Amazon Sells Custom AI Chips to Challenge Nvidia's Market Control
5h ago|4 min readStandard
Fazen Markets Editorial Desk
Collective editorial team · methodology
ai-chipssemiconductors
Fazen Markets Editorial Desk
Collective editorial team · methodology
Trades XAUUSD 24/5 on autopilot. Verified Myfxbook performance. Free forever.
Risk warning: CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. The majority of retail investor accounts lose money when trading CFDs. Vortex HFT is informational software — not investment advice. Past performance does not guarantee future results.
Amazon has confirmed plans to sell its custom-designed inferentia and trainium AI accelerator chips to external customers, directly challenging Nvidia's dominance in the critical AI hardware market. The strategy was reported by Barrons on 18 June 2026, marking a significant escalation in the cloud provider's competition with the premier chip designer. This commercial move follows years of internal development for Amazon Web Services and arrives as Nvidia trades at $210.69, up 1.58% on the day, and Amazon shares hover at $244.39 as of mid-morning UTC today.
Nvidia has established a commanding position in the data center GPU market, with analysts estimating its share exceeds 80% for AI training workloads. The last major challenge to its architectural dominance came in late 2022 when multiple cloud providers, including Google with its TPU and Microsoft with its Maia chip, announced custom silicon projects. These were largely kept for internal use, preserving Nvidia's stranglehold on the commercial market.
The current catalyst is the explosive demand for inference, the phase of AI where trained models answer queries. Inference workloads are forecast to represent over 70% of AI compute spending by 2027, as models like those from OpenAI and Anthropic are deployed at scale. Nvidia's general-purpose GPUs, while powerful, are often seen as over-engineered and expensive for this specific task. Amazon's decision to commercialize its chips now targets this massive, growing segment where cost-per-inference is the primary metric.
The market data illustrates the scale of the competition Amazon is entering. Nvidia's stock has a market capitalization exceeding $5.2 trillion, buoyed by its AI monopoly. In contrast, Amazon trades with a market cap of approximately $2.5 trillion. Amazon's custom chips, like the Trainium2, claim to deliver up to 4x better price-performance for training large language models compared to previous generations. For inference, the Inferentia2 chip promises 50% higher throughput and 40% lower latency than its predecessor.
A key comparison lies in cloud provider spending. Prior to this move, AWS, Microsoft Azure, and Google Cloud were collectively forecast to spend over $150 billion on data center hardware in 2026, with a significant portion flowing to Nvidia. By selling its own chips, Amazon aims to capture a direct revenue stream while potentially reducing its own capital expenditure on third-party hardware. The financial impact will be measured in gross margins; Nvidia's data center gross margin sits above 70%, a target for any successful challenger.
The direct second-order effect is pressure on Nvidia's pricing power and market share in the inference segment. Companies like Meta and Tesla, which operate massive AI inference clusters, could become early customers for Amazon's more cost-optimized chips. This would directly reduce Nvidia's total addressable market. It also benefits semiconductor foundry TSMC, as Amazon's chips are manufactured on its advanced processes, diversifying its customer base beyond Nvidia and AMD.
The acknowledged risk is software lock-in. Nvidia's CUDA platform is the entrenched ecosystem for AI developers. Amazon's success hinges on its ability to provide a competitive software stack, including frameworks like PyTorch and TensorFlow. Without strong software, even superior hardware can struggle for adoption. Recent flow data shows increased institutional interest in semiconductor supply chain companies, like ASML and LAM Research, as the market diversifies away from a single vendor.
The first major catalyst is Amazon's Q2 2026 earnings call, scheduled for late July. Management commentary on early customer interest and projected revenue from chip sales will be critical. The second catalyst is the launch timeline for Nvidia's next-generation Rubin architecture, expected in late 2026, which will define the new performance benchmark.
Key levels to watch include Nvidia's stock price holding above its 200-day moving average near $205 and Amazon's ability to break through resistance at its yearly high of $245.73. The performance of the iShares Semiconductor ETF (SOXX) relative to the Nasdaq will indicate whether the market views this as a net positive for the broader sector or a zero-sum share shift. The conditional outcome hinges on whether Amazon can secure design wins with at least one other major cloud provider or large-scale enterprise by year-end.
Google's Tensor Processing Units (TPUs) have been available externally since 2018, but primarily through Google Cloud's platform-as-a-service model. Customers rent access to TPU pods but cannot purchase the physical chips. Amazon's strategy is more direct, selling the chips outright for deployment in customer-owned data centers or other clouds. This is a more aggressive commercial challenge to Nvidia's hardware business model.
For AI model developers, increased competition in the hardware market should lead to lower compute costs over time. This could improve the economics of running large-scale inference services. It also provides more architectural choice, allowing companies to optimize different parts of their workload on different chips. However, it may add complexity to their engineering teams, who must now support and optimize for multiple hardware backends.
Yes, but indirectly. Amazon is likely to use this strategy to lower its own infrastructure costs for AWS. Those savings could be passed on to customers using AWS's AI services, like SageMaker and Bedrock, in the form of lower pricing or more generous free tiers. The competitive pressure may also force Google Cloud and Microsoft Azure to re-evaluate their pricing for AI-centric virtual machine instances to remain competitive.
Amazon's commercial chip sale directly attacks Nvidia's most profitable market, aiming to fragment the AI hardware ecosystem and capture value.
Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.
Vortex HFT is our free MT4/MT5 Expert Advisor. Verified Myfxbook performance. No subscription. No fees. Trades 24/5.
Position yourself for the macro moves discussed above
Start TradingSponsored
Open a demo account in 30 seconds. No deposit required.
CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money.