Amazon Web Services has started negotiations to sell its AI chips — Trainium and Inferentia processors — to data centers outside of AWS. This is a significant strategy shift. Until now, these chips were available exclusively within Amazon's cloud ecosystem.
CEO Andy Jassy called this move a $50 billion opportunity. Therefore, AWS is no longer just competing with Nvidia on performance: it is now also entering the hardware distribution market. However, the challenge is complex. Nvidia holds a dominant share in global AI infrastructure, with an established partner network and a software ecosystem — led by CUDA — that is hard to replicate.
For Italian cloud-native SMEs, this scenario opens up concrete prospects. In fact, greater competition among AI chip suppliers tends to reduce computing costs in the medium term. We at SHM Studio we monitor these dynamics to guide our clients' infrastructure choices toward solutions AI more efficient and sustainable. In short: the AI chip market is opening up, and SMBs would do well to follow its evolution.
Amazon's change of direction in the AI chip market
Until a few months ago, AI chips developed by Amazon — the processors Trainium for training and Inferentia for inference — were exclusively accessible through the AWS infrastructure. No external data center could buy them or integrate them into their own systems. Today, according to reports by TechCrunch , this logic is changing.
AWS has started formal negotiations to sell its chips to third-party data centers. CEO Andy Jassy has publicly estimated the value of this opportunity at 50 billion dollars . This is a figure that signals a precise ambition: no longer just a cloud provider, but an AI hardware supplier on a global scale.
Therefore, Amazon is no longer competing with Nvidia solely on the level of cloud services. It is now entering the physical processor distribution segment. This is terrain where Nvidia has built a structural advantage in recent years.
Why Nvidia remains a difficult opponent to beat
Nvidia currently controls an estimated 70% to 80% share of the AI chip market, according to analyses by Gartner . The advantage is not just technological. It's ecosystemic.
The platform CUDA — Nvidia's parallel programming framework — is integrated into almost all major machine learning frameworks: PyTorch, TensorFlow, JAX. Furthermore, the network of hardware and software partners built by Nvidia over time represents a considerable barrier to entry.
Amazon, on the other hand, focuses on an alternative ecosystem based on Neuron SDK . However, the maturity of this stack is still lower than CUDA. Consequently, convincing data centers to replace or supplement Nvidia GPUs with Trainium chips will require no small amount of technical evangelism.
Despite this, the strategic signal is clear. Amazon has no intention of staying confined to its own cloud. The move is somewhat reminiscent of the strategy adopted by Amazon itself with AWS in the 2000s: monetizing an internal infrastructure by turning it into a commercial product.
The AI chip market in 2026: a sector in redefinition
The context for this move is already buzzing. In 2025, several hyperscalers accelerated the development of proprietary chips. Google with TPUs, Microsoft with Project Maia, and Meta with its own custom accelerators. So, the trend of reducing dependence on Nvidia is not exclusive to Amazon.
What distinguishes the AWS strategy is the willingness to selling externally , not just for internal use. This creates a new market segment. In fact, independent data centers—those that do not belong to the major hyperscalers—might find an alternative supplier in Amazon at potentially more competitive prices.
On top of that, mixing up where hardware comes from has turned into a major geopolitical goal. The drama around semiconductor supply chains, which really flared up between 2022 and 2024, pushed a lot of players to look for backup suppliers. Amazon is stepping right into that gap at just the right time.
Concrete impact for Italian cloud-native SMEs
For small and medium-sized Italian enterprises operating in cloud environments — or considering integrating AI solutions in their processes — this news has practical implications, even if not immediate.
First off, more rivalry among AI chip makers usually brings down computing costs down the road. So, machine learning tasks—from sorting docs to predicting sales trends—could end up way more budget-friendly even for smaller businesses.
Secondly, having chips that can act as alternatives to Nvidia could mean more flexibility in what cloud providers offer. For instance, AWS might roll out Trainium instances for less than Nvidia GPU-powered ones, encouraging a switch over to more efficient AI setups.
Finally, for SMEs that already use AWS as their main infrastructure, the news suggests a strengthening of the Amazon ecosystem. Therefore, investing in skills for the Neuron SDK and Trainium instances could prove advantageous over the next two years.
What the official press releases don't say
Andy Jassy's statements are optimistic, as expected from a CEO presenting a new business line. However, some loose ends remain.
The first concerns the software compatibility . Selling chips to third-party data centers means they must be able to manage them independently. Without a mature and documented software ecosystem, the risk of slow adoption is real. We at SHM Studio we often observe this dynamic in projects of digital transformation : the technology is available, but adoption requires time and skills.
The second knot concerns the competitive positioning . Selling chips to competing data centers means, to some extent, empowering infrastructures that could compete with AWS itself. It is a strategic tension that Amazon will need to manage carefully.
The third element concerns the timing . Negotiations are still ongoing. There is not yet a product commercialized on a large scale. Consequently, SMEs should not expect immediate operational changes.
What to observe in the next 12-18 months
The most relevant time window for evaluating the impact of this strategy extends until the end of 2027. Certain indicators deserve special attention.
- Partnership announcements between AWS and independent European or American data centers.
- Neuron SDK Updates that increase compatibility with standard frameworks.
- Price variations of cloud instances based on Amazon chips compared to Nvidia ones.
- Nvidia's reaction : potential exclusive agreements or price reductions to retain data center partners.
For those who manage SEO strategies , google ads campaigns or LinkedIn campaigns with AI components, the cost of inference is already a significant variable today. Similarly, those who develop applications on web platforms with integrated language models will find this evolution something to keep an eye on.
To explore how these dynamics influence the technological choices of SMEs, the team at SHM Studio is available for a consultation . Furthermore, on our Blog we regularly publish analyses on AI, cloud, and digital infrastructure for the Italian market.
In short: Amazon's move is ambitious and structurally consistent with its history. However, the path to directly challenging Nvidia is still long. Italian SMEs would do well to follow the evolution, without expecting immediate revolutions. The AI chip market is reopening — and this, in the medium term, is good news for those investing in applied artificial intelligence to the business.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
