While limited quantities have already been delivered, mass deployment would mark an important step for Chinese firms seeking domestic alternatives to US-made semiconductors.

The move comes at a time when Chinese developers face tighter restrictions on access to Nvidia hardware. The US government recently informed Nvidia that sales of its H20 AI chip to China require an export licence. That’s left developers in China looking for options that can support large-scale training and inference workloads.

The Huawei Ascend 910C chip isn’t built on the most advanced process nodes, but it represents a workaround. The chip is essentially a dual-package version of the earlier 910B, with two processors to double the performance and memory. Sources familiar with the chip say it performs comparably to Nvidia’s H100.

Rather than relying on cutting-edge manufacturing, Huawei has adopted a brute-force approach, combining multiple chips and high-speed optical interconnects to scale up performance. This approach is central to Huawei’s CloudMatrix 384 system, a full rack-scale AI platform for training large models.

The CloudMatrix 384 features 384 Huawei Ascend 910C chips deployed in 16 racks comprising of 12 compute racks and four networking. Unlike copper-based systems, Huawei’s platform is uses optical interconnects, enabling high-bandwidth communication between components of the system. According to analysis from SemiAnalysis, the architecture includes 6,912 800G LPO optical transceivers to form an optical all-to-all mesh network.

This allows Huawei’s system to deliver approximately 300 petaFLOPs of BF16 compute power – outpacing Nvidia’s GB200 NVL72 system, which reaches around 180 BF16 petaFLOPs. The CloudMatrix also claims advantages in higher memory bandwidth and capacity, offering more than double the bandwidth and over 3.6 times the high-bandwidth memory (HBM) capacity.

The gains, however, are not without drawbacks. The Huawei system is predicted to be 2.3 times less efficient per floating point operation than Nvidia’s GB200 and has lower power efficiency per unit of memory bandwidth and capacity. Despite the lower performance per watt, Huawei’s system still provides the infrastructure needed to train advanced AI models at scale.

Sources indicate that China’s largest chip foundry, SMIC, is producing some of the main components for the 910C using its 7nm N+2 process. Yield levels remain a concern, however, and some of the 910C units reportedly include chips produced by TSMC for Chinese firm Sophgo. Huawei has denied using TSMC-made parts.

The US Commerce Department is currently investigating the relationship between TSMC and Sophgo after a Sophgo-designed chip was found in Huawei’s earlier 910B processor. TSMC has maintained that it has not supplied Huawei since 2020 and continues to comply with export regulations.

In late 2023, Huawei began distributing early samples of the 910C to selected technology firms and opened its order books. Consulting firm Albright Stonebridge Group suggested the chip is likely to become the go-to choice for Chinese companies building large AI models or deploying inference capacity, given the ongoing export controls on US-made chips.

While the Huawei Ascend 910C may not match Nvidia in power efficiency or process technology, it signals a broader trend. Chinese technology firms are developing homegrown alternatives to foreign components, even if it means using less advanced methods to achieve similar outcomes.

As global AI demand surges and export restrictions tighten, Huawei’s ability to deliver a scalable AI hardware solution domestically could help shape China’s artificial intelligence future – especially as developers look to secure long-term supply chains and reduce exposure to geopolitical risk.

(Photo via Unsplash)

See also: Huawei’s AI hardware breakthrough challenges Nvidia’s dominance

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 Huawei to begin mass shipments of Ascend 910C amid US curbs appeared first on AI News.

The Alibaba-owned company has used chips from domestic suppliers, including those tied to its parent, Alibaba, and Huawei Technologies to train large language models using the Mixture of Experts (MoE) method. The results were reportedly comparable to those produced with Nvidia’s H800 chips, sources claim. While Ant continues to use Nvidia chips for some of its AI development, one sources said the company is turning increasingly to alternatives from AMD and Chinese chip-makers for its latest models.

The development signals Ant’s deeper involvement in the growing AI race between Chinese and US tech firms, particularly as companies look for cost-effective ways to train models. The experimentation with domestic hardware reflects a broader effort among Chinese firms to work around export restrictions that block access to high-end chips like Nvidia’s H800, which, although not the most advanced, is still one of the more powerful GPUs available to Chinese organisations.

Ant has published a research paper describing its work, stating that its models, in some tests, performed better than those developed by Meta. Bloomberg News, which initially reported the matter, has not verified the company’s results independently. If the models perform as claimed, Ant’s efforts may represent a step forward in China’s attempt to lower the cost of running AI applications and reduce the reliance on foreign hardware.

MoE models divide tasks into smaller data sets handled by separate components, and have gained attention among AI researchers and data scientists. The technique has been used by Google and the Hangzhou-based startup, DeepSeek. The MoE concept is similar to having a team of specialists, each handling part of a task to make the process of producing models more efficient. Ant has declined to comment on its work with respect to its hardware sources.

Training MoE models depends on high-performance GPUs which can be too expensive for smaller companies to acquire or use. Ant’s research focused on reducing that cost barrier. The paper’s title is suffixed with a clear objective: Scaling Models “without premium GPUs.” [our quotation marks]

The direction taken by Ant and the use of MoE to reduce training costs contrast with Nvidia’s approach. CEO Officer Jensen Huang has said that demand for computing power will continue to grow, even with the introduction of more efficient models like DeepSeek’s R1. His view is that companies will seek more powerful chips to drive revenue growth, rather than aiming to cut costs with cheaper alternatives. Nvidia’s strategy remains focused on building GPUs with more cores, transistors, and memory.

According to the Ant Group paper, training one trillion tokens – the basic units of data AI models use to learn – cost about 6.35 million yuan (roughly $880,000) using conventional high-performance hardware. The company’s optimised training method reduced that cost to around 5.1 million yuan by using lower-specification chips.

Ant said it plans to apply its models produced in this way – Ling-Plus and Ling-Lite – to industrial AI use cases like healthcare and finance. Earlier this year, the company acquired Haodf.com, a Chinese online medical platform, to further Ant’s ambition to deploy AI-based solutions in healthcare. It also operates other AI services, including a virtual assistant app called Zhixiaobao and a financial advisory platform known as Maxiaocai.

“If you find one point of attack to beat the world’s best kung fu master, you can still say you beat them, which is why real-world application is important,” said Robin Yu, chief technology officer of Beijing-based AI firm, Shengshang Tech.

Ant has made its models open source. Ling-Lite has 16.8 billion parameters – settings that help determine how a model functions – while Ling-Plus has 290 billion. For comparison, estimates suggest closed-source GPT-4.5 has around 1.8 trillion parameters, according to MIT Technology Review.

Despite progress, Ant’s paper noted that training models remains challenging. Small adjustments to hardware or model structure during model training sometimes resulted in unstable performance, including spikes in error rates.

(Photo by Unsplash)

See also: DeepSeek V3-0324 tops non-reasoning AI models in open-source first

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 Ant Group uses domestic chips to train AI models and cut costs appeared first on AI News.

In recent submissions to the US government, the companies warned that Chinese models, such as DeepSeek R1, are becoming more sophisticated and competitive. The submissions, filed in March 2025 in response to a request for input on an AI Action Plan, highlight the growing challenge from China in technological capability and price.

China’s growing AI presence

Chinese state-supported AI model DeepSeek R1 has piqued the interest of US developers. According to OpenAI, DeepSeek demonstrates that the technological gap between the US and China is narrowing. The company described DeepSeek as “state-subsidised, state-controlled, and freely available,” raises concerns about the model’s ability to influence global AI development.

OpenAI compared DeepSeek to Chinese telecommunications company Huawei, warning that Chinese regulations could allow the government to compel DeepSeek to compromise sensitive US systems or infrastructure. Concerns about data privacy were also raised, with OpenAI pointing out that Chinese rules could force DeepSeek to disclose user data to the government, and enhance China’s ability to develop more advanced AI systems.

The competition from China also includes Ernie X1 and Ernie 4.5, released by Baidu, which are designed to compete with Western systems.

According to Baidu, Ernie X1 “delivers performance on par with DeepSeek R1 at only half the price.” Meanwhile, Ernie 4.5 is priced at just 1% of OpenAI’s GPT-4.5 while outperforming it in multiple benchmarks.

DeepSeek’s aggressive pricing strategy is also raising concerns with the US companies. According to Bernstein Research, DeepSeek’s V3 and R1 models are priced “anywhere from 20-40x cheaper” than equivalent models from OpenAI. The pricing pressure could force US developers to adjust their business models to remain competitive.

Baidu’s strategy of open-sourcing its models is also gaining traction. “One thing we learned from DeepSeek is that open-sourcing the best models can greatly help adoption,” Baidu CEO Robin Li said in February. Baidu plans to open-source the Ernie 4.5 series starting June 30, which could accelerate adoption and further increase competitive pressure on US firms.

Cost aside, early user feedback on Baidu’s models has been positive. “[I’ve] been playing around with it for hours, impressive performance,” Alvin Foo, a venture partner at Zero2Launch, said in a post on social media, suggesting China’s AI models are becoming more affordable and effective.

US AI security and economic risks

The submissions also highlight what the US companies perceive as risks to security and the economy.

OpenAI warned that Chinese regulations could allow the government to compel DeepSeek to manipulate its models to compromise infrastructure or sensitive applications, creating vulnerabilities in important systems.

Anthropic’s concerns centred on biosecurity. It disclosed that its own Claude 3.7 Sonnet model demonstrated capabilities in biological weapon development, highlighting the dual-use nature of AI systems.

Anthropic also raised issues with US export controls on AI chips. While Nvidia’s H20 chips meet US export restrictions, they nonetheless perform well in text generation – a important feature for reinforcement learning. Anthropic called on the government to tighten controls to prevent China from gaining a technological edge using the chips.

Google took a more cautious approach, acknowledging security risks yet warned against over-regulation. The company argues that strict AI export rules could harm US competitiveness by limiting business opportunities for domestic cloud providers. Google recommended targeted export controls to protect national security but without disruption to its business operations.

Maintaining US AI competitiveness

All US three companies emphasised the need for better government oversight and infrastructure investment to maintain US AI leadership.

Anthropic warned that by 2027, training a single advanced AI model could require up to five gigawatts of power – enough to power a small city. The company proposed a national target to build 50 additional gigawatts of AI-dedicated power capacity by 2027 and to streamline regulations around power transmission infrastructure.

OpenAI positioned the competition between US and Chinese AI as a contest between democratic and authoritarian AI models. The company argued that promoting a free-market approach would drive better outcomes and maintain America’s technological edge.

Google focused on urging practical measures, including increased federal funding for AI research, improved access to government contracts, and streamlined export controls. The company also recommended more flexible procurement rules to accelerate AI adoption by federal agencies.

Regulatory strategies for US AI

The US companies called for a unified federal approach to AI regulation.

OpenAI proposed a regulatory framework managed by the Department of Commerce, warning that fragmented state-level regulations could drive AI development overseas. The company supported a tiered export control framework, allowing broader access to US-developed AI in democratic countries while restricting it in authoritarian states.

Anthropic called for stricter export controls on AI hardware and training data, warning that even minor improvements in model performance could give China a strategic advantage.

Google focused on copyright and intellectual property rights, stressing that its interpretation of ‘fair use’ is important for AI development. The company warned that overly restrictive copyright rules could disadvantage US AI firms compared to their Chinese competitors.

All three companies stressed the need for faster government adoption of AI. OpenAI recommended removing some existing testing and procurement barriers, while Anthropic supported streamlined procurement processes. Google emphasised the need for improved interoperability in government cloud infrastructure.

See also: The best AI prompt generator: Create perfect AI prompts

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 Is America falling behind in the AI race? appeared first on AI News.

Efficiently managing and coordinating AI inference requests across a fleet of GPUs is a critical endeavour to ensure that AI factories can operate with optimal cost-effectiveness and maximise the generation of token revenue.

As AI reasoning becomes increasingly prevalent, each AI model is expected to generate tens of thousands of tokens with every prompt, essentially representing its “thinking” process. Enhancing inference performance while simultaneously reducing its cost is therefore crucial for accelerating growth and boosting revenue opportunities for service providers.

A new generation of AI inference software

NVIDIA Dynamo, which succeeds the NVIDIA Triton Inference Server, represents a new generation of AI inference software specifically engineered to maximise token revenue generation for AI factories deploying reasoning AI models.

Dynamo orchestrates and accelerates inference communication across potentially thousands of GPUs. It employs disaggregated serving, a technique that separates the processing and generation phases of large language models (LLMs) onto distinct GPUs. This approach allows each phase to be optimised independently, catering to its specific computational needs and ensuring maximum utilisation of GPU resources.

“Industries around the world are training AI models to think and learn in different ways, making them more sophisticated over time,” stated Jensen Huang, founder and CEO of NVIDIA. “To enable a future of custom reasoning AI, NVIDIA Dynamo helps serve these models at scale, driving cost savings and efficiencies across AI factories.”

Using the same number of GPUs, Dynamo has demonstrated the ability to double the performance and revenue of AI factories serving Llama models on NVIDIA’s current Hopper platform. Furthermore, when running the DeepSeek-R1 model on a large cluster of GB200 NVL72 racks, NVIDIA Dynamo’s intelligent inference optimisations have shown to boost the number of tokens generated by over 30 times per GPU.

To achieve these improvements in inference performance, NVIDIA Dynamo incorporates several key features designed to increase throughput and reduce operational costs.

Dynamo can dynamically add, remove, and reallocate GPUs in real-time to adapt to fluctuating request volumes and types. The software can also pinpoint specific GPUs within large clusters that are best suited to minimise response computations and efficiently route queries. Dynamo can also offload inference data to more cost-effective memory and storage devices while retrieving it rapidly when required, thereby minimising overall inference costs.

NVIDIA Dynamo is being released as a fully open-source project, offering broad compatibility with popular frameworks such as PyTorch, SGLang, NVIDIA TensorRT-LLM, and vLLM. This open approach supports enterprises, startups, and researchers in developing and optimising novel methods for serving AI models across disaggregated inference infrastructures.

NVIDIA expects Dynamo to accelerate the adoption of AI inference across a wide range of organisations, including major cloud providers and AI innovators like AWS, Cohere, CoreWeave, Dell, Fireworks, Google Cloud, Lambda, Meta, Microsoft Azure, Nebius, NetApp, OCI, Perplexity, Together AI, and VAST.

NVIDIA Dynamo: Supercharging inference and agentic AI

A key innovation of NVIDIA Dynamo lies in its ability to map the knowledge that inference systems hold in memory from serving previous requests, known as the KV cache, across potentially thousands of GPUs.

The software then intelligently routes new inference requests to the GPUs that possess the best knowledge match, effectively avoiding costly recomputations and freeing up other GPUs to handle new incoming requests. This smart routing mechanism significantly enhances efficiency and reduces latency.

“To handle hundreds of millions of requests monthly, we rely on NVIDIA GPUs and inference software to deliver the performance, reliability and scale our business and users demand,” said Denis Yarats, CTO of Perplexity AI.

“We look forward to leveraging Dynamo, with its enhanced distributed serving capabilities, to drive even more inference-serving efficiencies and meet the compute demands of new AI reasoning models.”

AI platform Cohere is already planning to leverage NVIDIA Dynamo to enhance the agentic AI capabilities within its Command series of models.

“Scaling advanced AI models requires sophisticated multi-GPU scheduling, seamless coordination and low-latency communication libraries that transfer reasoning contexts seamlessly across memory and storage,” explained Saurabh Baji, SVP of engineering at Cohere.

“We expect NVIDIA Dynamo will help us deliver a premier user experience to our enterprise customers.”

Support for disaggregated serving

The NVIDIA Dynamo inference platform also features robust support for disaggregated serving. This advanced technique assigns the different computational phases of LLMs – including the crucial steps of understanding the user query and then generating the most appropriate response – to different GPUs within the infrastructure.

Disaggregated serving is particularly well-suited for reasoning models, such as the new NVIDIA Llama Nemotron model family, which employs advanced inference techniques for improved contextual understanding and response generation. By allowing each phase to be fine-tuned and resourced independently, disaggregated serving improves overall throughput and delivers faster response times to users.

Together AI, a prominent player in the AI Acceleration Cloud space, is also looking to integrate its proprietary Together Inference Engine with NVIDIA Dynamo. This integration aims to enable seamless scaling of inference workloads across multiple GPU nodes. Furthermore, it will allow Together AI to dynamically address traffic bottlenecks that may arise at various stages of the model pipeline.

“Scaling reasoning models cost effectively requires new advanced inference techniques, including disaggregated serving and context-aware routing,” stated Ce Zhang, CTO of Together AI.

“The openness and modularity of NVIDIA Dynamo will allow us to seamlessly plug its components into our engine to serve more requests while optimising resource utilisation—maximising our accelerated computing investment. We’re excited to leverage the platform’s breakthrough capabilities to cost-effectively bring open-source reasoning models to our users.”

Four key innovations of NVIDIA Dynamo

NVIDIA has highlighted four key innovations within Dynamo that contribute to reducing inference serving costs and enhancing the overall user experience:

• GPU Planner: A sophisticated planning engine that dynamically adds and removes GPUs based on fluctuating user demand. This ensures optimal resource allocation, preventing both over-provisioning and under-provisioning of GPU capacity.
• Smart Router: An intelligent, LLM-aware router that directs inference requests across large fleets of GPUs. Its primary function is to minimise costly GPU recomputations of repeat or overlapping requests, thereby freeing up valuable GPU resources to handle new incoming requests more efficiently.
• Low-Latency Communication Library: An inference-optimised library designed to support state-of-the-art GPU-to-GPU communication. It abstracts the complexities of data exchange across heterogeneous devices, significantly accelerating data transfer speeds.
• Memory Manager: An intelligent engine that manages the offloading and reloading of inference data to and from lower-cost memory and storage devices. This process is designed to be seamless, ensuring no negative impact on the user experience.

NVIDIA Dynamo will be made available within NIM microservices and will be supported in a future release of the company’s AI Enterprise software platform. 

See also: LG EXAONE Deep is a maths, science, and coding buff

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 Dynamo: Scaling AI inference with open-source efficiency appeared first on AI News.

Cambricon’s breakthrough into profitability signals a significant shift in the US-China AI chip race, transforming from a 2016 startup into China’s most valuable artificial intelligence company, now valued at approximately 300 billion yuan ($41 billion).

While this represents only a fraction of Nvidia’s $3 trillion market capitalisation, it marks China’s growing capability to develop sophisticated AI chips domestically.

The company’s financial turnaround is particularly noteworthy in the context of technological competition between the world’s two largest economies. After years of losses, Cambricon reported its first quarterly profit in the final quarter of 2024, with net profits ranging from 240 million yuan to 328 million yuan, despite posting a 724 million yuan loss in the first nine months.

The market’s response to this shifting dynamic in the US-China AI chip race has been remarkable. Cambricon’s shares on the Shanghai Stock Exchange’s Star Market have surged more than 470% over the past year, climbing from 120.80 yuan to 695.96 yuan.

The company projects a 70% revenue increase to 1.2 billion yuan in 2024, driven by China’s aggressive buildup of computing infrastructure to support its AI ambitions.

At the technical level, Cambricon has positioned itself as China’s answer to US chip restrictions with its 7-nanometre AI chips. The company’s flagship Cambricon-1A processor has gained significant traction in the domestic market, particularly in products from major technology companies like Huawei Technologies.

The stakes in the US-China AI chip race continue to rise, with analysts at Changjiang Securities projecting that China’s AI semiconductor market will reach 178 billion yuan by 2025. Beijing’s push for semiconductor self-sufficiency and increasing investments from domestic technology companies in AI infrastructure are fueling this growth.

Recent US regulations announced in January 2025 have intensified the race, restricting Chinese access to advanced AI technology and limiting it to American companies and their allies. In response, major Chinese technology companies are investing heavily in domestic computing infrastructure.

ByteDance, TikTok’s parent company, has committed 4.5 billion yuan to a new computing centre in Datong City, Shanxi province. This highlights the growing market opportunity for domestic chip manufacturers.

While Cambricon’s progress represents a significant advancement in the US-China AI chip race, challenges remain. The company must continue to narrow the technological gap with international competitors while maintaining its growth trajectory.

However, supportive government policies and growing domestic demand provide a favourable environment for continued development. Cambricon’s inclusion in the SSE 50 Index, which tracks the Shanghai Stock Exchange’s most valuable companies, underscores its strategic importance to China’s technology sector.

As global tensions persist and access to foreign technology becomes more restricted, developing domestic AI chip capabilities has become increasingly important for China’s technological advancement and economic security.

The post US-China AI chip race: Cambricon’s first profit lands appeared first on AI News.

“Artificial intelligence is quickly becoming central to security and economic strength,” the White House fact sheet declares, framing the controls as a decisive action “to ensure that US technology undergirds global AI use and that adversaries cannot easily abuse advanced AI.”

The new AI chips export controls split the global technology landscape into three distinct tiers, fundamentally reshaping how nations can access and develop AI capabilities. Access to advanced AI processors remains unrestricted for 18 key allies, so-called tier-one nations, including Japan, Britain, and the Netherlands.

However, the administration has implemented strict AI chips export quotas for other nations, creating a new global AI development hierarchy. The 18 allies possess “robust technology protection regimes and technology ecosystems aligned with the national security and foreign policy interests of the US,” the policy document states.

For other countries, the restrictions impose precise limitations – chip orders maxing out at roughly 1,700 advanced GPUs can proceed without licences, primarily benefiting academic and research institutions.

Impact on global AI development

The reverberations through the AI industry were immediate. Nvidia, whose AI accelerators power many of the world’s most advanced artificial intelligence systems, saw its shares decline 2%. Vice President of Government Affairs Ned Finkle warned that the export curb “threatens to derail innovation and economic growth worldwide.”

The stakes are exceptionally high for Nvidia, which derives 56% of its revenue from international markets. Cloud computing giants face a complex recalibration of their AI infrastructure.

Under the new framework, US-headquartered providers must adopt a precise mathematical approach to their global operations: no more than 50% of their AI computing power can be deployed outside the country, with a maximum of 25% beyond tier-one countries, and just 7% in any single non-tier-one nation.

US-China AI technology battle intensifies

The timing and scope of these AI chip export controls reveal their primary target: China’s rapidly advancing AI capabilities. The White House document explicitly warns about “countries of concern” that “actively employ AI — including US-made AI” in ways that could “undermine US AI leadership.” 

With China accounting for 17% of Nvidia’s sales, the commercial impact aligns directly with the administration’s strategic goals. China’s Commerce Ministry’s swift response – promising to “take necessary measures to safeguard its legitimate rights and interests” – signals a new chapter in the technological cold war between the world’s leading AI powers.

The restrictions specifically target China’s ability to develop advanced AI systems, particularly those that could enable “the development of weapons of mass destruction, supporting powerful offensive cyber operations, and aiding human rights abuses.”

Global response and future implications

The US’s European allies have raised concerns about the broad reach of the controls. EU Executive Vice-President Henna Virkkunen and Commissioner Maroš Šefčovič emphasized the need for continued access to advanced AI technology, stating they are “looking forward to engaging constructively with the next US administration” to maintain “a secure transatlantic supply chain on AI technology and supercomputers.”

US National Security Adviser Jake Sullivan frames the controls within a broader technological revolution: “The US has to be prepared for rapid increases in AI’s capability in the coming years, which could have a transformative impact on the economy and our national security.”

Set to take effect in 120 days, the AI chip export controls represent more than just Biden’s final policy move – they establish a new paradigm for global AI development. As former Trump administration national security official Meghan Harris notes, “How effective the rule ends up being in the next 10 to 15 years is now up to the incoming team.”

The regulations mark a defining moment in both US-China relations and global AI development, creating boundaries and alliances that will shape the future of artificial intelligence well beyond the current administration. With these controls, Biden’s final act may be remembered as the moment that redefined the global AI technology landscape.

See also: South Korea wants to develop 50 types of AI chips by 2030

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 US-China tech war escalates with new AI chips export controls appeared first on AI News.

“AI has been advancing at an incredible pace,” Huang said. “It started with perception AI — understanding images, words, and sounds. Then generative AI — creating text, images, and sound. Now, we’re entering the era of ‘physical AI,’ AI that can perceive, reason, plan, and act.”

With NVIDIA’s platforms and GPUs at the core, Huang explained how the company continues to fuel breakthroughs across multiple industries while unveiling innovations such as the Cosmos platform, next-gen GeForce RTX 50 Series GPUs, and compact AI supercomputer Project DIGITS. 

RTX 50 series: “The GPU is a beast”

One of the most significant announcements during CES 2025 was the introduction of the GeForce RTX 50 Series, powered by NVIDIA Blackwell architecture. Huang debuted the flagship RTX 5090 GPU, boasting 92 billion transistors and achieving an impressive 3,352 trillion AI operations per second (TOPS).

“GeForce enabled AI to reach the masses, and now AI is coming home to GeForce,” said Huang.

Holding the blacked-out GPU, Huang called it “a beast,” highlighting its advanced features, including dual cooling fans and its ability to leverage AI for revolutionary real-time graphics.

Set for a staggered release in early 2025, the RTX 50 Series includes the flagship RTX 5090 and RTX 5080 (available 30 January), followed by the RTX 5070 Ti and RTX 5070 (February). Laptop GPUs join the lineup in March.

In addition, NVIDIA introduced DLSS 4 – featuring ‘Multi-Frame Generation’ technology – which boosts gaming performance up to eightfold by generating three additional frames for every frame rendered.

Other advancements, such as RTX Neural Shaders and RTX Mega Geometry, promise heightened realism in video games, including precise face and hair rendering using generative AI.

Cosmos: Ushering in physical AI

NVIDIA took another step forward with the Cosmos platform at CES 2025, which Huang described as a “game-changer” for robotics, industrial AI, and AVs. Much like the impact of large language models on generative AI, Cosmos represents a new frontier for AI applications in robotics and autonomous systems.

“The ChatGPT moment for general robotics is just around the corner,” Huang declared.

Cosmos integrates generative models, tokenisers, and video processing frameworks to enable robots and vehicles to simulate potential outcomes and predict optimal actions. By ingesting text, image, and video prompts, Cosmos can generate “virtual world states,” tailored for complex robotics and AV use cases involving real-world environments and lighting.

Top robotics and automotive leaders – including XPENG, Hyundai Motor Group, and Uber – are among the first to adopt Cosmos, which is available on GitHub via an open licence.

Pras Velagapudi, CTO at Agility, comments: “Data scarcity and variability are key challenges to successful learning in robot environments. Cosmos’ text-, image- and video-to-world capabilities allow us to generate and augment photorealistic scenarios for a variety of tasks that we can use to train models without needing as much expensive, real-world data capture.”

Empowering developers with AI models

NVIDIA also unveiled new AI foundation models for RTX PCs, which aim to supercharge content creation, productivity, and enterprise applications. These models, presented as NVIDIA NIM (Neural Interaction Model) microservices, are designed to integrate with the RTX 50 Series hardware.

Huang emphasised the accessibility of these tools: “These AI models run in every single cloud because NVIDIA GPUs are now available in every cloud.”

NVIDIA is doubling down on its push to equip developers with advanced tools for building AI-driven solutions. The company introduced AI Blueprints: pre-configured tools for crafting agents tailored to specific enterprise needs, such as content generation, fraud detection, and video management.

“They are completely open source, so you could take it and modify the blueprints,” explains Huang.

Huang also announced the release of Llama Nemotron, designed for developers to build and deploy powerful AI agents.

Ahmad Al-Dahle, VP and Head of GenAI at Meta, said: “Agentic AI is the next frontier of AI development, and delivering on this opportunity requires full-stack optimisation across a system of LLMs to deliver efficient, accurate AI agents.

“Through our collaboration with NVIDIA and our shared commitment to open models, the NVIDIA Llama Nemotron family built on Llama can help enterprises quickly create their own custom AI agents.”

Philipp Herzig, Chief AI Officer at SAP, added: “AI agents that collaborate to solve complex tasks across multiple lines of the business will unlock a whole new level of enterprise productivity beyond today’s generative AI scenarios.

“Through SAP’s Joule, hundreds of millions of enterprise users will interact with these agents to accomplish their goals faster than ever before. NVIDIA’s new open Llama Nemotron model family will foster the development of multiple specialised AI agents to transform business processes.”

Safer and smarter autonomous vehicles

NVIDIA’s announcements extended to the automotive industry, where its DRIVE Hyperion AV platform is fostering a safer and smarter future for AVs. Built on the new NVIDIA AGX Thor system-on-a-chip (SoC), the platform allows vehicles to achieve next-level functional safety and autonomous capabilities using generative AI models.

“The autonomous vehicle revolution is here,” Huang said. “Building autonomous vehicles, like all robots, requires three computers: NVIDIA DGX to train AI models, Omniverse to test-drive and generate synthetic data, and DRIVE AGX, a supercomputer in the car.”

Huang explained that synthetic data is critical for AV development, as it dramatically enhances real-world datasets. NVIDIA’s AI data factories – powered by Omniverse and Cosmos platforms – generate synthetic driving scenarios, increasing the effectiveness of training data exponentially.

Toyota, the world’s largest automaker, is committed to using NVIDIA DRIVE AGX Orin and the safety-certified NVIDIA DriveOS to develop its next-generation vehicles. Heavyweights such as JLR, Mercedes-Benz, and Volvo Cars have also adopted DRIVE Hyperion.

Project DIGITS: Compact AI supercomputer

Huang concluded his NVIDIA keynote at CES 2025 with a final “one more thing” announcement: Project DIGITS, NVIDIA’s smallest yet most powerful AI supercomputer, powered by the cutting-edge GB10 Grace Blackwell Superchip.

“This is NVIDIA’s latest AI supercomputer,” Huang declared, revealing its compact size, claiming it’s portable enough to “practically fit in a pocket.”

Project DIGITS enables developers and engineers to train and deploy AI models directly from their desks, providing the full power of NVIDIA’s AI stack in a compact form.

Set to launch in May, Project DIGITS represents NVIDIA’s push to make AI supercomputing accessible to individuals as well as organisations.

Vision for tomorrow

Reflecting on NVIDIA’s journey since inventing the programmable GPU in 1999, Huang described the past 12 years of AI-driven change as transformative.

“Every single layer of the technology stack has been fundamentally transformed,” he said.

With advancements spanning gaming, AI-driven agents, robotics, and autonomous vehicles, Huang foresees an exciting future.

“All of the enabling technologies I’ve talked about today will lead to surprising breakthroughs in general robotics and AI over the coming years,” Huang concludes.

(Image Credit: NVIDIA)

See also: Sam Altman, OpenAI: ‘Lucky and humbling’ to work towards superintelligence

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 advances AI frontiers with CES 2025 announcements appeared first on AI News.

Reportedly led by a dozen AI researchers, scientists, and investors, the new training techniques, which underpin OpenAI’s recent ‘o1’ model (formerly Q* and Strawberry), have the potential to transform the landscape of AI development. The reported advances may influence the types or quantities of resources AI companies need continuously, including specialised hardware and energy to aid the development of AI models.

The o1 model is designed to approach problems in a way that mimics human reasoning and thinking, breaking down numerous tasks into steps. The model also utilises specialised data and feedback provided by experts in the AI industry to enhance its performance.

Since ChatGPT was unveiled by OpenAI in 2022, there has been a surge in AI innovation, and many technology companies claim existing AI models require expansion, be it through greater quantities of data or improved computing resources. Only then can AI models consistently improve.

Now, AI experts have reported limitations in scaling up AI models. The 2010s were a revolutionary period for scaling, but Ilya Sutskever, co-founder of AI labs Safe Superintelligence (SSI) and OpenAI, says that the training of AI models, particularly in the understanding language structures and patterns, has levelled off.

“The 2010s were the age of scaling, now we’re back in the age of wonder and discovery once again. Scaling the right thing matters more now,” they said.

In recent times, AI lab researchers have experienced delays in and challenges to developing and releasing large language models (LLM) that are more powerful than OpenAI’s GPT-4 model.

First, there is the cost of training large models, often running into tens of millions of dollars. And, due to complications that arise, like hardware failing due to system complexity, a final analysis of how these models run can take months.

In addition to these challenges, training runs require substantial amounts of energy, often resulting in power shortages that can disrupt processes and impact the wider electriciy grid. Another issue is the colossal amount of data large language models use, so much so that AI models have reportedly used up all accessible data worldwide.

Researchers are exploring a technique known as ‘test-time compute’ to improve current AI models when being trained or during inference phases. The method can involve the generation of multiple answers in real-time to decide on a range of best solutions. Therefore, the model can allocate greater processing resources to difficult tasks that require human-like decision-making and reasoning. The aim – to make the model more accurate and capable.

Noam Brown, a researcher at OpenAI who helped develop the o1 model, shared an example of how a new approach can achieve surprising results. At the TED AI conference in San Francisco last month, Brown explained that “having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer.”

Rather than simply increasing the model size and training time, this can change how AI models process information and lead to more powerful, efficient systems.

It is reported that other AI labs have been developing versions of the o1 technique. The include xAI, Google DeepMind, and Anthropic. Competition in the AI world is nothing new, but we could see a significant impact on the AI hardware market as a result of new techniques. Companies like Nvidia, which currently dominates the supply of AI chips due to the high demand for their products, may be particularly affected by updated AI training techniques.

Nvidia became the world’s most valuable company in October, and its rise in fortunes can be largely attributed to its chips’ use in AI arrays. New techniques may impact Nvidia’s market position, forcing the company to adapt its products to meet the evolving AI hardware demand. Potentially, this could open more avenues for new competitors in the inference market.

A new age of AI development may be on the horizon, driven by evolving hardware demands and more efficient training methods such as those deployed in the o1 model. The future of both AI models and the companies behind them could be reshaped, unlocking unprecedented possibilities and greater competition.

See also: Anthropic urges AI regulation to avoid catastrophes

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, a

The post New AI training techniques aim to overcome current challenges appeared first on AI News.

Japanese technology experts are developing advanced models that grasp the unique nuances of the Japanese language and culture—essential for industries such as healthcare, finance, and manufacturing – where precision is key.

But this effort isn’t Japan’s alone. Consulting giants like Accenture, Deloitte, EY Japan, FPT, Kyndryl, and TCS Japan are partnering with NVIDIA to create AI innovation hubs across the country. The centres are using NVIDIA’s AI software and specialised Japanese language models to build tailored AI solutions, helping industries boost productivity in a digital workforce. The goal? To get Japanese companies fully on board with enterprise and physical AI.

One standout technology supporting the drive is NVIDIA’s Omniverse platform. With Omniverse, Japanese companies can create digital twins—virtual replicas of real-world assets—and test complex AI systems safely before implementing them. This is a game-changer for industries such as manufacturing and robotics, allowing businesses to fine-tune processes without the risk of real-world trial and error. This use of AI is more than just innovation; it represents Japan’s plan for addressing some major challenges ahead.

Japan faces a shrinking workforce presence as its population ages. With its strengths in robotics and automation, Japan is well-positioned to use AI solutions to bridge the gap. In fact, Japan’s government recently shared its vision of becoming “the world’s most AI-friendly country,” underscoring the perceived role AI will play in the nation’s future.

Supporting this commitment, Japan’s AI market hit $5.9 billion in value this year; a 31.2% growth rate according to IDC. New AI-focused consulting centres in Tokyo and Kansai give Japanese businesses hands-on access to NVIDIA’s latest technologies, equipping them to solve social challenges and aid economic growth.

Top cloud providers like SoftBank, GMO Internet Group, KDDI, Highreso, Rutilea, and SAKURA Internet are also involved, working with NVIDIA to build AI infrastructure. Backed by Japan’s Ministry of Economy, Trade and Industry, they’re establishing AI data centres across Japan to accelerate growth in robotics, automotive, healthcare, and telecoms.

NVIDIA and SoftBank have also formed a remarkable partnership to build Japan’s most powerful AI supercomputer using NVIDIA’s Blackwell platform. Additionally, SoftBank has tested the world’s first AI and 5G hybrid telecoms network with NVIDIA’s AI Aerial platform, allowing Japan to set a worldwide standard. With these developments, Japan is taking big strides toward establishing itself as a leader in the AI-powered industrial revolution.

(Photo by Andrey Matveev)

See also: NVIDIA’s share price nosedives as antitrust clouds gather

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 AI Summit Japan: NVIDIA’s role in Japan’s big AI ambitions appeared first on AI News.

This breakthrough represents a significant step in China’s ongoing efforts to become self-reliant in AI technology, especially in light of escalating US limitations on access to advanced semiconductors for its competitors.

According to the company’s Institute of AI, one of the models, TeleChat2-115B and another unnamed model were trained on tens of thousands of Chinese-made chips. This achievement is especially noteworthy given the tighter US export rules that have limited China’s ability to purchase high-end processors from Nvidia and other foreign companies. In a statement shared on WeChat, the AI institute claimed that this accomplishment demonstrated China’s capability to independently train LLMs and signals a new era of innovation and self-reliance in AI technology.

The scale of these models is remarkable. China Telecom stated that the unnamed LLM has one trillion parameters. In AI terminology, parameters are the variables that help the model in learning during training. The more parameters there are, the more complicated and powerful the AI becomes.

Chinese companies are striving to keep pace with global leaders in AI based outside the country. Washington’s export restrictions on Nvidia’s latest AI chips such as the A100 and H100, have compelled China to seek alternatives. As a result, Chinese companies have developed their own processors to reduce reliance on Western technologies. For instance, the TeleChat2-115B model has approximately 100 billion parameters, and therefore can perform as well as mainstream platforms.

China Telecom did not specify which company supplied the domestically-designed chips used to train its models. However, as previously discussed on these pages, Huawei’s Ascend chips play a key part in the country’s AI plans.

Huawei, which has faced US penalties in recent years, is also increasing its efforts in the artificial intelligence field. The company has recently started testing its latest AI processor, the Ascend 910C, with potential clients waiting in the domestic market. Large Chinese server companies, as well as internet giants that have previously used Nvidia chips, are apparently testing the new chip’s performance. Huawei’s Ascend processors, as one of the few viable alternatives to Nvidia hardware, are viewed as a key component of China’s strategy that will lessen its reliance on foreign technology.

In addition to Huawei, China Telecom is collaborating with other domestic chipmakers such as Cambricon, a Chinese start-up specialising in AI processors. The partnerships reflect a broader tendency in China’s tech industry to build a homegrown ecosystem of AI solutions, further shielding the country from the effects of US export controls.

By developing its own AI chips and technology, China is gradually reducing its dependence on foreign-made hardware, especially Nvidia’s highly sought-after and therefore expensive GPUs. While US sanctions make it difficult for Chinese companies to obtain the latest Nvidia hardware, a black market for foreign chips has emerged. Rather than risk operating in the grey market, many Chinese companies prefer to purchase lower-powered alternatives such as previous-gen models to maintain access to Nvidia’s official support and services.

China’s achievement reflects a broader shift in its approach to AI and semiconductor technology, emphasising self-sufficiency and resilience in an increasingly competitive global economy and in face of American protectionist trade policies.

(Photo by Mark Kuiper)

See also: Has Huawei outsmarted Apple in the AI race?

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 China Telecom trains AI model with 1 trillion parameters on domestic chips appeared first on AI News.

This past summer, investors questioned whether top AI stocks could sustain their sky-high valuations, given the lack of returns on massive AI spending. As Autumn approaches, major AI sectors—such as chips, LLMs, and AI devices—received renewed confidence. Nonetheless, there are an increasing number of reasons to be cautious.

Cerebras: A chip contender with a major risk

Chip startup Cerebras is challenging Nvidia’s dominance by developing processors designed to power smarter LLMs. Nvidia, a major player in the AI boom, has seen its market cap skyrocket from $364 billion at the start of 2023 to over $3 trillion.

Cerebras, however, relies heavily on a single customer: the Abu Dhabi-based AI firm G42. In 2023, G42 accounted for 83% of Cerebras’ revenue, and in the first half of 2024, that figure increased to 87%. While G42 is backed by major players like Microsoft and Silver Lake, its dependency poses a risk. Even though Cerebras has signed a deal with Saudi Aramco, its reliance on one client may cause concerns as it seeks a $7-8 billion valuation for its IPO.

OpenAI’s record-breaking funding – but with strings attached

OpenAI made the news when it raised $6.6 billion at a $157 billion valuation, becoming the largest investment round in Silicon Valley history. However, the company has urged its investors not to back competitors such as Anthropic and Elon Musk’s xAI—an unusual request in the world of venture capital, where spread betting is common. Critics, including Gary Marcus, have described this approach as “running scared.”

OpenAI’s backers also include “bubble chasers” such as SoftBank and Tiger Global, firms known for investing in companies at their peak, which frequently results in huge losses. With top executives such as CTO Mira Murati departing and predicted losses of $5 billion this year despite rising revenues, OpenAI faces significant challenges.

Meta’s big bet on AI wearables

Meta entered the AI race by unveiling Orion, its augmented reality glasses. The wearables promise to integrate AI into daily life, with Nvidia’s CEO Jensen Huang endorsing the product. However, at a production cost of $10,000 per unit, the price is a major obstacle.

Meta will need to reduce costs and overcome consumer hesitation, as previous attempts at AI-powered wearables—such as Snapchat’s glasses, Google Glass, and the Humane AI pin—have struggled to gain traction.

The road ahead

What’s next for AI? OpenAI must prove it can justify a $157 billion valuation while operating at a loss. Cerebras needs to reassure investors that relying on one client isn’t a dealbreaker. And Meta must convince consumers to adopt a completely new way of interacting with AI.

If these companies succeed, this moment could mark a turning point in the AI revolution. However, as tech history shows, high-stakes markets are rarely easy to win.

(Photo by Growtika)

See also: Ethical, trust and skill barriers hold back generative AI progress in EMEA

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 The risks behind the generative AI craze: Why caution is growing appeared first on AI News.

During the regular trading session on Tuesday, NVIDIA’s share price experienced a near-10% drop. The fall wiped £212 billion from its market value, marking the largest single-day loss for a US company in history.

While the wider market experienced a sell-off fueled by concerns over weak US manufacturing data, NVIDIA was hit particularly hard after Bloomberg reported that the US Department of Justice issued subpoenas to NVIDIA and other tech firms. 

Officials are reportedly concerned that NVIDIA’s business practices may be hindering client flexibility in switching to alternative semiconductor suppliers. Additionally, there are concerns about potential penalties imposed on buyers who opt not to exclusively utilise NVIDIA’s AI chips. Such actions would represent an escalation of the ongoing US antitrust investigation, bringing the government a step closer to formally charging NVIDIA.

In response, NVIDIA asserted its belief that its success is based “on merit, as reflected in our benchmark results and value to customers, who can choose whatever solution is best for them.” 

This latest downturn adds to the recent volatility experienced by NVIDIA and other AI-related stocks, such as Google, Apple, and Amazon. Investors are grappling with uncertainty surrounding the timeline for tangible benefits and concrete returns from the much-touted AI revolution.

Analysts suggest that investors are seeking greater clarity on the trajectory of gross margins as production of NVIDIA’s new Blackwell chip increases. Furthermore, they are eager for more concrete evidence that AI is delivering tangible returns for customers.

After a 9.5% decline on Tuesday alone and a 14% drop since last week’s earnings report, NVIDIA’s stock has shown marginal signs of recovery in today’s trading session, registering a modest 0.64% increase at the time of writing.

Looking ahead, NVIDIA will need to convince investors of its growth potential not only for 2025 but also for 2026. While Wall Street currently focuses on Blackwell chip shipments, there is increasing interest in the company’s next-generation chip offering.

(Photo by Sebastian Molina)

See also: xAI breaks records with ‘Colossus’ AI training system

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’s share price nosedives as antitrust clouds gather appeared first on AI News.