The FDE Brief · Monthly

AWS committed $1B to a forward-deployed engineering organization. Microsoft committed $2.5B to one of its own.

  • AWS put $1B into a forward-deployed engineering group on June 30: thousands of engineers in teams of five or six, with the NBA, NFL, and Southwest as early customers.
  • Microsoft launched Frontier on July 2, a $2.5B business of roughly 6,000 experts, with LSEG, Unilever, and Novo Nordisk as launch customers.
  • OpenAI had committed $4B and Anthropic $1.5B to the same model in May.
  • OpenAI's Deployment Company agreed to acquire Northslope, founded by ex-Palantir engineers, subject to regulatory clearance. It would be its second such deal since May.
  • Anthropic's $1.5B venture opened on July 15 as Ode: about 100 engineers, backed by nine investors who own hundreds of companies between them.
  • The Pentagon launched War Force to recruit hundreds of engineers into two-year tours paying up to nearly $200K.
  • Pay kept climbing: about $215K median at Palantir, $560K to $785K for senior engineers at the labs, with postings up more than 1,000%.

In May, OpenAI committed $4B to a services arm and Anthropic $1.5B. In July, AWS and Microsoft announced comparable programs of their own, OpenAI bought a second team, and Anthropic opened Ode. That leaves buyers two questions: whether the people are available at all, and which cloud and models the ones they bring in will build on.

The reads

AWSAWS
Cloud providersJune 30, 2026

AWS makes forward deployment a $1B product line

On June 30, AWS committed $1B to a dedicated Forward Deployed Engineering organization, funded off Amazon's own balance sheet, that will embed thousands of engineers in pods of five or six working alongside AI agents. Led by Francessca Vasquez, it counts the NBA, NFL, Southwest Airlines, and the Allen Institute among early customers. Amazon→

The read

OpenAI and Anthropic raised outside money for their services arms. AWS funded this from its own balance sheet, which puts it in the cost of selling cloud rather than in a separate business. AWS says the team completes the engagement and leaves. The systems it builds run on AWS models and infrastructure.

MicrosoftMicrosoft
ScaleJuly 2, 2026

Microsoft launches a $2.5B Frontier Company

On July 2, Microsoft launched the Frontier Company, a $2.5B operating business of roughly 6,000 industry and engineering experts led by Rodrigo Kede Lima, with London Stock Exchange Group, Unilever, Land O'Lakes, and Novo Nordisk as launch customers. Commercial chief Judson Althoff said it goes beyond what has been labeled as forward-deployed engineering. Microsoft→

The read

Frontier launched with four named enterprise customers, not just a statement of intent. Althoff's description of the work as going beyond forward-deployed engineering suggests the larger vendors intend to define it under their own labels rather than Palantir's. Either way, the company supplying the engineers also supplies the model.

OpenAIOpenAINorthslopeNorthslope
AcquihireJuly 8, 2026

OpenAI's deployment arm acquires Northslope, a firm built by Palantir alumni

OpenAI's Deployment Company agreed to acquire Northslope, an applied AI startup founded by former Palantir forward-deployed engineers and backed by $22M in Series A funding. The deal is the Deployment Company's second acquisition since its May launch (following Tomoro) and adds hundreds of engineers to OpenAI's team. Terms were not disclosed; the transaction is subject to regulatory clearance. Axios→

The read

Within two months OpenAI has agreed to buy two firms built by people who left Palantir, Northslope and Tomoro. Christian & Timbers says Palantir still employs a large share of the engineers with this experience.

Ode with AnthropicOde with Anthropic
Services ventureJuly 15, 2026

Anthropic's services venture opens as Ode

On July 15, Anthropic, Blackstone, and Hellman & Friedman launched Ode with Anthropic, the operating brand for the $1.5B venture announced in May. It starts with about 100 engineers under CEO Chris Taylor and CTO Eddie Siegel, both co-founders of Fractional AI, the firm Anthropic bought in May to seed the team. Goldman Sachs, General Atlantic, Leonard Green, Apollo, and GIC are also backers. Ode joins the Claude Partner Network alongside Accenture, Deloitte, and PwC. BusinessWire→

The read

Ode's backers are private-equity firms that own hundreds of companies between them, so its first customers come from its own investor base rather than from open-market sales. No other arm launched this year started with that. How many become customers is undisclosed.

U.S. Dept. of DefenseU.S. Dept. of Defense
GovernmentJune 30, 2026

The Pentagon starts hiring forward-deployed engineers

On June 30, the Department of Defense and the Office of Personnel Management launched War Force, a drive to recruit hundreds of software and AI engineers into two-year tours paying up to nearly $200K, posted on USAJobs, to embed technical talent that deploys AI inside military operations. OPM→

The read

Palantir developed this model on government contracts twenty years ago, and the government is now hiring for it directly under the same name. Defense work carries procurement rules and clearance requirements, which limit which companies can place people inside those systems.

PalantirPalantir
The marketJune 2026

Pay and postings both point to a shortage

New compensation data put Palantir's forward-deployed median around $215K total, with senior FDEs at frontier labs reaching $560K to $785K and equity now 55 to 70% of top packages. Bloomberry's analysis of roughly 1,000 live FDE postings, using Revealera data, found hiring up more than 1,000% year over year across dozens of AI firms. Perspective AI→

The read

AWS, Microsoft, and the Pentagon have each announced hiring in the hundreds or thousands. None of the announcements says how quickly those roles can be filled, or how far the requirements overlap. Posted pay is one indication of how tight the experienced end of the market already is.

a16za16z
TrainingJuly 1, 2026

a16z starts training forward-deployed engineers

Andreessen Horowitz's eight-week FDE Fellowship began its first cohort in July, based in San Francisco and drawing engineers from across its portfolio. It is the clearest sign the industry is trying to train these people rather than only bid for them. a16z→

The read

A few dozen fellows does not move a market that needs thousands. It is more useful as an indicator: venture firms do not build training programs for roles they expect to be temporary. Supply takes years to build, and trained engineers remain the constraint.

More from July 20267 more, in brief
  • June 30, 2026

    AWS extends forward deployment to its partner network

    Alongside its $1B launch, AWS began building AWS-trained teams at chosen partners to deliver the same work. AWS→

  • June 11, 2026

    DXC and Anthropic will train tens of thousands of Claude-certified FDEs

    A multi-year alliance to train a dedicated workforce of Claude-certified engineers from DXC's staff. DXC→

  • June 11, 2026

    OpenAI acquires Ona to run Codex inside customers' clouds

    OpenAI is buying Ona (formerly Gitpod) to fold secure, persistent cloud execution into Codex, so its agents can run inside a customer's own environment: deployment plumbing for the agent era. OpenAI→

  • June 22, 2026

    Baseten raises $1.5B Series F at up to $13B

    The AI inference platform's raise, led by Altimeter, Conviction, and Spark, is a proxy for what production-deployment infrastructure is worth as pilots become live systems. Baseten→

  • June 22, 2026

    Groq closes $650M for its inference cloud

    More capital into the layer that runs models in production, the infrastructure forward-deployed teams operationalize once a pilot becomes real. Groq→

  • June 4, 2026

    Northzone: not every AI company needs an FDE

    VC firm Northzone argues the role is oversubscribed and best suited to companies still discovering their product, a useful check on a month of billion-dollar announcements. Northzone→

  • May 4, 2026

    From engineers to entities

    An analyst argues AI vendors are shifting from individual forward-deployed engineers to larger deployment organizations, the multi-billion-dollar ventures announced this year. AI Razor's Edge→

Zoom out

The largest vendors built their own deployment organizations

For a year this was a story about the AI labs. Palantir established the model, then OpenAI and Anthropic committed billions in May to services arms built around it. In July the two largest cloud providers announced their own, Google Cloud was already hiring for the role, and Anthropic opened Ode on the 15th. A job that existed at one company a few years ago is now a standard part of how large software vendors sell.

AWS and Microsoft funded their programs internally and staff them from their own organizations. OpenAI assembled its arm by acquisition. Ode is a separate company with outside investors. Buyers are comparing different cost structures, different model and infrastructure requirements, and different views on whether an engagement is meant to end.

What we’re watching

  • Lock-in versus self-sufficiency. Where a provider commits to leaving a customer self-sufficient, watch whether the systems it builds can be moved to another stack.
  • What it ends up being called. Watch whether forward-deployed engineering survives as the name or splits into vendor brands: Frontier, Deployment Company, Applied AI.
  • Whether supply actually moves. Watch whether fellowships, integrator training, and cloud hiring add engineers, or just raise the price of the ones who already exist.
  • Defense as its own market. Procurement rules and clearance requirements limit who can take part, which may split this work off from the rest.
  • Where Palantir's alumni end up. OpenAI has bought two firms founded by former Palantir engineers. Watch whether the labs absorb most of this pool.
  • Companies that run on more than one cloud. Watch whether they settle on one vendor's engineers or keep the job in-house.

The FDE Brief is written by Plank, once a month. Every item links to a primary source; the read is our analysis, not a summary. See all editions or send tips and corrections.