Yazılar

H3C Warns of Nvidia AI Chip Shortage as Chinese Demand Surges

/in /tarafından

One of China’s top server manufacturers, H3C, has reportedly warned of a looming shortage of Nvidia’s H20 AI chip—the most advanced processor currently available for sale in China under U.S. export rules. The notice, seen by Reuters, indicates rising concerns about disruptions in the international supply chain and signals possible obstacles for China’s AI development ambitions.

According to the document dated Tuesday, H3C informed clients that existing inventories of the H20 chip are “nearly depleted,” with new shipments only expected around mid-April. The notice blamed geopolitical tensions and raw material disruptions for the uncertainty. It also highlighted challenges in future supply planning beyond April 20 due to changes in shipping routes, production complications, and evolving policy environments.

Shortly after publication, Nvidia declined to comment, and H3C issued a clarification stating that “neither the company nor any of its departments have issued this notice or its related content.” However, industry sources confirmed that the chip is becoming increasingly difficult to obtain.

. AI Boom Driving Surging Demand
Demand for the H20 has sharply risen since January, driven by the popularity of Chinese startup DeepSeek’s cost-efficient AI models. Major Chinese tech giants such as Tencent, Alibaba, and ByteDance have ramped up orders in response, according to previous reporting by Reuters.

An anonymous distributor told Reuters that despite prior assurances about availability, many buyers were ultimately told the chips had already been sold at higher prices. “We were told the chips would be available, but when it came time to actually purchase them, we were informed they had already been sold at higher prices,” the source said.

. Geopolitical Pressures and U.S. Export Controls
The H20 chip was specifically designed by Nvidia to comply with tightened U.S. export restrictions implemented in October 2023. These rules ban the sale of the company’s most powerful chips—such as the A100 and H100—to China over national security concerns. The U.S. believes advanced AI chips could potentially be used to bolster China’s military capabilities.

Despite these restrictions, Nvidia is believed to have shipped around 1 million units of the H20 to China in 2024, generating over $12 billion in revenue.

Washington is also reportedly considering further limitations on even these lower-tier chips, raising the risk of additional constraints on China’s access to cutting-edge AI technologies.

. China’s Response and Domestic Alternatives
H3C is one of Nvidia’s key OEM partners in China, alongside Inspur, Lenovo, and xFusion—a spinoff from Huawei focused on x86 servers. The looming shortage could accelerate China’s efforts to develop and adopt domestic alternatives such as Huawei’s Ascend chips and Cambricon’s AI processors.

H3C’s internal communication reportedly emphasized a “profit-first” distribution model for upcoming chip deliveries, favoring long-term, high-margin customers. This signals that the chip shortage could hit smaller AI startups and new entrants the hardest, potentially reshaping China’s fast-evolving AI ecosystem.

US Commerce Department Bureaus Reportedly Ban China’s DeepSeek on Government Devices

/in /tarafından

In recent weeks, bureaus within the U.S. Commerce Department have informed staff that the Chinese-developed Artificial Intelligence (AI) model, DeepSeek, is banned from being used on government devices. This move, confirmed by a message seen by Reuters and sources familiar with the matter, is part of ongoing efforts to safeguard government information systems. A mass email sent to staffers explained the directive, urging them to avoid downloading or accessing any applications, desktop apps, or websites associated with DeepSeek. The email specifically warned, “To help keep Department of Commerce information systems safe, access to the new Chinese-based AI DeepSeek is broadly prohibited on all GFE” (government-furnished equipment).

The Commerce Department’s decision comes amid growing concerns over cybersecurity risks tied to foreign technologies. DeepSeek, an AI model developed in China, has raised alarms due to its rapid proliferation and the potential security vulnerabilities it might pose. While the Commerce Department has not provided further comment on the matter, the ban highlights the increasing scrutiny of Chinese technology in the U.S. government sector.

The specific scope of the ban across other U.S. government departments remains unclear, with Reuters unable to determine the full extent of the restrictions. However, the move reflects broader U.S. efforts to limit the use of Chinese technologies in critical sectors, particularly those involving sensitive data and national security. This step follows a series of actions aimed at curbing the influence of Chinese tech firms on American infrastructure.

The controversy surrounding DeepSeek has also reverberated in financial markets. In January, the introduction of DeepSeek’s low-cost AI models contributed to a major selloff in global equity markets, as investors expressed concerns that China’s advancements in AI could challenge the United States’ dominance in the field. This growing fear of losing competitive ground in AI technology has only amplified regulatory scrutiny in the U.S., particularly as AI continues to play an increasingly pivotal role in both the economy and national security.

Microsoft Scales Back on Data Center Leases Amid AI Spending Concerns

/in /tarafından

Microsoft has pulled back from leasing new data center capacity in the U.S. and Europe, abandoning projects that would have used 2 gigawatts of electricity over the past six months. According to analysts at TD Cowen, the tech giant’s decision is driven by an oversupply of data center capacity relative to its current demand forecast, particularly in light of its shifting approach to supporting OpenAI’s ChatGPT workloads.

Shifting Focus and Market Impact

Investor skepticism has risen regarding the large-scale artificial intelligence (AI) investments made by U.S. tech giants, partly due to slower-than-expected returns and competition from Chinese startup DeepSeek, which offers AI solutions at significantly lower costs. As part of its pullback, Microsoft has decided not to support additional AI workloads, particularly those associated with OpenAI’s ChatGPT, a move that has been closely watched by industry analysts.

Microsoft’s withdrawal from certain data center projects has led to competitors stepping in to fill the void. Alphabet’s Google and Meta Platforms have moved to backfill the data center capacity, with Google focusing on international markets and Meta stepping in for U.S. projects. Despite these shifts, Microsoft remains committed to growing its infrastructure, with plans to invest $80 billion in AI infrastructure during this fiscal year, in line with its ongoing AI strategy.

Continuing Investment and Future Outlook

While Microsoft’s share price saw a slight decline of over 1% on Wednesday, the company reassured investors that its infrastructure growth plans will remain strong across all regions. The company has already scrapped leases with at least two private data center operators, a decision that aligns with its strategic pacing and adjustments to its AI needs.

Executives from both Microsoft and Meta defended their massive AI investments after the reveal of DeepSeek’s cost-effective technology in January, emphasizing that these investments are crucial to remaining competitive in the rapidly evolving AI space. Alphabet has also committed to increasing its AI spending this year, planning $75 billion, a 29% increase over Wall Street’s expectations.

Conclusion

Microsoft’s decision to scale back on data center leases highlights the evolving landscape of AI infrastructure spending, as companies adjust their strategies in response to market competition and changing demand. Despite this pullback, Microsoft’s commitment to AI remains strong, with a continued focus on investing heavily in the technology’s future.