This year marks AWS’s 20th anniversary. Looking back over these two decades, what stands out to you most about how cloud computing has transformed the way businesses build and innovate?

What I am most proud of is not only that AWS launched storage as one of its first services, but that we also fundamentally changed the economic model of the information technology (IT) industry.

When I was CTO at Amazon.com, reducing the cost of using a database required a long-term commitment, often of five years or more. However, it was difficult to know exactly how much capacity Amazon would need over such a long period, so we had to estimate our future requirements and often overestimated them.

Under that model, vendors effectively controlled the relationship. Customers had to make long-term commitments to secure lower costs, while vendors had little incentive to remain focused on the customer once they had received a multi-million-dollar payment.

Amazon has long followed the principle set out by Mr. Jeff Bezos, Founder and Executive Chairman of Amazon, in his first letter to shareholders: to be the Earth’s most customer-centric company. When AWS was launched, this principle was applied to the IT industry by putting customers in control rather than vendors.

Under the AWS model, customers pay only for the resources they actually use, without long-term contracts, and can reduce their usage or switch to another provider if they believe they are paying too much. This customer-centric economic model has changed the computing industry.

Greater transparency around costs has also changed how companies manage their digital systems. When costs are paid upfront, engineers may have limited visibility into the cost of individual resources. By making these costs explicit, companies can build more cost-aware architectures.

Cost is also a useful proxy for sustainability: the less customers pay, the fewer resources they use. As a result, more companies are asking their IT teams to report how much energy they consume and whether that consumption can be cut.

In particular, how do you view the position of emerging markets like Vietnam in terms of AI, cloud, and tech adoption both within ASEAN and globally?

Vietnam has several strengths that position it well in the development of AI, cloud, and other emerging technologies, particularly its education system and engineering workforce. 

It produces around 50,000 new engineers each year, with the total number now estimated at some 560,000. This is helping the country become an increasingly important software development hub. In the past, companies looking to outsource software development often turned to India. Today, Vietnam is increasingly becoming another destination, not only because of cost but also because of the quality and training of its engineers.

As AI and generative AI develop rapidly, the priority is to equip more engineers and businesses with the skills to use these technologies effectively. Many companies are asking what they need to do with AI, but the focus should be on understanding what the technology can actually do and determining which models are appropriate for specific applications. This is what AWS is working to provide, including free education for customers and their engineers.

The focus should not simply be on comparing models such as Anthropic, OpenAI, DeepSeek, and Qwen. Customers need to understand what these technologies can actually do and which models are appropriate for specific applications. This is one of the reasons AWS built Amazon Bedrock, which brings together different AI models and gives customers access to a range of technologies.

AWS is not a consumer company. It does not build consumer chatbots, but provides technologies that enable other companies to build chatbots, integrate AI into their workflows, and develop other applications.

Cost is another important consideration. If a customer works with only one AI provider, it is effectively tied to that particular model and its different model sizes. With Bedrock, customers can access multiple models and experiment to determine which one is most appropriate for a specific application.

According to our latest research, 26 per cent of Vietnamese businesses have already adopted AI - and that number is growing fast. The next step is to move from pilot programs into production. This requires companies to experiment with different models and assess their cost and performance. For applications such as creating a bestseller list or summarizing a conversation, for example, a highly-expensive, high-end model may not be necessary. An open source or more specialized model may be more appropriate.

Another important consideration is the development of culturally aware models. For Vietnam, where there are many different types of languages, enabling people to interact with AI in their local language and cultural context is important for building trust. AWS is working with companies in Vietnam to help them develop their own models using Vietnamese data through Amazon SageMaker.

There are already examples of local models in other markets, such as Abu Dhabi developing Falcon, Singapore with SEA-LION, Japan developing 15 models incorporating Japanese characteristics, and Chinese models reflecting local language and cultural contexts.

Culturally aware models are important because they allow businesses to build systems and applications that are better adapted to local needs. This is also closely linked to trust.

This illustrates how local models can help build trust in AI systems. Large language models are designed to generate plausible responses rather than guarantee factual accuracy. As statistical word generators, they can make mistakes and may not always indicate when they do not know an answer. Building trust therefore requires appropriate guardrails to ensure that AI systems provide reliable responses rather than simply generating plausible information.

Vietnam is also developing local models. But there is a difference in how AI is being approached across markets. In the US, the approach is largely driven by commercial models, with large companies, enterprise contracts, and products and services being sold to businesses. In parts of Asia, including China and South Korea, governments are taking a different approach by viewing AI as a fundamental service that should be accessible to everyone.

The adoption of agentic AI is still in its early stages in Vietnam. What are the key factors that can help local businesses move forward and unlock its full potential? 

Vietnamese businesses need to ensure that people are educated about AI and understand that it is a tool to help people.Humans remain particularly strong in creativity, judgment, empathy, and understanding the broader context, including interpreting what people actually mean. AI should therefore amplify human capabilities while taking over routine tasks.

The level of human involvement should depend on the level of risk. For lower-risk applications, such as marketing content or meeting summaries, limited human review may be sufficient. However, when AI involves sensitive information or supports healthcare, drug discovery, or other high-risk applications, stronger human review and verification are required.

As risk increases, businesses need stronger verification, and in some cases machines may also need to be used to check other machines. This is particularly important with agentic AI. Language models are effective at text manipulation, but agents can take action on a user’s behalf. As a result, the issue is not only whether the generated content can be trusted, but also what actions the agents are taking. Businesses therefore need to maintain control and establish appropriate guardrails.

Vietnam is training more developers, and access to talent is not considered a major problem. However, having enough developers is different from having developers with advanced AI capabilities. That is an area where further progress is needed.

AI is also changing what developers need to learn. At re:Invent 2025, I introduced the “Renaissance Developer” framework, which focuses on skills that are becoming increasingly important for developers.

The first is lifelong learning. Technology, programming languages, and tools continuously evolve, so developers need to keep learning and adapting.

Another important skill is systems thinking: understanding what is being built, the problem it is intended to solve, and how it fits into the broader system. Communication is also becoming increasingly important. Developers need to work with businesses and customers, rather than focusing only on technical implementation.

Developers also need to return to first principles and understand the actual problem they are trying to solve. AI may suggest what should be built, but developers need to determine whether AI is the right tool for that particular problem.

Ownership is equally important. If you build a system, you are responsible for it. Developers cannot simply blame AI when something goes wrong, particularly in areas such as financial services or healthcare.

The role of developers is also moving beyond deep specialization. Engineers increasingly need to combine technical expertise with a broader understanding of the systems they are building and how different components interact.

Finally, developers need to continuously adapt to new tools. Programming environments have evolved significantly over time, and new tools will continue to emerge. Engineers need to evolve with them.

That is something I see happening in Vietnam, where engineers play an important role in building the next generation of systems.

Most businesses in Vietnam are small and medium-sized enterprises (SMEs). Can these factors be fully applied to all business groups, or are they relevant only to large enterprises?

AI is not limited to large enterprises. Businesses of all sizes can use AI to improve efficiency, particularly in areas where employees spend significant time on routine work.

For example, in a law firm, an AI agent can search for and collect information needed for a legal brief, while a human remains responsible for interpreting the information and preparing the final document. AI can therefore support employees without replacing human responsibility.

Efficiency is one application. AI can also accelerate business operations. When an engineer investigates a customer issue, for example, AI can collect relevant information from customer databases, log files, and other sources, allowing the engineer to make decisions more quickly. AI can also support innovation by helping companies identify new products, services, and business opportunities. The focus is therefore not only on improving efficiency and the bottom line, but also on creating new opportunities for growth.

At the same time, companies need to continuously evaluate how AI is being used. One financial services company in Europe introduced AI chatbots in customer service and replaced half of its customer service staff. However, many customers contacting the company were facing difficult personal circumstances, including financial problems caused by purchases made by their children or loss of employment. These situations required empathy and human understanding that the AI system could not provide. The company subsequently removed the chatbots and rehired customer service staff.

This illustrates the importance of using AI where it helps and amplifies human capabilities. If customer service agents have access to relevant customer information through AI, they can work more efficiently and make better-informed decisions. However, human judgment and empathy remain essential in situations that require an understanding of individual circumstances.

AWS recently launched the Hanoi Local Zone. What does this mean for Vietnamese businesses and developers?

AWS Local Zone in Hanoi helps address data sovereignty and local data requirements, making it easier for customers to meet those requirements. But it is not only about that.

Having computing resources that can run locally can be very important for many customers because it helps reduce latency.

With Hanoi Local Zone, customers can keep their local data and run local compute on top of that. They can also run containers there. For customers with significant data processing requirements, particularly where that processing is on the path to their end customers, the Local Zone provides an ideal environment.

Over time, we will add more capabilities to the Local Zone. Of course, customers are now asking to run their AI models in the Local Zone as well, and we are working on technologies to enable customers to do that in the future.

You introduced the “Renaissance Developer” concept. What does the developer look like in 2026 especially in Vietnam in particular and globally? 

The “Renaissance Developer” concept is not only about developers, but also about how companies reward people who use these technologies. Productivity should no longer be measured simply by lines of code or how quickly a product is launched. Companies also need to place greater emphasis on broader capabilities, ownership, and communication rather than deep specialization in a single area.

At Amazon, cost and business requirements are closely connected. Decisions about which services need to be highly available are ultimately business requirements, so architects need to communicate with the business to understand those priorities.

For Amazon’s retail platform, services such as search, browsing, shopping carts, checkout, and reviews need to be highly reliable and continuously available. A second tier includes services such as recommendations and similar products, which are important but less critical. A third tier includes less essential features, such as bestseller lists, where temporary unavailability would have a smaller impact.

These different requirements involve different costs. Engineers need to explain those costs to the business and balance quality, availability, and investment. This ability to communicate with the business is an important part of the “Renaissance Developer” concept and helps ensure that what engineers build is sustainable for the business.

I also see younger engineers in Vietnam becoming increasingly concerned about climate change. Recent climate-related events, including floods in northern Vietnam and disasters linked to melting ice in Nepal, are increasing awareness of these issues. Younger engineers are increasingly considering whether the technologies they use are safe, sustainable, and based on renewable energy. At AWS, we are running on 100 per cent renewable energy and have more than 500 wind and solar projects around the world. We are also working to ensure that we consume less water than we give back.

Sustainability remains a work in progress, becoming an increasingly important consideration for engineers, companies and boards when choosing technologies. They are looking at how much energy technologies consume and whether those choices are sustainable for the future.

As physical AI and robotics continue to develop, where do you see the biggest opportunities for Vietnam?

Vietnam’s strong manufacturing sector provides significant opportunities for the development of physical AI and robotics. Manufacturing and supply chains have traditionally been labor-intensive, and introducing more robotics could help improve working conditions while creating new opportunities for innovation.

Physical AI and robotics will become increasingly important in areas where Vietnam has built capabilities over the past 25 years, particularly manufacturing and supply chains. These technologies can support workers by automating routine tasks and enabling them to focus on other activities.

Another important application is supporting an ageing population. Younger generations want to pursue their own careers while continuing to care for their parents. Physical technologies could help older people remain independent for longer in their own homes without relying as much on others.

Generative AI and robotics could also help address loneliness. AI-enabled devices that can communicate with people could provide companionship and support in everyday life. Such technologies do not necessarily need to resemble humans or animals; their value lies in helping people remain connected and independent.

Physical AI could also be integrated into everyday objects to support elderly people living independently. As healthcare improves and people live longer, technologies that help older people maintain independence and reduce loneliness will become increasingly important, not only in Asia but around the world.