Amazon Develops Own AI Chips to Reduce Nvidia Dependency, Plans Sales to External Customers

Amazon is continuing its strategy to reduce its dependence on Nvidia with its own AI chips. The company's cloud division is launching new versions of its custom-developed processors and attempting to offer hardware that can compete with Nvidia's products. In June 2026, Amazon expanded its ambitions: the company held discussions to sell its custom AI chips to external customers – a step beyond its previous internal use.

Amazon's Chip Development as Counter-Strategy

Amazon has been developing its own AI processors for several years to strengthen its position in the competition for artificial intelligence infrastructure. The chips aim to provide computing power for machine learning – a segment currently dominated by Nvidia's graphics processing units (GPUs). Through in-house development, Amazon can potentially reduce costs and retain technical control over its cloud services.

The decision to sell these chips to third-party companies in future marks a strategic shift. Until now, Amazon had used the processors exclusively for its own cloud platform Amazon Web Services (AWS). External sales would not only generate additional revenue, but also spread development costs across a broader base and expand the ecosystem around Amazon chips.

Nvidia's Market Position Under Pressure

Nvidia currently holds a dominant position in the market for AI hardware. The company's GPUs are regarded as the standard for training large language models and other compute-intensive AI applications. Analysts already viewed Amazon's chip development in March 2025 as a potential threat to this market leadership. Whether Amazon can actually capture significant market share depends on several factors: the technical performance of the chips, their cost-effectiveness, and customers' willingness to switch to an alternative to Nvidia's established platform.

For major customers such as cloud providers or tech companies, developing their own chips or using alternatives can be attractive to regain negotiating power against Nvidia and reduce supply chain risks. Google and Microsoft are also pursuing comparable strategies with their own AI processors.

Infrastructure Bottlenecks Hamper Expansion

Despite chip development and massive investments in Nvidia hardware, Amazon is struggling with fundamental infrastructure problems. According to a Business Insider report from 21 October 2026, Amazon is buying billions of dollars' worth of Nvidia AI hardware, yet some new server racks are sitting unused in warehouses because data centers lack the electrical capacity to operate them. This power shortage limits the company's ability to deploy even existing GPU capacity – a problem that exists regardless of chip manufacturer.

The situation illustrates that competition in the AI infrastructure market is not decided solely by chip availability. Data center capacity, power supply, cooling, and network connectivity are equally critical factors. Although Amazon is investing heavily in expanding its data centers, lead times for new facilities and power connections can take years.

Dual Strategy: In-House Development and Nvidia Purchases

Amazon is currently pursuing a dual strategy: on the one hand, the company is developing and marketing its own AI chips; on the other hand, it continues to purchase Nvidia products in large quantities. This parallel approach reflects the reality that AWS customers often specifically demand Nvidia GPUs because existing software and models are optimized for this platform. A complete switch to in-house chips is therefore neither technically nor commercially realistic in the short term.

For Nvidia, Amazon's strategy nonetheless creates long-term pressure: every workload running on Amazon chips instead of Nvidia GPUs reduces revenue potential. At the same time, the availability of alternative chips could limit Nvidia's pricing power.

Outlook for Investors

Whether Amazon can capture significant market share in the GPU business remains to be seen. The technical hurdle is high: Nvidia's software ecosystem, particularly the CUDA platform for parallel computing, is deeply embedded in AI development. Amazon would need to deliver not only comparable hardware but also build an ecosystem that moves developers and companies to switch.

For investors, the constellation presents different perspectives: Amazon's chip initiative can improve AWS margins in the long term, provided in-house production is cheaper than purchasing from Nvidia. At the same time, infrastructure bottlenecks show that growth in the AI sector is not limited solely by chip availability. Nvidia, meanwhile, faces increasing competition from in-house developments by major customers but retains the lead in external AI chip sales for now.

The share price performance of both companies will depend on how quickly Amazon can catch up technically and commercially and whether Nvidia can defend its lead in performance and software integration. In the long term, the AI chip market should be large enough to support multiple providers – but the distribution of market share remains uncertain.