Microsoft, AWS, OpenAI Race to Embed AI Engineers Inside Customers

Microsoft launched a new $2.5 billion unit on July 2, 2026, called Microsoft Frontier Company, that will embed 6,000 AI deployment engineers inside customer organizations. Amazon Web Services announced a $1 billion version of the same idea two days earlier. OpenAI and Anthropic had already moved down this path in May, with $4 billion and $1.5 billion ventures respectively, bringing the four-provider total to roughly $9 billion.

The four initiatives are different in structure but identical in thesis: AI models sold on their own have not translated into enterprise productivity gains, and the bottleneck is now the human talent required to deploy them. The playbook the four companies are following, forward-deployed engineering, was pioneered by Palantir two decades ago for military and intelligence work. Microsoft commercial business CEO Judson Althoff framed the new unit as the largest such effort yet in the industry.

Microsoft’s $2.5B Frontier Company Leads the Latest Push

Microsoft announced the new subsidiary on Thursday, July 2, calling it Microsoft Frontier Company. The unit pulls together 6,000 engineers, AI specialists, and industry consultants from inside the company, and will send them inside customer organizations to design, build, and run AI systems on the customer’s own infrastructure. Rodrigo Kede Lima, who had been running Microsoft’s Asia business across 20 countries and a team of 30,000, will lead it as president.

Microsoft has not confirmed whether the $2.5 billion represents new spending or repurposed budget, or over what period it will be spent. The company has also not detailed how the new unit will affect its existing consulting and services arms. Microsoft commercial business CEO Judson Althoff positioned the unit against every comparable effort underway elsewhere in the industry.

This goes beyond what has been labeled as Forward Deployed Engineering (FDE) and will be the largest, most capable, outcome-driven engineering organization in the industry.

Althoff, who runs Microsoft’s commercial business, pointed to Palantir as the company that popularized the FDE title, noting that the U.S. military has long relied on Palantir software and that the company sent FDEs to U.S. bases in Afghanistan. Microsoft already runs large forward-deployed engineering programs with Accenture and a $1 billion, five-year alliance with EY, and the Frontier Company is its biggest, most concentrated bet on the model so far.

The early customer roster spans finance, consumer goods, agriculture, and global consulting, per Microsoft’s announcement as reported by TechCrunch:

• London Stock Exchange Group
• Unilever
• Land O’Lakes
• Accenture

Read the full breakdown of the six thousand engineers Microsoft is embedding, and the detailed look at Frontier Company’s origins and Palantir model.

AWS Got There Two Days Earlier With a $1 Billion Pod

Amazon Web Services announced its own version on June 30, two days before Microsoft did. The cloud provider committed $1 billion to a new Forward Deployed Engineering organization that will seed the unit with thousands of engineers. AWS framed the move as agentic-first, designed to compress AI deployment timelines from months to days by pairing human engineers with AI agents. AWS vice president of frontier AI engineering and services Francessca Vasquez called the unit the first such effort from a hyperscaler.

An initial pod of roughly five or six engineers will be embedded within an AWS customer at a time, working alongside the customer’s own engineering, security, and business teams. AWS said the engagement is structured to leave behind self-sufficient teams, with deployed systems, knowledge graphs, runbooks, and trained internal champions, rather than billable hours. The cloud provider lists six organizations already working with its FDE teams: the Allen Institute, Cox Automotive, the National Basketball Association, the National Football League, Ricoh, and Southwest Airlines. AWS has been building AI solutions for customers since 2017 and ran a Generative AI Innovation Center for three years before standing up the FDE unit. See AWS’s $1 billion Forward Deployed Engineering announcement and the breakdown of AWS pods and named customers.

OpenAI and Anthropic Built the Template, With Private Equity Money

OpenAI went first in May 2026, when it announced a separate company called the OpenAI Deployment Company. The new entity is majority-owned and controlled by OpenAI but structured as a standalone business with its own operating model, leadership, and customer focus. It launches with more than $4 billion of initial investment from a partnership led by private equity firm TPG, with Advent, Bain Capital, and Brookfield as co-lead founding partners.

The partner roster extends well beyond the four lead firms. Founding partners include B Capital, BBVA, Emergence Capital, Goanna, Goldman Sachs, SoftBank Corp., Warburg Pincus, and WCAS. The venture also works closely with Bain & Company, Capgemini, and McKinsey & Company as consulting and systems integration partners. OpenAI separately agreed to acquire Tomoro, an applied AI consulting firm, bringing approximately 150 experienced Forward Deployed Engineers and Deployment Specialists into the new company from day one.

Anthropic followed with a parallel structure in May, teaming with Goldman Sachs, Blackstone, and Hellman & Friedman on a $1.5 billion venture focused on embedding engineers inside mid-sized companies. The thesis the four providers are now converging on is that selling the model is the easy part, and that the actual work of plugging AI into a real enterprise is what determines whether the technology pays off. Goldman Sachs global head of asset and wealth management Marc Nachmann made that argument in an interview with CNBC.

Having the model alone doesn’t change your workflows or how you operate. You need people who can combine the technology with what’s actually happening in the business and implement those changes.

Marc Nachmann, Goldman Sachs’ global head of asset and wealth management, said that in an interview with CNBC. Here is how the four initiatives compare:

Full details on OpenAI’s Deployment Company announcement and partner list.

Why the AI Models Alone Weren’t Closing the Deal

The deployment push is the response to a sales problem the AI industry has been unable to wish away. Microsoft 365 Copilot, the software vendor’s flagship AI assistant for Office, has yet to gain anything approaching ubiquity in the business world. GitHub Copilot, the coding agent Microsoft also owns, has ceded market share to newer players.

Microsoft stock has slumped 21% this year, by far the worst performance among mega-cap tech companies. The March quarter brought about $2.1 billion in revenue from enterprise and partner services, up 2.5% from a year earlier, a thin gain against the tens of billions Microsoft has sunk into AI data centers. One concern on Wall Street is that AI models that quickly compose code might threaten mature software companies, the same companies whose services revenue supports Microsoft’s professional services arm.

Across the industry, businesses have adopted tools like ChatGPT, Claude, Gemini, and Copilot, only to find that impressive demos do not automatically translate into results inside a real company, with its own data, governance rules, and entrenched ways of working. So the AI providers are sending their own engineers to fix what the tools cannot.

The model the four providers are converging on was pioneered two decades ago by Palantir. AWS Generative AI Innovation Center engineers spent three years working on thousands of customer solutions before the FDE unit was stood up, including a partnership with BMW to reduce service disruptions across 23 million connected vehicles, with Jabil to build a manufacturing assistant for the factory floor, and with Lyft to resolve driver support issues 87% faster. The new FDE units formalize and expand what those teams were already doing on a smaller scale.

By the numbers

• $2.5 billion: Microsoft’s commitment to Microsoft Frontier Company
• $1 billion: AWS’s commitment to its Forward Deployed Engineering unit
• $4 billion+: Initial investment in OpenAI’s Deployment Company
• $1.5 billion: Anthropic’s venture with Goldman Sachs, Blackstone, and Hellman & Friedman
• 21%: Microsoft’s year-to-date stock decline, the worst among mega-cap tech

For context on the layoffs Microsoft announced the same week, see the 4,800 jobs cut across the company.

The Lock-In Problem Hiding Inside Model Flexibility

Microsoft’s pitch to enterprise customers is built on two promises. The first is privacy and trust: a customer’s data and hard-won knowledge stay the customer’s alone, and Microsoft will not feed them into training its models in ways that would hand the same advantages to rivals. The second is choice: customers can run whichever AI model fits the job, from OpenAI, Anthropic, Microsoft, or open-source providers, without being locked into one. Microsoft CEO Satya Nadella laid out the model-flexibility vision in a June 14 essay with a warning that applied to the AI industry itself.

“The last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see.” Nadella wrote. He argued that a company should be able to exchange one AI model for another without losing the institutional knowledge built up around it, and said that is the test for whether a business still controls its own future.

The risk Nadella’s essay does not address is that the deployment services themselves may end up doing the locking in. Even if a customer can theoretically swap in a competitor’s AI model, working with Microsoft’s 6,000 embedded engineers means systems naturally end up running on Microsoft’s cloud platform and related technologies, making it very difficult to jump ship. Microsoft Frontier Company is described by Microsoft as a purpose-built organization with its own leadership and financial accountability, but it is not a separate legal entity, meaning its contracts and customer relationships sit inside Microsoft. For more on how Microsoft is positioning the model-flexibility pitch, see Althoff’s interview on Palantir’s role in inventing FDE.

What Enterprise IT Leaders Are Actually Getting

For enterprise IT leaders, the four initiatives look different on paper but converge on the same delivery model. AWS FDE pairs an initial pod of five or six engineers with the customer’s own teams for what the company calls an agentic-first engagement that compresses timelines from months to days. The customer receives deployed systems running inside its own AWS account, plus knowledge graphs, runbooks, architectural documentation, and trained internal champions ready to operate independently, per the AWS announcement.

Microsoft Frontier Company starts from a deeper existing footprint, with engineers already embedded inside much of the Fortune 500 through Microsoft’s existing consulting and services work. OpenAI’s Deployment Company brings a different network: the private equity sponsors and consulting partners represent more than 2,000 businesses globally, which OpenAI says gives the new company a broad view of where AI can create value and which deployment patterns can scale. Whether the deployment service leaves the customer self-sufficient or quietly deepens their dependence on the vendor’s cloud stack is the question none of the four announcements has resolved.

Frequently Asked Questions

What is Microsoft’s Frontier Company?

Microsoft Frontier Company is a new internal organization the company announced on July 2, 2026. Microsoft committed $2.5 billion and pulled together 6,000 engineers and AI specialists from across the company to embed inside customer organizations. Rodrigo Kede Lima, formerly president of Microsoft Asia, leads the unit as its president. Microsoft has not confirmed whether the $2.5 billion is new spending or repurposed budget, or how the new unit will affect existing consulting services.

How much are AI companies spending on deployment services?

Microsoft committed $2.5 billion to Frontier Company, AWS committed $1 billion to its Forward Deployed Engineering unit, OpenAI’s Deployment Company launched with more than $4 billion from a partnership led by TPG, and Anthropic’s joint venture with Goldman Sachs, Blackstone, and Hellman & Friedman has $1.5 billion behind it. The combined commitment from the four providers is more than $9 billion, all announced between May and July 2026.

What is a forward-deployed engineer?

A forward-deployed engineer, or FDE, is a technical employee embedded directly inside a customer’s operations to design, build, deploy, and operate technology on-site rather than selling a tool and walking away. The role was pioneered by data analytics vendor Palantir more than a decade ago, with Palantir sending FDEs to U.S. military bases in Afghanistan to work alongside soldiers using Palantir software. AWS credits Palantir with popularizing the job title across the broader software industry.

Will embedded AI engineers lock customers into one vendor?

All four providers say no. Microsoft CEO Satya Nadella wrote in a June 14 essay that customers should be able to swap AI models without losing institutional knowledge, and Microsoft’s Frontier Company pitch explicitly offers model choice across OpenAI, Anthropic, Microsoft, and open-source providers. The unresolved risk is that the deployment services themselves create lock-in: once a customer’s systems run on a vendor’s cloud platform and are supported by that vendor’s embedded engineers, switching costs rise even if the underlying AI model is portable.

Why are AI companies sending humans instead of selling AI tools?

Because the tools have not sold themselves. Microsoft 365 Copilot has not gained anything approaching ubiquity in the business world, GitHub Copilot has ceded market share to newer coding agents, and Microsoft stock has slumped 21% this year, by far the worst performance among mega-cap tech companies. Across the industry, businesses that adopted ChatGPT, Claude, Gemini, and Copilot have found that demos do not automatically translate into production results inside companies with their own data, governance, and legacy workflows.