constraints. As China accelerates its push toward AI independence, developments like these are likely to draw attention across both the tech and financial landscapes.

The post Ant Group develops AI models using Chinese chips to lower training costs appeared first on TechWire Asia.

Nvidia revealed new AI chips at its annual GTC conference on Tuesday. CEO Jensen Huang introduced two key products: the Blackwell Ultra chip family, which is expected to ship in the second half of this year, and Vera Rubin, a next-generation GPU set to launch in 2026.

The release of OpenAI’s ChatGPT in late 2022 has significantly boosted Nvidia’s business, with sales increasing more than sixfold. Nvidia’s GPUs play an important role in the training of advanced AI models, giving the company a market advantage. Cloud providers like Microsoft, Google, and Amazon will be evaluating the new chips to see if they provide enough performance and efficiency gains to justify further investment in Nvidia technology. “The computational requirement, the scaling law of AI, is more resilient, and in fact, is hyper-accelerated,” Huang said.

The new releases reflect Nvidia’s shift to an annual release cycle for chip families, moving away from its previous two-year pattern.

Nvidia expands role in AI infrastructure partnership

Nvidia’s announcements come as the company deepens its involvement in the AI Infrastructure Partnership (AIP), a collaborative effort to build next-generation AI data centres and energy solutions. On Wednesday, BlackRock and its subsidiary Global Infrastructure Partners (GIP), along with Microsoft and MGX, announced updates to the partnership. Nvidia and Elon Musk’s AI company, xAI, have joined the initiative, strengthening its position in AI infrastructure development.

Nvidia will serve as a technical advisor to the AIP, contributing its expertise in AI computing and hardware. The partnership aims to improve AI capabilities and focus on energy-efficient data centre solutions.

Since its launch in September 2024, AIP has attracted strong interest from investors and corporations. The initiative’s initial goal is to unlock $30 billion in capital, with a target to generate up to $100 billion in total investment potential through a mix of direct investment and debt financing.

Early projects will focus on AI data centres in the United States and other OECD countries. GE Vernova and NextEra Energy are recent members of the partnership, bringing experience in energy infrastructure. GE Vernova will assist with supply chain planning and energy solutions to support AI data centre growth.

Vera Rubin chip family

Nvidia’s next-generation GPU system, Vera Rubin, is scheduled to ship in the second half of 2026, consisting of two main components: a custom CPU, Vera, and a new GPU called Rubin, named after astronomer Vera Rubin. Vera marks Nvidia’s first custom CPU design, built on an in-house core named Olympus. Previously, Nvidia used off-the-shelf Arm-based designs. The company claims Vera will deliver twice the performance of the Grace Blackwell CPU introduced last year.

Rubin will support up to 288 GB of high-speed memory and deliver 50 petaflops of performance for AI inference – more than double the 20 petaflops handled by Blackwell chips. It will feature two GPUs working together as a single unit. Nvidia plans to follow up with a “Rubin Next” chip in 2027, combining four dies into a single chip to double Rubin’s processing speed.

Blackwell Ultra chips

Nvidia also introduced new versions of its Blackwell chips under the name Blackwell Ultra, created to increase token processing, allowing AI models to process data faster. Nvidia expects cloud providers to benefit from Blackwell Ultra’s improved performance, claiming that the chips could generate up to 50 times more revenue than the Hopper generation, which was introduced in 2023.

Blackwell Ultra will be available in multiple configurations, including a version paired with an Nvidia Arm CPU (GB300), a standalone GPU version (B300), and a rack-based version with 72 Blackwell chips. Nvidia said the top four cloud companies have already deployed three times as many Blackwell chips as Hopper chips. Nvidia also referred to its history of increasing AI computing power with each generation, from Hopper in 2022 to Blackwell in 2024 and the anticipated Rubin in 2026.

DeepSeek and AI reasoning

Nvidia addressed investor concerns about China’s DeepSeek R1 model, which launched in January and reportedly required less processing power than comparable US-based models. Huang framed DeepSeek’s model as a positive development, noting that its ability to perform “reasoning” requires more computational power. Nvidia said its Blackwell Ultra chips are designed to handle reasoning models more effectively, improving inference performance and responsiveness.

Broader AI strategy

The GTC conference in San Jose, California, drew about 25,000 attendees and featured presentations from hundreds of companies that use Nvidia hardware for AI development. General Motors, for example, announced plans to use Nvidia’s platform for its next-generation vehicles.

Nvidia also introduced new AI-focused laptops and desktops, including the DGX Spark and DGX Station, designed to run large models like Llama and DeepSeek. The company also announced updates to its networking hardware, which ties GPUs together to function as a unified system, and introduced a software package called Dynamo to optimise chip performance.

Nvidia plans to continue naming its chip families after scientists. The architecture following Rubin will be named after physicist Richard Feynman and is scheduled for release in 2028.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Nvidia introduces new AI chips at GTC and joins AI infrastructure partnership appeared first on TechWire Asia.

Nvidia’s CEO Jensen Huang believes that global tech collaboration will remain strong, even as the US considers stricter policies on advanced computing products. During a recent visit to Hong Kong, Huang addressed concerns about the evolving political landscape, emphasising that science thrives on openness.

“Open science and global collaboration—cooperation across math and science—have been around for a very long time. They are the foundation of social and scientific advancement,” he said. “That’s not going to change.”

His outlook comes at a critical time. President-elect Donald Trump has reignited debates over tariffs and reshoring chip manufacturing, proposing measures that could significantly impact the global semiconductor sector.

Trump has long-supported tariffs as a tool for reshaping trade and manufacturing. During a recent appearance on Joe Rogan’s podcast, he criticised the CHIPS Act—a bipartisan effort signed in 2022 to boost US semiconductor production—calling it “so bad.” Instead of subsidies, Trump suggested imposing tariffs on semiconductors from Taiwan, arguing that this would push companies such as TSMC to build more facilities in the US.

However, experts are sceptical. William Reinsch, a senior adviser at the Center for Strategic and International Studies, pointed out that TSMC is already building a fab in Arizona. “Tariffs aren’t going to make that move any faster. If anything, they might complicate the effort,” he said.

Potential impacts on Nvidia and the tech industry

If Trump moves forward with tariffs, companies like Nvidia and AMD, which rely heavily on Taiwanese chips, could face rising costs. While their expenses might be passed down to customers, the ripple effects would likely be felt across the tech industry.

Huang was measured in his response to present uncertainties. “Whatever happens, we’ll balance compliance with laws and policies, continue to advance our technology, and support customers worldwide,” he said.

During his visit Huang also discussed broader issues, such as the energy demands of AI technologies. “If the world uses more energy to power the AI factories of the world, we’re a better world when that happens,” he said. He suggested sustainable solutions, such as placing AI supercomputers in remote areas powered by renewable energy.

“My hope and dream is that, in the end, we’ll all see that using energy for intelligence is the best use of energy,” Huang said, underscoring AI’s potential to address global challenges—from carbon capture to designing better wind turbines.

The stakes of reshoring chip manufacturing

Reshoring chip production has become a national security priority for the US, especially after the pandemic exposed vulnerabilities in global supply chains. As of 2021, 44% of US logic chip imports came from Taiwan. A major disruption in Taiwanese manufacturing could cause logic chip prices to surge by as much as 59%, according to a 2023 US International Trade Commission report.

The CHIPS Act aims to mitigate such risks, and companies have already started building new US facilities. Still, Trump’s proposed tariffs could introduce new challenges, potentially cutting profit margins for US-based companies like Nvidia.

Reactions to Trump’s tariff proposals are divided. Dan Newman, CEO of Futurum Group, suggests the idea may be more political posturing than a concrete plan. “Trump is unlikely to move forward with anything that hurts the economy,” he said.

However, Columbia Business School’s Lori Yue argued there’s a high chance Trump could impose tariffs. She added that deregulation related to AI under a second Trump administration might offset some of the financial strain on chip companies.

A new era for AI and computing

Huang closed his trip to Hong Kong on a hopeful note. Speaking at the Hong Kong University of Science and Technology after receiving an honorary doctorate in engineering, he told graduates they are entering a transformative era.

“The age of AI has started,” Huang said. “The whole world is reset. You’re at the starting line with everybody else. An industry is being reinvented. You now have the instruments necessary to advance science in so many different fields.”

While uncertainties about US technology policies remain, Huang’s message was clear: innovation will continue, driven by a new generation ready to redefine what is possible.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

The post Nvidia CEO Jensen Huang’s optimism amid US tech policies appeared first on TechWire Asia.

In a strategic maneuver blending innovation with geopolitical compliance, Nvidia Corp has recently unwrapped a variant of its gaming processor, the GeForce RTX 4090 D chip, designed to align with stringent US export controls, enabling its lawful sale in the Chinese market. This technological ballet comes at a time when Nvidia, poised for a market splash this month with the RTX 4090 D, concurrently charts a course for the second quarter of the year, initiating mass production of the H20 chip—an AI powerhouse designed to adhere to the intricate dance of US export regulations for China. 

Yet, a challenging crescendo emerges amid this orchestrated symphony of silicon and policy. Reports echo the hesitancy of Chinese companies to embrace the diluted allure of Nvidia’s chips–especially considering the forthcoming release of the downgraded H20. Reports now indicate that Chinese clients are cautiously exploring domestic alternatives, fearing the specter of potential tightening in US restrictions. 

The plot thickens in this delicate dance between technological prowess and geopolitical maneuvering, leaving the stage open for an unpredictable interplay of innovation and regulation.

What has Nvidia been doing for China?

In adherence to the latest US regulations governing chip exports, Nvidia has consistently introduced specialized AI chips and graphics cards explicitly tailored for the Chinese market. It all started with the A800 and H800, introduced as alternatives for Chinese customers in November 2022, about a month after the US first restricted exports of advanced microchips and equipment to China.

Unfortunately, a year into the October 2022 export control, the US further tightened its restrictions, which led to the barring of the shipments of advanced A800 and H800 AI chips to China. In response to the October 2023 renewed restrictions, Nvidia planned three other chips to comply with new US export rules – the H20, L20, and L2. 

In November last year, citing two sources familiar with the matter, Reuters reported that Nvidia has communicated to its clientele that the speculated H20 GPU, crafted to navigate through the export control measures imposed by the Biden administration on China, is anticipated to be unavailable until February or March this year.

The H20, L20, and L2 include most of Nvidia’s newest features for AI work. Still, some. Still, some of their computing power measures were cut back to comply with new US rules, according to SemiAnalysis’ analysis of the chips’ specifications. But the question remains: does China even want the downgraded version of Nvidia’s AI chips?

China would instead source a domestic alternative

For a start, according to sources cited by The Wall Street Journal (WSJ), since November 2023, major cloud service providers (CSP) in China, such as Alibaba and Tencent, have been testing samples of Nvidia’s special chips.

These Chinese enterprises have conveyed to Nvidia that the quantity of chips they plan to order in 2024 will be significantly lower than their initial plans.

In short, the new challenge is that significant Chinese CSPs are not interested in buying these chips’ lower-performing versions. A report by TrendForce indicated that Chinese enterprises have been testing the highest-performance version, H20, of Nvidia’s “special edition” AI chips. 

“Some testers have mentioned that this chip enables efficient data transfer among multiple processors, making it a better choice than domestic alternatives for building chip clusters required for processing AI computational workloads,” the report reads. Nevertheless, testers highlight the necessity for additional H20 to offset the performance difference compared to earlier Nvidia chips, leading to increased expenses.

According to the report by the WSJ, the advantage in the performance of Nvidia’s “downgraded” chips over domestic Chinese alternatives is diminishing in the short term, making locally manufactured chips more appealing to buyers. The report indicates that influential entities such as Alibaba and Tencent are shifting some advanced semiconductor orders to domestic companies, leaning towards internally developed chips. 

This shift in sourcing behavior is also noticeable among the other major chip purchasers, including Baidu and ByteDance. According to TrendForce data, around 80% of the high-end AI chips Chinese cloud computing companies use are currently sourced from Nvidia. However, this proportion may decrease to 50% or 60% in the next five years.

In the long term, TrendForce believes Chinese customers will begin to express uncertainty about Nvidia’s ability to consistently supply chips due to the potential tightening of chip export controls by US regulatory authorities. Focusing on independent AI chip development, Chinese CSPs like Baidu and Alibaba actively invest in autonomous AI chip initiatives. 

Baidu introduced its first self-developed ASIC AI chip, Kunlunxin, in early 2020, with plans for the second generation in 2021 and the third in 2024. After acquiring Zhongtian Micro Systems and establishing T-Head Semiconductor, Alibaba began creating its ASIC AI chips, including the Hanguang 800. 

TrendForce reports that while T-Head initially collaborated with external companies for ASIC chip design, post-2023, Alibaba is anticipated to increasingly rely on internal resources to strengthen the independent design capabilities of its next-gen ASIC chips, particularly for Alibaba Cloud’s AI infrastructure.

The post Nvidia faces dilemma as China shuns downgraded AI chips appeared first on TechWire Asia.

In a tech upheaval last August, Huawei ignited waves of discussion in the US and China by unveiling a smartphone with a 7nm processor crafted by Shanghai-based Semiconductor Manufacturing International Corp. (SMIC). A meticulous teardown conducted by a Canada-based research outfit, TechInsights, later revealed that the chip inside the Mate 60 Pro, though not at the absolute forefront, remarkably skirted the cutting edge—contrary to the intended impact of stringent US trade curbs. 

This unexpected revelation triggered jubilation within the Chinese tech sphere, and sparked a lively debate in the US about the efficacy of imposed sanctions. In the Mate 60 smartphone, Huawei showcased innovation and solidified its position at the vanguard of China’s journey to break free from Western technologies and foster homegrown alternatives. 

The allure of the Mate 60 captivated Chinese consumers in the last quarter of 2023, propelling Huawei past the symbolic US$100 billion revenue milestone. In this triumph, the company regained financial stature and chipped away at the towering dominance of Apple Inc.’s iPhone. After all, the US government has been blocking China’s access to semiconductors of 14-nm or better since 2022. 

The Mate 60 Pro smartphone by Huawei, powered by the domestically developed Kirin 9000S chip manufactured by China (SMIC), reflects significant progress in China’s quest to improve its domestic chip production ecosystem. “It is very concerning, but we’re going to take the strongest action to protect the US,” the US Commerce Secretary Gina M. Raimondo told Bloomberg on December 11 during her visit to China.

Raimondo’s remarks might have come off as a face-saving measure considering the efforts made by the US to prevent China from manufacturing such components. Until recently, SMIC was known for producing parts with a maximum of 14nm. However, a teardown study of the Mate 60 Pro by TechInsights for Bloomberg indicates that China can now have 7nm components, despite opposition from the US and its allies. It’s easy to extrapolate from there that the US tech clampdown has failed.

Huawei: smartphone and laptops with more advanced chips is the start of many more

During the company’s annual conference in China last month, Huawei CEO Richard Yu declared that the company will launch more leading and disruptive new products in 2024. “We will continue to exceed everyone’s expectations,” he said.

In the same month, Huawei took the world by surprise again, this time with a new laptop, equipped with an even more advanced chip: the 5-nm application processor (AP).

According to the Huawei website, the new laptop, the Qingyun L540, is powered by the Kirin 9006C AP chip made through a 5-nm process. It has an eight-core architecture and a clock speed of up to 3.13 gigahertz (GHz). Venturing into the realm of 5-nm technology marks a significant stride for the Shenzhen-based tech giant, placing it closer to the cutting-edge processes predominantly focused on 3nm nodes.

The 5nm feat by TSMC, not SMIC

Following a teardown of the Mate 60 Pro that TechInsights conducted for Bloomberg News in September 2023, leading to the revelation that the new Kirin 9000s chip was fabricated in China by SMIC, a similar dismantle was done on Huawei’s Qingyun L540 notebook, released last month.

The debut of the 14-inch ultralight in early December, featuring a 5nm processor, sparked intense speculation about China’s domestic semiconductor manufacturing capabilities, suggesting a more advanced state than previously believed.

This time, the teardown revealed that the enigmatic 5nm Kirin 9006C processor was made by chip giant Taiwan Semiconductor Manufacturing Co. (TSMC) rather than a Chinese manufacturer. An investigation into the origin of the underlying silicon showed that “analysis confirms with high confidence that the HiSilicon Kirin 9006C is indeed manufactured by TSMC, dispelling debates about SMIC involvement.”

The findings also show that the notebooks’ processor strongly resembles the HiSilicon Kirin 9000, which was packaged in the 35th week of 2020, the team concludes. “The die uses the same BEOL [back end of line] stack and process node that TechInsights has reported for the Kirin 9000 processor.”

Huawei’s acquisition of a three-year-old processor remains unclear, but the Chinese company has been building semiconductor reserves since facing global access restrictions. Despite being on the Washington Entity List since 2019, TSMC only stopped fulfilling Huawei orders in 2020 to comply with heightened US trade restrictions.

The post Huawei: no breakthrough, just a Taiwan-made chip powering the latest laptop appeared first on TechWire Asia.

In the closing chapters of 2023, Nvidia Corp, the current juggernaut of AI chips, pulled back the curtain on its ambitious venture into tailored innovation for the Chinese market. Navigating a complex terrain, Nvidia faced the formidable task of crafting chips that catered to the distinctive needs of its Chinese clientele and danced within the strict confines of Washington’s export restrictions. In December, 2023, the company unveiled a reimagined gaming chip designed expressly for the gaming landscape in China.

Boasting a “quantum leap in performance, efficiency, and AI-driven graphics,” according to the company, the chip is set to hit Chinese shelves in January, 2024, as confirmed by an Nvidia spokesperson to Reuters. “The GeForce RTX 4090 D has been designed to comply fully with US government export controls,” the spokesperson said.

Under the new US export control regulations, Nvidia cannot ship its flagship consumer gaming graphic card, the RTX 4090, to China, the world’s largest semiconductor market, making the RTX 4090 D a notable release. The debut of the GeForce RTX 4090 D signifies Nvidia’s first official launch of a China-focused chip since the Biden Administration’s introduction of export rules in October, 2023.

The Nvidia GTX 4090 D chip listed on the company’s China website packs approximately 10% fewer processing cores compared to its international counterpart, the 4090, sold in other countries. During Nvidia’s CEO Jensen Huang’s visit to Malaysia last month, he affirmed the company’s commitment to crafting compliant versions of its top-tier products, tailored for the Chinese market. 

Huang said his company is “extensively engaging” with the US government while developing graphics products that align with the export regulations imposed by the Biden administration, aiming to curb the rapid progress of China’s AI capabilities. The rules also resulted in the blockage of two modified AI chips, the A800 and H800, designed by Nvidia for the Chinese market to comply with the initial export rules unveiled in October 2022. 

On December 11, last year, US Commerce Secretary Raimondo, in an interview with Reuters, clarified that Nvidia has the green light to sell AI chips to China, except those boasting the highest processing power. In contrast to the prohibited RTX 4090, the China-centric RTX 4090 D is reported to be “5% slower in gaming and creating.”

Priced at 12,999 yuan (US$1,842), the China-targeted RTX 4090 D commands a premium of 350 yuan (US$50) compared to the second most advanced chip in the product series available to Chinese customers.

Is Nvidia working on any more AI chips for China?

In early November, the semiconductor industry newsletter, SemiAnalysis, suggested that Nvidia, in reaction to the recalibrated export rules unveiled in October 2023, might unveil three new AI chips designed for the Chinese market. Nevertheless, as per Reuters‘ report, Nvidia informed its Chinese customers of a delay in launching one of the chips, pushing the release to the first quarter of this year. 

Nvidia’s GeForce RTX 4090 D could be the first of the three new AI chips to very distinctly not flout US trade restrictions, and yet allow China to develop its AI programs. As of now, the other two chips have yet to appear on Nvidia’s China website. According to Dylan Patel, chief analyst at SemiAnalysis, one of the China-specific GPUs is over 20% faster than the H100 in LLM inference and is more similar to the new GPU that Nvidia is launching early this year than it is to the H100. 

Specification details on Nvidia’s new GPUs show that the AI chip giant is “perfectly straddling the line on peak performance and performance density with these new chips to get them through the new US regulations,” he added.

The post Nvidia unveils China-compliant gaming chip amid US export controls appeared first on TechWire Asia.

Malaysia’s semiconductor industry is considered one of the fastest-growing sectors in the world. The country is already a hub for the semiconductor supply chain, supplying key components to the industry over the past few years.

Malaysia has also witnessed continued investments in the semiconductor industry. Some of the major players manufacturing chips and other semiconductor components in the country include Intel, AMD, Broadcom, Infineon, Bosch, GlobalFoundries and many more.

The country currently controls 13% of the global market for packaging, assembly and testing services for semiconductors. It is also the sixth largest source of semiconductor exports in the world. This makes the country an ideal supplier in the semiconductor supply chain, with the US being a major trade partner in the industry too.

According to a report by Nikkei Asia, chip companies Jabil, Micron, Bosch, Western Digital and Lam Research are all expanding their manufacturing footprints in the area around Penang in Malaysia, stretching from Kulim in the north to the Batu Kawan Industrial Park in the south. DHL Express is constructing several logistics centers in the area, after launching direct cargo flights five days a week between Penang and Hong Kong, a chip-trading hub close to mainland China, in mid-2021.

Semiconductor industry in Malaysia

Given the development of the chip industry in Malaysia, China may now see opportunities to rely on the country not only to solve its supply chain problems but also as a solution to the sanctions that have been imposed on it.

Last year, the US government enforced chip sanctions on China, intending to slow down the country’s capabilities in developing AI. However, there were some loopholes in the sanctions that allowed Chinese firms to keep buying and building the chips used to train some of the world’s most advanced AI algorithms.

The US only realized the loopholes when companies like Huawei were able to develop a 5G phone despite sanctions. Companies like Nvidia were also still doing business with China but with a different version of chips, which were developed based on the limits of the sanctions. For example, the A800 and H800 chips were heavily sourced by Tencent and Alibaba in China.

Jensen Huang, CEO of Nvidia, who was in Malaysia recently, reaffirmed that China remains an important market for the company. Huang explained the situation, saying, ‘The US has determined to regulate our technology, the highest end of our technology, and limit its access to China. So, the regulations specify the maximum performance we can ship to China, and we will follow the regulations precisely – and very explicitly do so.” He added that Nvidia is currently developing new chips specifically for the Chinese market.

Huang also confirmed that Malaysia is a key player not just in the semiconductor supply chain but also in the data center industry in the region.

In Malaysia particularly, the data center infrastructure, “a layer of computing which is one of the most important parts of AI and cloud, is very successful,” Huang said. “And so I think that we’re going to see Southeast Asia participate across the entire technology stack, and Malaysia play a role in cloud infrastructure,” he added.

Another loophole?

According to a report by Reuters, several Chinese semiconductor design companies are tapping Malaysian firms for chip assembly. The report said that the Chinese companies are asking Malaysian chip packaging firms to assemble graphics processing units (GPUs).

Assembling chips does not contravene any US restrictions. The sources also said that the assembling is not the same as the fabrication of the chip wafers. While the sources of the Reuters report declined to name the companies involved in the assembling, a few contracts have already been agreed.

The US recently updated its sanctions on chips towards China, which focused on covering some of the loopholes that were present in the previous sanctions announced in 2022. The focus of the sanctions remains mainly on the sale of sophisticated chipmaking equipment.

Malaysia, a major hub in the semiconductor supply chain, is seen as well placed to grab further business as Chinese chip firms diversify outside of China for their chip assembling needs.

The Reuters report quoted Unisem chairman John Chia as saying “Many Chinese chip design houses have come to Malaysia to establish additional sources of supply outside of China to support their business in and out of China.”

Unisem provides semiconductor assembly and test services in Malaysia. Its packaging and testing facilities are located in Ipoh, Perak, Malaysia, Chengdu, China, and Batam, Indonesia. It also has wafer bumping facilities in Ipoh, Perak, Malaysia and Chengdu, China. China’s Huatian Technology is a majority shareholder.

Apart from chip design, China is Malaysia’s largest trade partner. Malaysia is also part of China’s “One Belt One Road” policy. Chinese investments remain high in the country, across a variety of industries.

Asked whether accepting orders to assemble GPUs from Chinese firms could potentially provoke US ire, Chia said Unisem’s business dealings were “fully legitimate and compliant,” and that the company did not have the time to worry over “too many possibilities.”

Unisem’s customers in Malaysia included those from the United States as well. The US Department of Commerce did not respond to requests for comment.

The post Could Malaysia’s semiconductor industry help China?  appeared first on TechWire Asia.

In the ever-evolving landscape of technology, the AI chips market stands at the forefront of innovation, driving unprecedented advancements across diverse industries. These specialized processors, crucial for accelerating AI computations, play pivotal roles in applications spanning data centers, cloud computing, edge devices, and personal electronics. As the demand for enhanced machine learning capabilities continues to soar, industry giants like Intel, AMD, and Nvidia are engaging in fierce competition to deliver cutting-edge AI chips, surpassing escalating performance expectations.

In the high-stakes race for AI accelerators, AMD, too, has thrown its hat into the ring, gearing up to challenge Nvidia’s dominance. The company, traditionally known for its CPUs and GPUs, only recently unveiled its most advanced GPU for AI, the MI300X, which is capable of performing 1.4 to 1.6 times better than the competition in the inference market. According to AMD’s CEO Lisa Sui, this represents a solid challenge to Nvidia, which currently dominates the market for AI chips with over 80% market share.

With AMD boldly predicting that the market for AI accelerators could soar beyond the US$400 billion mark within the next four years, the tech arena is about to witness a seismic shift as AMD aims to reshape the future of accelerated computing. But as anticipated by market experts, more players, especially tech giants, will be making their own forays into the AI chips market, or will at least up the ante of their current portfolio – mainly to challenge Nvidia’s dominance, mainly on the grounds that nobody but a monopolist enjoys existing in a monopoly. 

At this year’s AI Everywhere event on December 14, Intel joined the likes of AMD by unveiling a constellation of new chips set to redefine the landscape of AI hardware still further.

Intel’s new AI chips

Stepping into the booming AI market with gusto, Intel’s latest offerings include revamped Xeon server chips—marking the second overhaul for those chips in less than a year. The Xeon chips have been redesigned to catapult Intel to the forefront of AI innovation by promising heightened performance and memory capabilities while requiring less electricity. 

The spotlight is also falling on Intel’s Ultra Core chips, which allow laptops and desktops to wield AI prowess directly. Amid its chip lineup, Intel added the Gaudi 3, the newest addition to a lineage that aims to challenge Nvidia’s industry-dominating H100. AI accelerators play a pivotal role in birthing chatbots and delivering many swiftly emerging services. With Gaudi 3 slated for a 2024 release, Intel is boldly asserting its superiority over the H100, setting the stage for an AI showdown.

Intel’s CEO, Pat Gelsinger, is betting on the transformative power of AI to breathe new life into the company as it navigates through the shadows of past missteps and a PC market lull. But the playing field has never been more intense. Intel’s arch-nemesis, AMD has deftly seized portions of the PC and server market while Intel has been forced to grapple with the unsettling reality of major clients meking their chips in-house.

In this tech maelstrom, Nvidia emerges as a formidable titan, dominating SPACE’s data center chip with its AI accelerators. The success of these products has propelled Nvidia’s valuation to US$1.1 trillion, setting the stage for a historic revenue overtake of Intel in 2023, as per industry analysts’ projections.

In the throes of this seismic shift, Intel, once the uncontested giant in the chipmaking realm, now charts a course through a fiercely competitive landscape where the winds of change are blowing at gale force.

Intel has AI chips “for all”

At the AI Everywhere event in the heart of New York City, Intel unveiled a portfolio of AI products to enable customers’ AI solutions everywhere — across the data center, cloud, network, edge, and PC. “Intel is on a mission to bring AI everywhere through exceptionally engineered platforms, secure solutions, and support for open ecosystems. Our AI portfolio gets even stronger with today’s launch of Intel Core Ultra, ushering in the age of the AI PC and AI-accelerated 5th Gen Xeon for the enterprise,” Gelsinger said.

Intel’s latest innovation, the Intel Core Ultra, marks a significant architectural shift after four decades, ushering in a new era for AI-powered PCs and applications. This transformative technology encompasses CPU computing, graphics, power efficiency, and battery life advancements and introduces cutting-edge AI features. Unsurprisingly, projections suggest that AI PCs will dominate 80% of the market by 2028.

Featuring Intel’s first client on-chip AI accelerator, the neural processing unit (NPU), the Core Ultra achieves 2.5 times better power efficiency for AI acceleration than its predecessor. Partnering with over 100 software vendors, Intel plans to bring hundreds of AI-boosted applications to the PC market, promising a diverse range of creative, productive, and enjoyable experiences. 

The coming of Gaudi 3

The Core Ultra-based AI PCs are available from select US retailers for the holiday season, with plans to expand to over 230 designs worldwide in the next year. Then there’s the newly introduced 5th Gen Xeon processor family, which signifies a substantial advancement in performance and efficiency across the data center, cloud, network, and edge computing. 

Compared to the previous generation, these processors boast a remarkable 21% average performance gain for general computing, enabling a 36% higher average performance per watt across various customer workloads. Upgrading to the 5th Gen Xeon can result in a 77% reduction in total cost of ownership (TCO) for customers on a typical five-year refresh cycle.

Intel claims that Xeon is the only mainstream data center processor with built-in AI acceleration, with the new 5th Gen Xeon delivering up to 42% higher inference and fine-tuning on models as large as 20 billion parameters. “It’s also the only CPU with a consistent and ever-improving set of MLPerf training and inference benchmark results,” Intel contends.

During the event, IBM announced that 5th Gen Intel Xeon processors achieved up to 2.7x better query throughput on its watsonx.data platform than previous-generation Xeon processors during testing. Google Cloud, which will deploy 5th Gen Xeon next year, noted that Palo Alto Networks experienced a 2x performance boost in its threat detection deep learning models using built-in acceleration in 4th Gen Xeon through Google Cloud.

 And indie game studio Gallium Studios turned to Numenta’s AI platform running on Xeon processors to improve inference performance by 6.5x over a GPU-based cloud instance, saving cost and latency in its AI-based game, Proxi11.

Concluding the event, Gelsinger shared an update on the upcoming Intel Gaudi 3, which is set to debut next year. Revealing it for the first time, he showcased the next-generation AI accelerator designed for deep learning and creating large-scale generative AI models. Intel’s Gaudi pipeline has experienced rapid expansion, driven by proven performance advantages and competitive Total Cost of Ownership (TCO) pricing.

 As the demand for generative AI solutions continues to rise, Intel aims to secure a significant share of the accelerator market in 2024, driven by its suite of AI accelerators and led by Gaudi.

“Can Intel chase down Nvidia?” We’ll know next year…

The post Intel unleashes AI chips for PCs and data centers, challenging AMD and Nvidia appeared first on TechWire Asia.

The demand for AI workloads has seen Microsoft taking matters into its own hands. As companies want better AI infrastructure to support their development of AI use cases, the need to deliver that infrastructure has become a challenge for tech companies around the world.

More AI use cases simply means the need for more compute power. And the need for more computing power means the need for more data centers and chips to process these workloads. But the problem now is, are there sufficient chips capable of doing all this?

While the shortage of chips is normally seen as the reason for difficulties and stalling in the progress of AI, there is also the increasing costs of chips as well as the challenge of making sure everything can work together with minimal complexity. This includes ensuring the cooling systems can support the amount of heat generated from data centers – which, with increased chip complexity, is no longer in any sense a certainty.

For Microsoft, AI will be key for the company’s direction in the future, especially in the areas in which it plans to develop solutions for customers. As such, Microsoft has unveiled two of its own custom-designed chips and integrated systems at its Ignite event. The Microsoft Azure Maia AI Accelerator will be optimized for AI tasks and generative AI,  while the Microsoft Azure Cobalt CPU will be an Arm-based processor tailored to run general-purpose compute workloads on the Microsoft Cloud.

“Cobalt is the first CPU designed by us specifically for Microsoft Cloud, and this 64-bit 128-core ARM-based chip is the fastest of any cloud provider. It’s already powering parts of Microsoft Teams, and Azure communication services, as well as Azure SQL. And next year we will make this available to customers,” said Satya Nadella, Microsoft CEO in his keynote address at the event.

“Starting with the Maia 100 design running cloud AI workloads like LLM training and inference, this chip is manufactured on a five-nanometre process, and has 105 billion transistors, making it one of the largest chips that can be made with current technology. And it goes beyond the chip, as we have designed Maia 100 as an end-to-end rack for AI,” added Nadella.

Expected to be rolled out early next year to Microsoft data centers, the chips will initially power the company’s services, such as Microsoft Copilot or Azure OpenAI Service. They will then join an expanding range of products from industry partners to help meet the exploding demand for efficient, scalable and sustainable compute power, and the needs of customers eager to take advantage of the latest cloud and AI breakthroughs.

The Microsoft Azure Maia AI Accelerator and Microsoft Azure Cobalt CPU

As the new Maia 100 AI Accelerator is expected to power some of the largest internal AI workloads running on Microsoft Azure, it only made sense for OpenAI to provide feedback on the development as well.

According to Sam Altman, CEO of OpenAI, the company has worked together with Microsoft in designing and testing the new chip with its models. For Altman, Azure’s end-to-end AI architecture, now optimized down to the silicon with Maia, paves the way for training more capable models and making those models cheaper for customers.

Looking at the hardware stack, Brian Harry, a Microsoft technical fellow leading the Azure Maia team, explained that vertical integration, which is the alignment of chip design with the larger AI infrastructure designed with Microsoft’s workloads in mind, can yield huge gains in performance and efficiency.

Meanwhile, Wes McCullough, corporate vice president of hardware product development at Microsoft pointed out that the Cobalt 100 CPU is built on Arm architecture, a type of energy-efficient chip design, and optimized to deliver greater efficiency and performance in cloud-native offerings. McCullough added that choosing Arm technology was a key element in Microsoft’s sustainability goal. It aims to optimize performance per watt throughout its data centers, which essentially means getting more computing power for each unit of energy consumed.

Partnership with Nvidia

Apart from the new chips, Microsoft is also continuing to build its AI infrastructure in close collaboration with other silicon providers and industry leaders, such as Nvidia and AMD. With Nvidia, Azure works closely on using the Nvidia H100 Tensor Core (GPU) graphics processing unit-based virtual machines for mid to large-scale AI workloads, including Azure Confidential VMs.

The NC H100 v5 virtual machine (VM) series, which is now available for public preview, is the latest addition to Microsoft’s portfolio of purpose-built infrastructure for high performance computing (HPC) and AI workloads. The new Azure NC H100 v5 series is powered by Nvidia Hopper generation H100 NVL 94GB PCIe Tensor Core GPUs and 4th Gen AMD EPYC Genoa processors, delivering powerful performance and flexibility for a wide range of AI and HPC applications.

Azure NC H100 v5 VMs are designed to accelerate a broad range of AI and HPC workloads, including:

• Mid-range AI model training and generative inferencing: unlike the massively scalable ND-series powered by the same Nvidia Hopper technology, the NC-series is optimized for training and inferencing AI models that require smaller data size and a smaller number of GPU parallelism. This includes generative AI models such as DALL-E, which creates original images based on text prompts, as well as traditional discriminative AI models such as image classification, object detection, and natural language processing focused on the accuracy of prediction rather than the generation of new data.
• Traditional HPC modeling and simulation workloads: Azure NC H100 v5 VMs are also an ideal platform for running various HPC workloads that require high compute, memory, and GPU offload acceleration. This includes scientific workloads such as computational fluid dynamics (CFD), molecular dynamics, quantum chemistry, weather forecasting and climate modeling, and financial analytics.

Nvidia also introduced an AI foundry service to supercharge the development and tuning of custom generative AI applications for enterprises and startups deploying on Microsoft Azure.

The Nvidia AI foundry service pulls together three elements — a collection of Nvidia AI foundation models, Nvidia NeMo framework and tools, and Nvidia DGX Cloud AI supercomputing services — that give enterprises an end-to-end solution for creating custom generative AI models. Businesses can then deploy their customized models with Nvidia AI Enterprise software to power generative AI applications, including intelligent search, summarization and content generation.

“Enterprises need custom models to perform specialized skills trained on the proprietary DNA of their company — their data,” said Jensen Huang, founder and CEO of Nvidia. “Nvidia’s AI foundry service combines our generative AI model technologies, LLM training expertise and giant-scale AI factory. We built this in Microsoft Azure so enterprises worldwide can connect their custom model with Microsoft’s world-leading cloud services.”

The post Microsoft finally builds its own chips appeared first on TechWire Asia.

In March this year, the rotating chairman of Huawei, Eric Xu, made a bold claim that the tech restrictions imposed by Washington would ultimately strengthen China’s domestic semiconductor industry, rather than weakening it. “I believe China’s semiconductor industry will not sit idly by, but take efforts around… self-strengthening and self-reliance,” he said during a March 2023 press conference. He also boldly stated that China’s chip industry will be “reborn” due to US sanctions.

He wasn’t entirely wrong, and he wasn’t alone in his prediction. Months after the US imposed a stringent set of export controls on October 7, 2022, China started making strides, specifically in semiconductor technology. China even went as far as introducing a 5G smartphone featuring a domestically manufactured miniaturized silicon chip, which many analysts previously thought was unachievable by Chinese firms in the wake of the US-led technology ban.

The strides from China signaled the limits of the US-led sanctions. So, what could the US do to better constrain China’s expansion? The Biden administration targeted the hottest commodity in the market: AI chips.

Last month, the US Department of Commerce unveiled new rules intending to close loopholes left after last year’s restrictions on AI chip exports went into effect.

The earlier restrictions banned the sale of the Nvidia H100, the processor of choice for AI firms worldwide, especially in the US and China. Following that, Chinese companies could buy a slightly slowed-down version called the H800 or A800 that was explicitly made to comply with US restrictions (primarily by slowing down an on-device connection speed, called an interconnect).

The updated rules have banned those ‘slower’ chips as well, leaving Chinese companies scrambling for stockpiles. Experts were quick to explain how the restrictions would immediately cut off a significant and growing market for AI semiconductors. Nvidia thinks the restrictions might hurt its business in the long term, but what is certain amid all the US efforts to cripple China’s technological advancement is that they will actually boost Chinese developments rather than hampering them.

Huawei could fill the AI chips gap in China

Huawei, more widely recognized worldwide for its telecommunications and smartphone business, has been actively developing an AI chip product line for the past four years. It unveiled its push into AI technology at Huawei Connect 2018,  releasing two chips, neural networks compute architecture, a development toolkit, and a cloud training framework across AI.

What would have stood out if it were unveiled amid the current fascination around AI is Huawei’s Ascend AI chip series, which then included the Ascend 910 and Ascend 310. The chips were made available in 2019, the year Huawei was hit by the most severe US restrictions. 

Back then, Huawei asserted that its chip was the most powerful AI processor globally, and according to Chinese media reports, the initial Ascend 910 chip was fabricated using a 7-nanometer manufacturing process. With that process, the chip couldn’t significantly challenge Nvidia’s supremacy, either in domestic and international markets. 

https://www.youtube.com/watch?v=2DOX1dEsU6I

Nvidia launched its A100 and H100 chips in 2020 and 2022, capturing the lion’s share of the global AI chip market, a trend further fueled by the rise of generative AI. The main product to rival Nvidia’s A100 chip would later be the Ascend 910B.

Analysts have also noticed that Nvidia held a significant advantage over Huawei as an incumbent player, mainly due to the existing AI projects relying on Nvidia’s well-established software ecosystem. 

Although Huawei has its own ecosystem called CANN, experts say it has limitations when training AI models compared to Nvidia’s ecosystem. But there have been some changes in China of late, and Huawei might again be in the race to grab a slice of Nvidia’s AI chip market share.

According to a report by Reuters this week, Baidu, one of China’s leading AI companies, had ordered 1,600 of Huawei’s 910B Ascend AI chips. Baidu had been a long-time client of Nvidia, so the move demonstrates how Chinese firms are exploring alternatives to the latter’s products due to the export restrictions by the US. 

Reuters indicated that the order by Baidu was for 200 servers.

When was Huawei’s Ascend 910B launched?

Huawei hasn’t made an official announcement regarding the Ascend 910B, but it is undoubtedly a revised version of the Ascend 910. Specific details about the chip have surfaced through comments from Chinese firms and academics, along with technical information available on Huawei’s website.

In August this year, Liu Qingfeng, the chairman of the Chinese AI company iFlytek, lauded Huawei for developing a GPU that he likened to Nvidia’s A100. He mentioned that iFlytek was collaborating with Huawei on hardware development. Subsequently, Chinese media outlet Yicai reported that the hardware in question was powered by the previously undisclosed Ascend 910B chip.

Within the same month, documents related to Ascend 910B, such as driver and firmware upgrade guides, began appearing on Huawei’s website. Even during iFlyTek’s recent earnings call, senior vice president Jiang Tao reiterated that the capabilities of the Ascend 910B were comparable to Nvidia’s A100.

“Analysts and insiders suggest that the 910B chips offer comparable raw computing power to Nvidia’s chips but may lag slightly in overall performance. Nevertheless, they are considered the most advanced domestically-produced option available in China,” the report by Reuters reads.

Perhaps this was what Huawei’s chief financial officer Meng Wanzhou meant when she said the company would go big on AI by “building a solid computing base for China – and a second option for the world.” 

The post Huawei is prepared to challenge Nvidia’s position in the AI chip industry appeared first on TechWire Asia.