Why Microsoft and AWS Are Challenging NVIDIA – The Truth Behind the Battle for AI Chip Priority

An intriguing situation is unfolding in the AI industry. News has emerged that Microsoft (MS) and Amazon Web Services (AWS) are supporting the ‘Gain AI Act’, which restricts NVIDIA’s exports to China. According to a report by The Wall Street Journal, this is seen not as a mere political stance but as a survival strategy in the AI market. It’s quite rare for NVIDIA’s two largest customers to support legislation that curbs NVIDIA itself.

To understand this situation, one must look at the current structure of the AI chip market. As of 2025, NVIDIA holds over 80% of the AI chip market share and has an almost monopolistic position, especially in the data center GPU market. The H100 and the latest H200 chips are priced between $25,000 and $40,000 each, yet supply still cannot meet demand. For hyperscalers like MS and AWS to expand their AI services, these high-performance chips are essential. However, as NVIDIA supplies chips globally, including to China, there is a persistent shortage in the U.S. market.

The core of the Gain AI Act is to ensure that chip manufacturers prioritize meeting domestic demand in the U.S. before exporting to China and other countries under arms embargoes. It includes provisions allowing hyperscalers like MS, AWS, Google, and Meta to secure priority access to semiconductor chips for their data centers. MS’s policy chief, Jerry Petrella, has publicly expressed strong support for this bill, and AWS has reportedly indicated its approval to Senate aides unofficially. Interestingly, Anthropic, a competitor to OpenAI, also supports this legislation.

Personally, I think this is a product of very complex interests. MS has invested over $13 billion in OpenAI and provides AI services via Azure Cloud. AWS, while developing its own AI chips, Trainium and Inferentia, still relies on NVIDIA chips. As of the fourth quarter of 2024, AWS’s AI service revenue increased by 185% year-over-year, and a stable chip supply is essential to sustain this growth.

NVIDIA’s Rebuttal and Market Reality

NVIDIA’s stance also has merit. NVIDIA argues that this bill unnecessarily interferes with the semiconductor market and could lead to more export restrictions in the future. They emphasize that there are already enough chips in the U.S., and the main bottleneck in the AI industry is power. Indeed, as of 2025, many data centers are struggling to expand due to power shortages. For instance, Google’s power consumption increased by 48% year-over-year in 2024 due to AI workloads, and MS is in a similar situation.

However, contrary to NVIDIA’s claims, the market still faces a chip shortage. For the latest H200 chips, there is a waiting period of over six months after ordering, and the situation is exacerbated by Chinese companies circumventing U.S. export restrictions through indirect purchases. Chinese tech giants like Baidu, Alibaba, and Tencent are purchasing GPUs in large quantities for AI development, impacting the supply available to U.S. companies.

NVIDIA’s fourth-quarter 2024 results show data center revenue at $47.6 billion, a 409% increase year-over-year, with a significant portion coming from Asia, including China. While they supply downgraded versions like the H800 and H20 due to export restrictions to China, the Chinese market remains a crucial revenue source for NVIDIA. It’s estimated that about 20-25% of NVIDIA’s total revenue in 2024 is related to China.

What’s particularly interesting is the silence from Google and Meta. These two companies have yet to express their stance on the Gain AI Act, but their strategies might explain why. Google has invested significantly in developing its own AI chip, the TPU (Tensor Processing Unit), which outperforms the H100 in specific AI workloads. Meta is also developing its own AI chips, aiming to reduce dependency on NVIDIA in the long term.

A New Phase in Global AI Competition

This situation indicates that global AI competition is entering a new phase. It’s expanding from mere technological development competition to a race to secure supply chains. In response to U.S. semiconductor export restrictions, China is accelerating its own AI chip development. Huawei’s Ascend chip and Baidu’s Kunlun chip are examples, and although they currently lag behind NVIDIA’s chips in performance, they are rapidly closing the gap.

The situation in South Korea also warrants attention. Samsung Electronics and SK Hynix dominate over 90% of the HBM (High Bandwidth Memory) market, a key component of NVIDIA GPUs. Notably, SK Hynix’s HBM3E is used as a core component in NVIDIA’s H200 chips, meaning U.S. semiconductor regulations on China directly affect Korean companies. SK Hynix’s HBM revenue in 2024 increased by over 300% year-over-year, with a significant portion coming from supplies to NVIDIA.

The stance of the Trump administration is also noteworthy. Administration officials, including David Sachs, appointed as the ‘AI Czar’, have reportedly conveyed to the Department of Commerce that it already has the authority to oversee chip exports, suggesting the impact of the Gain AI Act might be limited. However, President Trump himself has yet to express his position on this bill, and his decision could significantly influence the future landscape of the AI industry.

Personally, I see this situation as an important signal of a shift in the power structure within the AI industry. While NVIDIA has so far dominated the industry through AI chip supply, hyperscalers like MS and AWS are starting to assert themselves more vocally. They are not merely chip-buying customers but key players in the AI ecosystem actively pursuing their interests.

In fact, AWS doubled its investment in its own AI chips in 2024, and MS is exploring alternatives to NVIDIA through a partnership with AMD. These moves seem to be creating cracks in NVIDIA’s monopolistic position. Although NVIDIA’s technical superiority is likely to continue in the short term, competition is expected to intensify in the long run.

Ultimately, the support for the Gain AI Act appears to be a strategic choice by MS and AWS to survive in the AI competition. They are demonstrating a willingness to secure a stable chip supply, even at the risk of straining their relationship with NVIDIA. It remains to be seen whether this bill will actually pass and what changes it will bring to the AI industry if it does. One thing is certain: AI competition is spreading beyond technology to encompass politics and the broader economy.

This article was written by reading a news article and adding personal opinions and analysis.

Disclaimer: This blog is not a news outlet, and the content reflects the author’s personal views. The responsibility for investment decisions lies with the investor, and the author bears no responsibility for investment losses based on this content.