News related to
Nvidia

Su held talks with Young Hyun Jun, Samsung’s vice chairman and chief executive of its Device Solutions division, during a visit to South Korea that also included discussions with SK Hynix chief executive Kwak Noh-jung. The meetings placed high-bandwidth memory, broader DRAM supplies and possible foundry co-operation at the centre of AMD’s efforts to secure capacity for its expanding AI business.

Speaking to reporters in Seoul, Su said AMD continued to explore opportunities with Samsung across both memory and contract chip manufacturing. She described Samsung and SK Hynix as critical partners for AMD’s data-centre ambitions and urged memory producers to add capacity as quickly as possible.

The talks come as shortages across parts of the memory market complicate expansion plans for chip designers and data-centre operators. Demand for high-bandwidth memory, or HBM, has surged because AI accelerators require substantially greater memory bandwidth than conventional computing processors, while strong demand has also tightened supplies of standard DRAM products.

Su said the supply environment remained tight and AMD was working with customers to optimise their memory requirements so available capacity could support more computing. She said the company had built a strong supply chain but was working with suppliers to accelerate its expansion.

AMD expects demand for AI computing to remain elevated for several years and plans to increase chip supply substantially in 2027. Su said during a stop in Taiwan this week that the company had extended its capacity-planning horizon from one or two years to between three and five years as it works with manufacturing partners.

That planning encompasses advanced wafers, packaging, substrates and memory, all of which have become potential constraints as companies race to build AI infrastructure. AMD has been expanding its Instinct accelerator business as it competes with Nvidia for spending by cloud providers and other large data-centre customers.

Samsung already occupies an important position in AMD’s supply plans. The companies signed a memorandum of understanding in March covering next-generation AI memory and computing technologies, with Samsung designated a primary HBM4 supplier for AMD’s Instinct MI455X graphics processor.

Their agreement also covers advanced DRAM products for AMD’s sixth-generation EPYC processors, code-named Venice, and its Helios rack-scale AI platform. Samsung has supplied HBM3E memory for AMD’s MI350X and MI355X accelerators, extending a technology relationship between the companies that spans graphics, mobile and computing products.

HBM4 is particularly important to the next generation of AI hardware because increasingly powerful accelerators need faster transfers of enormous datasets while controlling energy consumption. Memory manufacturers have consequently directed more investment towards HBM, leaving capacity across other memory categories under pressure.

Samsung has been seeking to increase its position in the high-margin HBM market while SK Hynix remains another major supplier to AI chip developers. Strong AI demand has supported memory prices and earnings, though the pace of conventional DRAM price increases is expected to moderate after steep gains earlier this year.

AMD’s discussions with Samsung could also extend beyond memory. Their March agreement provided for talks on a possible foundry partnership under which Samsung could manufacture future AMD products. Su said on Wednesday that AMD was continuing to examine opportunities in that area, without announcing a manufacturing contract.

The possibility would give AMD another potential production option as advanced semiconductor capacity remains heavily sought after. Taiwan Semiconductor Manufacturing Co remains a central manufacturing partner for AMD, and Su met suppliers in Taiwan before travelling to South Korea as part of her capacity discussions.

The Hangzhou-based AI developer released Ascend versions of several key components on September 30, including its DeepGEMM matrix-multiplication library and DeepEP communications software. The projects are designed to work with Huawei’s Ascend processors while keeping public interfaces aligned with versions built around Nvidia hardware, reducing the amount of code developers may need to rewrite when moving workloads.

That portability targets a critical part of Nvidia’s competitive advantage. Nvidia’s CUDA software platform has become deeply embedded in AI development, giving programmers mature libraries, tools and optimisation techniques built over years. Competing chips therefore have to overcome not only differences in raw computing performance but also the cost and complexity of shifting software away from CUDA.

DeepSeek’s DeepJIT project takes a further step by providing a shared runtime interface for Nvidia CUDA graphics processors and Huawei Ascend neural-processing units. It handles compilation, caching, loading and launching of specialised AI kernels while retaining backend-specific code where necessary. The approach does not make the two architectures interchangeable, but it can reduce friction for developers maintaining software for both.

DeepGEMM-Ascend is fully API-compatible with DeepSeek’s original DeepGEMM package and supports BF16, FP8 and FP4 calculations used in demanding AI workloads. DeepSeek says the library uses Ascend-specific optimisation methods to approach the hardware’s performance limits while shielding programmers from lower-level complexities including memory layouts, alignment constraints and address calculations.

Its DeepEP-Ascend library similarly aligns its public buffer interfaces with the Nvidia version while using Huawei’s communications technologies underneath. DeepSeek has also added Ascend support to DeepSelect, a specialised component used for sparse-attention and sampling operations, and released Ascend sparse-attention kernels for its model infrastructure.

Huawei, meanwhile, has been opening more of the software surrounding Ascend. Its CANN platform — the company’s counterpart to the programming infrastructure required to exploit AI accelerators — supports frameworks including PyTorch and works with open-source projects such as Triton, vLLM and veRL.

Huawei said last month that CANN had moved to sustained community-driven open-source development, with external programmers accounting for 61 per cent of its CANN developers. The company said the community had more than 5,200 monthly active developers, while more than 40 models had been natively pre-trained using Ascend and CANN.

The combination matters because China’s challenge to Nvidia has increasingly shifted from producing an alternative accelerator to assembling a usable full computing stack. Hardware can win orders only if model developers can train and run workloads efficiently without sacrificing too much engineering time or performance.

DeepSeek’s software also gives Huawei an influential independent developer helping test and improve that stack. The AI laboratory became prominent for building capable models with an emphasis on computational efficiency, and its infrastructure code is closely watched by developers seeking lower-cost approaches to training and inference.

The effort comes as US export controls continue to restrict China’s access to Nvidia’s most advanced processors, creating stronger incentives for technology groups to adopt domestic alternatives. Those restrictions have helped expand the strategic importance of Huawei’s Ascend line, even though China still faces constraints in advanced semiconductor manufacturing.

Nvidia nevertheless retains formidable advantages. CUDA has a vast installed developer base and an extensive collection of mature libraries, while Nvidia’s leading accelerators remain central to major AI training systems outside China. Compatibility layers and open-source libraries also require sustained testing, documentation and optimisation before they can match the reliability of an established platform across diverse workloads.

A first of its scale for a Southeast Asian-founded company, the collaboration marks a major milestone that will significantly boost Southeast Asia’s profile in the global AI ecosystem. By combining Aolani’s capabilities and NVIDIA’s technology, the strategic collaboration will unlock financing that will further accelerate and democratise access to AI compute in Asia for global AI natives, sovereigns, and local enterprises and SMEs.

This collaboration brings Aolani’s total AI factory capacity to more than 100 megawatts (MW). 22,000 GPUs are expected to be deployed across two AI factories in Malaysia and the Philippines in early 2027. This will bring Aolani’s total deployment to over 48,000 GPUs.

An NVIDIA Cloud Partner since 2023, Aolani will be deploying NVIDIA Blackwell Ultra GPUs, a leading AI system designed to help organisations build and scale advanced AI workloads today.

Aolani will deliver NVIDIA-powered cloud services to AI-native companies, enterprises, startups, developers and institutions across Southeast Asia. A revenue-sharing and credit-support model will align infrastructure deployment with customer demand, helping broaden access to large-scale accelerated computing while giving Aolani a capital-efficient path to scale its AI factories across the region.

Driving Southeast Asia as a leading global powerhouse for AI innovation

Southeast Asia is one of the most robust AI markets in the world, with AI adoption rates among companies demonstrating stronger momentum than global average^[1]. With more than US$50 billion invested by global hyperscalers in AI-ready data centres and cloud infrastructure^[2], the region’s growth is driven by its emerging position as a global arena for cloud and AI innovation.

While Malaysia has established itself as one of Asia’s premier AI factory hubs, the Philippines represents a high-potential emerging market for AI infrastructure. Aolani’s deployment in the Philippines is expected to be among the first large-scale AI infrastructure platforms in the country, broadening access to advanced AI compute and supporting the buildout of essential AI infrastructure in Asia.

Nicholas Chia, Chief Executive Officer at Aolani said: “This collaboration with NVIDIA marks a defining moment in Aolani’s mission to deliver world-class AI compute at the speed that matches the pace of AI development, and on infrastructure that is compliant, accountable, and trusted to perform. Southeast Asia is rapidly emerging as a key global node for the next wave of global AI buildout, and we are well-positioned to lead this. We are excited to work with new partners and customers to build the Southeast Asian AI ecosystem and cement the region’s central role in the global AI landscape.”

Media Contact

H/Advisors on behalf of Aolani
[email protected]

Sources:

[1] McKinsey & Company, “AI in Southeast Asia: An Era of Opportunity,” 2024, https://www.mckinsey.com/featured-insights/future-of-asia/ai-in-southeast-asia-an-era-of-opportunity

[2] Singapore Economic Development Board, “AI in Southeast Asia: An Era of Opportunity,” https://www.edb.gov.sg/en/business-insights/market-and-industry-reports/ai-in-southeast-asia-an-era-of-opportunity.html

https://www.aolani.com
https://sg.linkedin.com/company/aolani-cloud

The issuer is solely responsible for the content of this announcement.

About Aolani

Where AI gets built in Asia. Headquartered in Singapore and founded in 2023, Aolani is backed by compliant and purpose-built infrastructure to deliver the performance capabilities for next-generation AI. Aolani’s AI factories enable organisations to build with confidence, scale ambitiously, and move at hyper-speed in the world’s fastest-growing AI market.

For more information, visit www.aolani.com and follow @AolaniCloud on LinkedIn.

Rotating chairman David Wang said at Huawei Connect 2026 in Shanghai on Thursday that the company plans to release the 960DT in the first quarter of 2027 and the Ascend 960PR in the third quarter. The roadmap places Huawei’s next generation of accelerators at the centre of its effort to build domestic alternatives for training and running increasingly demanding AI models.

Huawei also unveiled the Ascend 960 SuperPoD and an upgraded version of its UnifiedBus interconnect technology, designed to link large numbers of processors so they can function as one computing system. The company said a combination of UnifiedBus and its Hi-ONE optical system could connect as many as 4,000 processors within a SuperPoD, extending high-speed optical links deeper into the machine.

The strategy reflects a broader shift towards system-level performance rather than relying solely on the capability of individual chips. Advanced AI workloads require thousands of processors to exchange data rapidly, making networking bandwidth, memory access and cluster efficiency increasingly important alongside raw chip performance.

Wang said Huawei has developed 11 semiconductors based on UnifiedBus technology for use across its large computing systems. Its biggest planned superclusters will be able to support up to one million AI processors, he said, while the company has already shipped more than 1,000 smaller supernodes to more than 370 customers.

Huawei did not identify those customers or give a detailed breakdown of shipments. It also did not provide independently verified benchmarks directly comparing the new 960-series chips with Nvidia’s leading accelerators, leaving questions over single-chip performance, software compatibility and manufacturing scale.

The company had previously said the Ascend 960 would offer twice the computing power, memory capacity, memory bandwidth and interconnect ports of the Ascend 950 generation. Huawei has also been developing the Atlas 960 SuperPoD and SuperCluster, with the larger system intended to combine more than one million neural processing units when it reaches the market.

Huawei’s current product roadmap also includes the Ascend 950DT, due in the fourth quarter of 2026. The company has said that processor will use its HiZQ 2.0 high-bandwidth memory, providing 144 GB of capacity and 4 terabytes per second of bandwidth. Those specifications are intended to support both model training and the decode stage of inference workloads.

US restrictions have limited China’s access to cutting-edge Nvidia processors and advanced semiconductor manufacturing equipment, giving domestic suppliers greater room to expand. Huawei has emerged as the most prominent of those suppliers, but Nvidia retains a major advantage through its CUDA software ecosystem and a broad global developer base.

Chinese AI developers have increasingly turned to domestic accelerators for inference, where trained models generate responses, even as many frontier laboratories continue to depend on Nvidia hardware for the most demanding training jobs. That split has made Huawei’s ability to improve both hardware and software a critical test of whether China can reduce its dependence on foreign computing infrastructure.

Huawei is also contending with shortages of high-bandwidth memory, a key component in AI accelerators. The company has designed memory for some newer Ascend products, but supply constraints across the industry have increased costs and complicated efforts to scale production quickly enough to meet demand.

The group is seeking to offset those limitations through architectural changes, advanced packaging, optical networking and tighter integration between chips, systems and software. Its SuperPoD approach is intended to compensate for weaker performance from individual processors by allowing much larger numbers of chips to work together with lower latency.

Bae said South Korea was still trying to close the gap with leading AI powers while creating new markets, making its position different from countries or companies at the technological frontier. “We cannot afford to slow down AI right now,” he wrote on social media, describing the government’s approach as one of “speed responsibility” rather than speed control.

His remarks place Seoul firmly on the acceleration side of a widening debate over whether developers should moderate the pace of capable AI systems. Anthropic chief executive Dario Amodei has called for a slowdown in frontier model development to allow more time for safety work, while OpenAI chief executive Sam Altman and other technology leaders have voiced support for stronger precautions as capabilities advance.

Bae said South Korea should continue expanding research, industrial adoption and home-grown model development while strengthening protections against deepfakes, cyber threats and privacy violations. Faster progress, he said, required safety and trust measures to move at the same pace rather than follow after deployment.

The minister pointed to a broader technology agenda that includes sovereign foundation models, systems capable of processing vision and language, and models able to interact with physical environments. He said the country also needed to prepare for AI becoming deeply embedded across industries and everyday life.

The government has made domestic foundation-model development a central part of its ambition to place South Korea among the world’s top three AI powers. The Ministry of Science and ICT has been running a sovereign AI foundation model programme intended to improve domestic capabilities and reduce dependence on foreign technology.

Three teams led by Upstage, SK Telecom and LG AI Research advanced after the programme’s second-stage evaluation this year. The ministry has said models developed under the initiative were built with comparatively limited resources but were released as open source as part of an effort to expand the domestic AI ecosystem.

Bae indicated that further details on the next phase of the foundation-model programme would be announced after ongoing reviews. His comments also referenced government work on security-focused AI models and frontier systems, alongside an “AI for All” initiative intended to spread adoption across the economy.

The policy push is being developed alongside South Korea’s AI Basic Act, which establishes rules covering both the development of artificial intelligence and the creation of a framework for trust. The law’s stated purpose includes protecting rights and human dignity while supporting quality of life and national competitiveness.

That legal structure gives Seoul a basis for pursuing faster adoption without presenting safety as a separate objective. Bae said competitiveness and sustainability were linked, arguing that development speed and safeguards should advance together.

The global dispute has intensified as leading executives and policymakers weigh the economic benefits of faster AI deployment against concerns over misuse and loss of control. Amodei has proposed independent safety evaluators, coordination among major laboratories and greater international cooperation, while Nvidia chief executive Jensen Huang has argued against slowing technological progress and favours advancing safety alongside innovation.

European Commission President Ursula von der Leyen on Wednesday backed calls for greater caution and said frontier AI laboratories would be invited for talks on managing risks. Meta chief executive Mark Zuckerberg, by contrast, said AI companies already had strong incentives to build systems safely and did not require an industry-wide slowdown.

Bae’s intervention highlights the competitive pressures facing countries outside the United States and China, where most frontier AI investment and model development remains concentrated. South Korea has strong semiconductor and technology industries, but its government has acknowledged the need to strengthen domestic models, computing capacity and commercial adoption to narrow the gap.

Foreign Ministry spokesman Guo Jiakun said on Monday that artificial intelligence was a consequential technology for humanity and should be developed openly and inclusively. Asked about calls from Dario Amodei, Sam Altman and Elon Musk for companies to slow frontier development, Guo warned that “fear-mongering, confrontation and vicious competition” would hinder sound global AI governance.

The response came as Washington’s debate over AI safety widened into a strategic argument about China. Anthropic chief executive Amodei has urged tougher safeguards for advanced systems and has argued that a Chinese lead in frontier AI could create national security risks for the United States. OpenAI chief executive Altman and Musk have also called for stronger precautions as systems become more capable.

Trump, however, has resisted proposals that could materially slow US development. He said this week that America already had tools to police the industry and stressed the importance of maintaining leadership over China. House Speaker Mike Johnson similarly rejected a moratorium, saying any pause could hand Beijing a competitive advantage and create national security risks.

The dispute places AI governance among the issues likely to shadow the Trump-Xi meeting scheduled for September 24 in Washington. The leaders are also expected to confront trade disputes, semiconductor restrictions and China’s controls on rare-earth exports, all of which affect the broader contest over computing capacity and advanced manufacturing.

Treasury Secretary Scott Bessent is due to meet Vice-Premier He Lifeng before the summit.

China has presented a different framework for AI governance, emphasising access, development and multilateral rules while opposing measures it regards as technological containment. At the BRICS summit this month, Xi proposed an open-source and inclusive AI initiative and said China would help establish a BRICS AI open-source community and digital ecosystem platform.

Beijing has nevertheless acknowledged the risks posed by increasingly powerful systems. Chinese officials have called for stronger global governance and risk awareness, while arguing that safety should not become a justification for blocking technological diffusion. The government has also supported the establishment of the World Artificial Intelligence Cooperation Organization, whose founding agreement was signed in Shanghai in July by 29 countries.

The argument over safety is closely linked to semiconductor controls. Washington has maintained restrictions on the export of advanced computing technology to China, although the Commerce Department in January shifted applications for Nvidia H200, AMD MI325X and comparable chips to case-by-case review under specified security conditions. Beijing has repeatedly criticised US technology restrictions as discriminatory and has threatened countermeasures when Chinese companies are targeted.

China’s Commerce Ministry last week also challenged a US cybersecurity advisory accusing China-based AI companies of using distillation campaigns against American models. The ministry said China was prepared for an intergovernmental AI dialogue with the United States but would respond if Washington used such allegations to suppress Chinese firms.

That exchange underscored the difficulty of separating AI safety from the strategic rivalry. US officials and technology executives increasingly frame advanced models as both commercial assets and national security capabilities, while China portrays broad restrictions as attempts to preserve US dominance.

There are areas of overlap. Both governments have said advanced AI presents risks requiring governance, and the two sides have agreed to launch an intergovernmental dialogue on the technology. Informal discussions among researchers and policy specialists have also continued, though official positions remain far apart on access to chips, model controls and the balance between safety and development.

The September 24 summit follows Trump’s May visit to China, when the two leaders agreed to pursue what the White House described as strategic stability and set up new bilateral trade and investment mechanisms. Artificial intelligence was not the centrepiece of those agreements, but it has since moved higher on the bilateral agenda as concerns over model capabilities, export controls and industrial competition have intensified.

Enflame’s shares closed 179 per cent above their initial public offering price on the STAR Market, lifting the company’s market value to about 171 billion yuan from 61.2 billion yuan at the offer price. The Tencent-backed company raised 6.12 billion yuan by selling 43.04 million new shares at 142.18 yuan each.

The surge came despite Enflame having never reported an annual profit since its establishment in 2018. Its 2025 revenue reached 990.2 million yuan while its net loss narrowed to about 1.2 billion yuan. The company has forecast revenue of 2.3 billion yuan to 3 billion yuan for the first nine months of 2026, alongside a net loss of 700 million yuan to 860 million yuan.

Revenue has expanded quickly from 301 million yuan in 2023 and about 722 million yuan in 2024, but losses have remained substantial as the company spends heavily on research and product development. Enflame recorded net losses of about 1.67 billion yuan in 2023 and 1.51 billion yuan in 2024 before narrowing the deficit last year, underscoring how far commercial growth has run ahead of earnings for the company.

Enflame said in its prospectus that it expects to reach break-even or profitability in 2026 or 2027. It plans to direct much of the IPO proceeds towards development and industrialisation of fifth- and sixth-generation artificial-intelligence chips, software and large-scale computing systems.

The company’s rapid revaluation also highlights the unusual economics of China’s artificial-intelligence chip boom. Investors have rewarded several young semiconductor groups whose sales are expanding quickly but whose research spending remains heavy and earnings records are limited or negative.

Moore Threads Technology, another Shanghai-listed graphics processor developer, created several billionaires after its market debut in December 2025, when its shares rose more than fourfold. Founder and chairman Zhang Jianzhong, a former head of Nvidia’s China operations, was valued at several billion dollars after the listing, while co-founders Zhang Yubo and Wang Dong also crossed the billionaire threshold based on their holdings.

MetaX Integrated Circuits produced a similar outcome after its Shanghai flotation. Chairman and chief executive Chen Weiliang, who previously worked at Advanced Micro Devices, became a billionaire as investors pushed the company’s valuation sharply higher despite continuing operating losses. Early investor Ge Weidong also joined the billionaire ranks through his stake.

The valuations reflect expectations that domestic chipmakers will capture a larger share of China’s fast-growing demand for artificial-intelligence computing as access to the most advanced US-designed processors remains constrained. Domestic producers are competing to supply cloud companies, data centres and developers running large language models, recommendation engines and other computing-intensive services.

Enflame’s commercial position, however, remains unusually concentrated. Tencent is its largest shareholder, with a 17.95 per cent stake after the offering, and sales linked to the technology group accounted for 83.79 per cent of Enflame’s 2025 revenue. The concentration gives Enflame a major anchor customer but also leaves its growth closely tied to purchasing decisions by one group.

The company employed about 860 people as of September 2025, according to disclosures compiled from its listing materials. Founded by former AMD executives Zhao Lidong and Zhang Yalin, Enflame develops accelerator chips, add-on cards, computing systems and software for artificial-intelligence training and inference in data centres.

Only 17.9 million shares, equivalent to 4.16 per cent of Enflame’s enlarged share capital, were freely tradable at the start of trading, limiting the initial public float and potentially amplifying price swings. The retail portion of the offering attracted thousands of times more demand than the shares available.

The government is seriously evaluating Huawei’s AI accelerators for the initiative, although no final procurement decision has been announced and the number of chips under consideration remains unclear. The project would be operated by state-linked Telekom Malaysia, which was selected through a public tender and already works with Huawei on cloud infrastructure.

The deliberations put Malaysia at the centre of a widening contest between Washington and Beijing over the technology underpinning artificial intelligence. US officials have warned that use of Huawei’s Ascend 910C could violate American export-control rules, while Malaysia has maintained that technology decisions should serve its national interests and comply with applicable laws.

Huawei’s 910C is the company’s flagship AI accelerator currently available at scale. Malaysia is considering that product rather than the newer Ascend 950 series, which remains in limited production. Huawei has sought to expand sales of its AI systems beyond China as US restrictions continue to limit its access to advanced foreign semiconductor manufacturing technology.

Telekom Malaysia’s existing AI computing service uses Nvidia processors, highlighting the competing technology options facing Kuala Lumpur. The government is also discussing another tender process that could give Nvidia and other US suppliers a more direct opportunity to compete for the sovereign AI contract.

Prime Minister Anwar Ibrahim has repeatedly argued that Malaysia should build domestic technological capability rather than remain solely a consumer of systems developed overseas. His government formally established AI Malaysia Berhad under the Digital Ministry in July to lead its ambition of becoming an “AI Nation” by 2030 and to strengthen national AI development, governance and deployment.

Sovereign AI has become particularly important to the government because the proposed infrastructure is intended to process information that could include sensitive government, military and intelligence material. Kuala Lumpur wants such data to remain within Malaysia, reducing dependence on foreign-controlled computing environments and giving national institutions greater authority over storage and access.

Anwar has also questioned the implications of the US Cloud Act, which allows American courts in specified circumstances to compel US technology companies to provide electronic data even when it is stored abroad. Speaking to university students in April, he described the law as unreasonable while acknowledging the difficulty of challenging Washington on the issue.

US officials have raised separate security concerns about Huawei equipment being used to handle intelligence-related material, including information supplied by foreign governments. A State Department spokesperson said secure technology was essential to economic prosperity and warned that suppliers susceptible to government manipulation could pose risks to critical infrastructure, privacy, intellectual property and national security. The statement did not name China or Huawei.

Malaysia’s National Security Council officials are expected to visit Washington later this month for discussions that will include the technology issue. The talks come as the United States promotes American AI hardware, software and applications across developing markets and considers government-backed financing packages to strengthen the international position of US suppliers.

The current evaluation follows an earlier controversy in May 2025, when a deputy communications minister said a Malaysian AI initiative would deploy 3,000 Huawei Ascend chips. Her remarks were withdrawn shortly afterwards, and the Investment, Trade and Industry Ministry clarified that the project concerned was privately driven, not endorsed or coordinated by the government.

At the time, the ministry said any government AI infrastructure programme would require legal, operational and reputational due diligence. It also reaffirmed Malaysia’s commitment to applicable export-control laws while asserting the country’s sovereign right to formulate policy according to national interests.

Malaysia’s National AI Action Plan for 2026-2030 calls for responsible adoption, stronger domestic capabilities and globally competitive locally developed solutions. The programme links AI deployment to economic productivity, public-sector transformation and governance, placing infrastructure choices within a strategy.