Amazon Web Services is committing $1bn to embed its own engineers inside customer companies. It is the first cloud giant to copy a playbook that Palantir built and that OpenAI and Anthropic have since adopted. The new unit, called Forward Deployed Engineering, will start with thousands of engineers working in small pods of five or six people. These pods will sit inside a single customer at a time, collaborating with the customer's business, engineering, and security teams.
Francessca Vasquez, the company’s vice-president of frontier AI engineering and services, set out the plan in an interview with CNBC. Her pitch came down to one word: speed. “The currency that the customers are always talking about right now is speed,” she said. She added that the model suits firms chasing quick returns for their executives and stakeholders. Vasquez framed the launch as a step change rather than a brand-new skill. “We’ve had capabilities over the years, but structurally this is like getting everybody together in one business unit with a common rubric of deployment,” she said. “It’s the first time we’re doing it in that way.”
What AWS is actually building
The new unit will start with what AWS calls “thousands” of engineers. It will send them out in small pods, each with five or six people, embedded inside a single customer at a time. Those engineers will also work alongside AI agents, the software tools that can carry out tasks on their own. The pods are meant to move fast. AWS said in a blog post that its engineers would sit with a customer’s business, engineering, and security teams, then hand back a self-sufficient team within weeks.
The concept of forward-deployed engineers originated at Palantir more than a decade ago. Palantir deployed software engineers directly inside government agencies and large corporations to help integrate their data analytics platforms. The idea has since spread to software firms that want faster adoption of their tools, and it now sits at the centre of the race to sell enterprise AI. By having engineers on-site, vendors can address technical hurdles in real time, customize solutions to specific workflows, and build trust with clients who may be wary of outsourcing critical AI infrastructure.
AWS is taking a different route compared to its rivals. In May 2026, Anthropic set up an AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs to help mid-sized firms roll out its Claude models. Days later, OpenAI launched its deployment company with TPG, Advent International, Bain Capital, and Brookfield, among others. Those rivals built their deployment arms as joint ventures, leaning on outside investors and consulting partners. AWS is funding the unit from its own balance sheet, with no partner firms attached. Google has made its own move too, with a $750mn partner fund aimed at agentic AI deployments.
Amazon has spent billions of dollars backing both Anthropic and OpenAI. It has also been clear about competing with them directly in places. An AWS spokesperson said the company still expected to work with the FDE arms of both labs, and promised more detail on its partner programmes soon. AWS has separately agreed to sell OpenAI’s models after Microsoft’s exclusivity lapsed.
Why a cloud giant wants bodies on the ground
The logic is about adoption, not headcount for its own sake. Companies have bought plenty of AI tools. Many have struggled to turn them into working systems. By placing engineers inside the customer, AWS hopes to close that gap and tie clients deeper into its cloud. The move also shows how AWS plans to defend its lead. Amazon is the biggest cloud provider by revenue, and it is the first hyperscaler to commit to an FDE unit at this scale. The bet is that hands-on help, not just cheaper compute, will decide who wins enterprise AI. Amazon has also pushed customers toward cheaper AI options as model costs climb.
The investment comes at a time when enterprises are facing mounting pressure to demonstrate return on AI investments. According to surveys, nearly 70% of companies that deployed AI in 2025 reported mixed outcomes, with many citing integration challenges as the primary bottleneck. By embedding engineers, AWS aims to reduce the time from purchase to production from months to weeks. This accelerated timeline is particularly valuable in industries like healthcare, finance, and manufacturing, where legacy systems and strict regulatory compliance often slow down deployment.
Not everyone will read the spend as a sure thing. Investors have grown wary of the huge sums flowing into AI, and they keep asking when the returns will land. A $1bn unit staffed by costly engineers adds to that bill. AWS is betting the outlay pays for itself in stickier, larger cloud contracts. The proof will sit in next year’s numbers, not in the launch. Amazon's history of long-term thinking, exemplified by its investment in AWS itself during the 2000s, suggests the company is willing to endure short-term costs for long-term strategic advantage.
There is a hiring story here as well. AWS wants thousands of engineers for the unit at a time when AI is eating into entry-level work. The roles it is creating are senior, client-facing, and hard to automate. That is a notable contrast with the junior jobs the same technology is removing. The need for human judgment, creative problem-solving, and deep understanding of client-specific contexts ensures that these positions are not easily replaced by AI agents. However, the integration of AI agents alongside human engineers points to a hybrid model where routine tasks are automated while complex decisions remain with people.
The customers already signed up
AWS named several early adopters. They include the Allen Institute, the National Basketball Association, the National Football League, and Ricoh. Vasquez said the next wave would come from heavily regulated industries that hold large, varied datasets. Those are the firms with the most to gain from faster deployment, and the most to lose from getting AI wrong. The Allen Institute, a nonprofit biomedical research organization, will likely use the embedded engineers to accelerate drug discovery and genomic analysis. The NBA and NFL, both data-rich sports leagues, may leverage AI for player performance analytics, injury prediction, and fan engagement. Ricoh, a global imaging and electronics company, could apply AI to document management and automation.
Beyond the named clients, AWS is likely targeting sectors such as financial services, where fraud detection and risk modeling require robust AI systems; healthcare, where diagnostic tools and patient data management need tight integration with existing electronic health records; and manufacturing, where predictive maintenance and quality control benefit from real-time data processing. These industries often handle sensitive data and operate under strict regulations like HIPAA, GDPR, and SOX, making on-site engineering support a critical differentiator.
The competitive landscape is intensifying. While AWS is the first hyperscaler to launch a dedicated FDE unit, it is not alone in offering hands-on support. Microsoft has its own FastTrack program for Azure, though it focuses on migration and optimization rather than AI deployment. Google Cloud provides Professional Services, but typically at a project-by-project basis rather than as a permanent embedded team. The scale of AWS's commitment—$1bn and thousands of engineers—signals a new phase in enterprise AI services, where customization and integration become as important as raw computing power.
For now, the move sharpens a question hanging over the whole sector. Businesses have spent heavily on AI and seen patchy results. Whoever turns that spending into working systems fastest will pull ahead. AWS has just bet $1bn that the answer is people, sent to sit at the customer’s desk. The success of this strategy will depend on whether the embedded engineers can effectively transfer knowledge and build self-sufficient teams within weeks, as promised. It also hinges on the willingness of customers to open up their operations to external engineers and share sensitive data. As AI becomes more pervasive, the boundary between vendor and client may continue to blur, with forward deployment becoming the new norm in enterprise technology partnerships.
Source:TNW | Anthropic News