AI Chips Archives - Gizmochina

AI Chip War Heats Up: U.S. Blocks Nvidia, China Responds with Total Ban

Joane — Fri, 07 Nov 2025 11:02:35 +0000

Key Highlights:

U.S. government has officially blocked Nvidia from selling its advanced Blackwell and B30A AI chips to China.
China has responded with a new rule banning foreign AI chips from all state-funded data centers.
Global AI chip war deepens as both nations tighten technology controls

White House Blocks Nvidia’s Advanced Blackwell Chip Sales to China

The Trump administration has decided not to allow Nvidia to sell its latest Blackwell AI chip to China. White House spokeswoman Karoline Leavitt confirmed the decision, stating, “The Blackwell chip is not something we’re interested in selling to China right now.”

This decision follows months of speculation about whether a scaled-down version of the chip might be cleared for sale. The move aligns with President Trump’s strategy to maintain America’s technological edge in artificial intelligence and prevent China from acquiring high-end U.S. hardware.

US Treasury Secretary Calls Blackwell the “Crown Jewel”

U.S. Treasury Secretary Scott Bessent described Nvidia’s Blackwell chip as the “crown jewel” of American AI innovation. He indicated that sales to China might be reconsidered only after the chip becomes outdated in 12 to 24 months. According to Bessent, Nvidia’s rapid pace of innovation means that the Blackwell could soon be several generations behind the company’s latest models.

Nvidia’s Market Collapse in China

Nvidia’s position in China has deteriorated sharply. Once controlling 95% of the Chinese AI data center market in 2022, the company now holds no market share. CEO Jensen Huang has expressed hope that Nvidia will eventually return to the Chinese market, but admitted there are no current plans to do so. Nvidia is reportedly redesigning its B30A chip in hopes of meeting future U.S. export conditions.

China Retaliates with Ban on Foreign AI Chips

In a direct response, Beijing has ordered that all new data centre projects receiving state funding must use only domestically produced AI chips. Projects that are less than 30% complete are required to remove or cancel any foreign chip orders, while those further along will be reviewed individually.

This directive represents one of China’s most aggressive steps toward achieving AI chip self-sufficiency. It is expected to benefit domestic manufacturers such as Huawei, Cambricon, and MetaX. However, industry analysts warn that Chinese AI chips still lag behind Nvidia’s products in performance and software ecosystem support, potentially slowing China’s AI development.

The Global AI Divide Widens

The escalating U.S.–China chip restrictions are reshaping the global AI landscape. The United States aims to secure national security and preserve its technological dominance, while China seeks independence from U.S. technology and to strengthen its domestic semiconductor ecosystem.

While American tech giants like Microsoft, Meta, and OpenAI continue to build massive AI data centres using Nvidia’s most advanced chips, China risks falling behind in computing capacity despite gains in local chip production.

Both the U.S. and China are pursuing self-reliance in AI technology, setting the stage for a deeper divide in global innovation. The U.S. is safeguarding its AI lead, while China is accelerating its domestic semiconductor push.

Nvidia and other American chipmakers are losing access to one of the world’s largest markets, while Chinese chip firms stand to benefit, though at the cost of slower advancement in high-end AI capabilities.

Read More:

(via)

The post AI Chip War Heats Up: U.S. Blocks Nvidia, China Responds with Total Ban appeared first on Gizmochina.

NVIDIA’s Market Share in China Drops from 95% to Zero Amid US Export Restrictions

Joane — Sat, 18 Oct 2025 14:15:30 +0000

NVIDIA has lost its entire share of the advanced chip market in China after once holding a 95% dominance. This sharp decline follows strict US export restrictions that prevent the company from selling its most advanced chips to Chinese companies. “At the moment, we’re 100 percent out of China,” confirmed NVIDIA CEO Jensen Huang during a recent event.

Sudden Market Collapse

The US government has banned the export of NVIDIA’s high-end AI chips to China since 2022. This includes the NVIDIA A100, NVIDIA H100, and NVIDIA H200. Although NVIDIA was allowed to sell a downgraded chip, the NVIDIA H20, the Chinese cyberspace regulator launched a security investigation, discouraging local companies from buying it. As a result, the company has been forced out of the Chinese market entirely.

Huang’s Warning

Jensen Huang criticized the export restrictions, calling them a “mistake.” He warned that cutting off China would push the market toward domestic competitors like Huawei Technologies. “What harms China could oftentimes also harm America, and even worse,” he said. Huang emphasized the importance of rethinking policies to maintain global leadership in technology.

China’s Rapid Self-Sufficiency Drive

China is moving fast to reduce its dependence on foreign chips. Domestic companies are investing heavily to develop their own AI chips and manufacturing technologies. Huawei has already introduced an advanced AI chip roadmap and new clustering methods to replace NVIDIA’s products. Other tech giants like Alibaba Group Holding, Tencent Holdings, ByteDance, and Baidu are also pouring resources into semiconductor research and development.

A Powerful AI Talent Base

Huang highlighted that China has nearly 50% of the world’s AI researchers. He warned that excluding the country from US technology could slow down global AI progress. He also noted China’s strong work culture, provincial competition, and deep talent pool that enable it to close the gap with the US quickly. According to Huang, China is only “nanoseconds behind” in chipmaking.

Strategic Impact on the US

Analysts warn that the restrictions may hurt US companies more than expected. By pushing NVIDIA out of the Chinese market, the US may be allowing Chinese firms to strengthen their supply chains and become more independent. This could reduce American companies’ global competitiveness and limit their access to one of the world’s largest chip markets.

Event and Industry Reaction

Huang made these remarks on October 6 at a Citadel Securities event in New York, where he spoke with Konstantine Buhler of Sequoia Capital. A video of the interview was released the following week and widely reported by the South China Morning Post. His comments reflect growing concern in the tech industry about the long-term consequences of the US-China tech war.

The collapse of NVIDIA’s market share in China highlights the impact of geopolitical tensions on the global semiconductor industry. While the US aims to protect its national security, the strategy may also be boosting China’s technological independence. As the chip race accelerates, the world is watching closely to see whether policy shifts will ease tensions or deepen the divide between the world’s two largest economies.

AMD’s Instinct MI300 series AI chips aim to redefine AI computing for Microsoft

Debasish — Wed, 14 Feb 2024 04:43:01 +0000

In a bid to revolutionize the AI computing landscape, AMD is set to introduce its latest innovation, the Instinct MI300 series accelerator. Reports from Windows Report suggest that AMD is poised to supply this cutting-edge technology to Microsoft at a reduced price, signaling a renewed collaboration between the two tech giants. This move comes at a time when Microsoft primarily relies on NVIDIA’s H100 Tensor Core GPU for artificial intelligence based tasks, underlining AMD’s ambitious push into the AI hardware market.

AMD Instinct MI300 series is an affordable alternative to NVIDIA’s H100

The AMD Instinct MI300X is positioned as an affordable alternative to NVIDIA‘s offerings, boasting a slew of groundbreaking features. Foremost among these is its innovative architecture, promising a significant leap in performance. Central to its capabilities is the incorporation of 192GB of high-performance HBM3 memory, facilitating faster data transmission and accommodating larger AI models. This is a critical aspect given the increasing complexity of AI applications, where speed and efficiency are paramount.

A notable aspect of the Instinct MI300X is its Chiplet design, housing 13 small chips based on 3D stacking. Among these are 24 Zen4 CPU cores, alongside integration of CDNA 3 and 8 HBM3 memory stacks. Leveraging 5nm and 6nm IP, this accelerator combines a total of 128GB HBM3 memory and a staggering 146 billion transistors. AMD’s strategic allocation of chiplets, with a majority based on 3D stacking, underscores its commitment to optimizing performance and efficiency.

AMD’s CEO, Su Zifeng, has emphasized the transformative potential of the MI300X, particularly in comparison to NVIDIA’s H100 GPU. Zifeng highlights the tangible benefits in user experience, citing the accelerator’s ability to deliver faster and more nuanced responses in AI-driven interactions. With AI applications becoming increasingly pervasive across industries, the demand for enhanced performance is paramount, a need that AMD aims to address with its latest offering. The Instinct MI300 series AI chips signify a significant milestone in AMD’s quest to establish itself as a key player in the AI hardware arena.

Tencent adapts to U.S. export bans, explores domestic alternatives for AI Chips

Joane — Thu, 16 Nov 2023 16:31:54 +0000

In response to the recent broadened U.S. restrictions on high-end chip sales to China, Tencent Holdings is strategically navigating the challenges posed by the semiconductor export bans to safeguard its cloud services. The company, acknowledging its substantial stockpile of AI chips from U.S. manufacturer Nvidia, is proactively planning to optimize their usage amidst the evolving regulatory landscape.

Tencent’s President, Martin Lau, emphasized the impact of the U.S. ban on exporting additional AI chips to China, revealing that the company is committed to efficiently utilizing its existing Nvidia chip stock for the continued development of its “Hunyuan” AI model. Despite the ban, Tencent assures continuity for at least a couple more generations, minimizing the near-term impact on its AI capabilities.

However, the export restrictions are expected to affect Tencent’s cloud services, particularly in the resale of AI chips to clients. Tencent currently relies on Nvidia, which dominates approximately 90% of the Chinese AI chip market. The company is exploring alternatives, including domestically produced chips, to mitigate the risks posed by the U.S. restrictions.

Tencent’s rival, Baidu, has reportedly taken steps to secure alternatives by ordering 1,600 Huawei Ascend 910B chips. In a bid to address the challenges posed by the ban on its H800 AI chips developed by Nvidia specifically for China, Tencent plans to reserve them for the crucial training phase of AI model development. The company is actively seeking ways to offload inference capability to lower-performance chips, preserving high-performance AI chips for training purposes.

Recognizing the need to reduce reliance on U.S.-made chips, Tencent is looking to increase its usage of domestically-produced chips. The company is not alone in this endeavor, as Chinese firms, including Baidu, are seeking alternatives and turning to domestic chipmakers like Huawei.

In response to the changing landscape, Nvidia is expected to announce new China-bound AI chips designed to comply with the export rules by featuring reduced computing power while retaining essential AI features.

Tencent’s move to explore domestic sources for training chips aligns with China’s broader efforts to invest billions in its semiconductor industry, countering the impact of U.S. high-end chip bans. Initiatives like the $41 billion fund launched in September and support for domestic chipmakers, such as YMTC, demonstrate China’s commitment to achieving self-sufficiency in the semiconductor sector despite external challenges.

Related:

(via)

The post Tencent adapts to U.S. export bans, explores domestic alternatives for AI Chips appeared first on Gizmochina.

Huawei’s new AI chips to compete against Qualcomm and Nvidia

Simranpal Singh — Wed, 10 Oct 2018 05:49:49 +0000

The world’s largest telecom-equipment maker, Huawei has unveiled two new AI chips at its annual global partners’ conference. Huawei has already surpassed Apple Inc. as the world’s no. 2 smartphone vendor. Now, they have another goal of exceeding sales and giving competition to the US chip makers like Qualcomm and Nvidia in the AI market.

Yesterday, Huawei announced a new Ascend series which is capable of machine learning as like of chips from Nvidia and Qualcomm. Along with the new chips, Huawei has also introduced the cloud computing services to help devices better integrate with different services. Its cloud computing platform will give competition to the homegrown Alibaba, Microsoft Corp and Amazon Web Services.

Huawei announced two new Artificial Intelligence chips called Ascend 910 and Ascend 310. Huawei aims Ascend 910 for use in the data centres whereas Ascend 310 is intended for internet-connected devices like smartphones, smartwatches and IoT devices.

“Huawei’s AI strategy is built on top of continued investment in basic technologies and talent training,” Chairman Eric Xu told the Huawei Connect conference in Shanghai.

It’s an overall push from Huawei to excel in the 5G technology both software and hardware wise. It will also help the Chinese company to mark a debut against the world’s top AI and cloud-based companies in the US. Huawei is shifting into AI chip business for enterprises and consumer applications. It’s a big push from Huawei, as until now they have managed to made chips for its smartphones.

Read More: Huawei Exec explains why the company doesn’t supply Kirin processors to other brands

“Going forward we need to think of new ways to prepare our business and industry for change. There are clear signs that AI will change or disrupt a whole host of industries,” Huawei Rotating Chairman Eric Xu said on Wednesday.

Let’s see the how Huawei ascend in the AI chip market.

(Via)

The post Huawei’s new AI chips to compete against Qualcomm and Nvidia appeared first on Gizmochina.