SeaOtter ( Kentaro Mori )

Posted on Oct 23

What Is a Forward Deployed Engineer (FDE)? Origins, Business Models, and How AI Labs and Japanese IT Firms Are Using It

#ai #career #openai #anthropic

Introduction

On October 22, 2026, OpenAI held a live event called "Working at the Frontier: Forward Deployed Engineering at OpenAI" and published a page describing the program.

https://openai.com/business/forward-deployed-engineering/

The term FDE (Forward Deployed Engineer) has become very common over the past year. At the same time, people often ask questions like these:

If the job is to go to a customer, listen to what they need, and build a system, how is that different from a solutions engineer or a contract developer?
The role apparently started at a company called Palantir. Why are AI companies adopting it one after another right now?
In the end, how does this model make money?

This article tries to answer those questions by organizing public information (company announcements, earnings materials, news reports, and job postings) in chronological order. Everything here reflects what was known as of October 23, 2026. FDE is a way of working that is still taking shape, so treat this as a first edition that will be updated as things change.

Key points

An FDE is not just "a developer stationed at a customer." Two things set it apart from traditional contract development: taking responsibility until the system is used in production and delivers results, and feeding what was learned in the field back into the company's own product or AI models.

It started at Palantir. Things built in the field were cleaned up by the product team, turned into product features, and monetized through software licenses.

In 2025, a widely discussed study reported that most enterprise generative AI pilots were not producing results. In 2026, OpenAI, Anthropic, Microsoft, AWS, and others committed close to $10 billion in total to FDE organizations.

There are roughly three ways to make money with FDEs: recover the cost through product licenses, recover it through deployment fees plus AI usage fees, and charge for outcomes or a fixed price instead of hours worked. Large Japanese IT companies are adopting the third model.

Terms used in this article

Term	Meaning
Production	The state where a system is used in everyday work, not just tried out
PoC / pilot	A small trial to check whether something works before full adoption
Person-month billing (ningetsu)	Pricing based on "one person working for one month." It has long been the standard in Japanese system integration
Usage-based pricing	Fees that depend on how much you use. For AI, this is usually the amount of text processed (tokens)
Gross margin	Revenue minus the direct cost of delivering the product or service. Software companies tend to be high; services companies that rely on people tend to be low
Private equity (PE)	Investment firms that buy stakes in companies, improve their operations, and profit from the increase in value

1. What did OpenAI announce?

First, here is what the OpenAI page says.

Item	Details
Role	FDEs join the customer's team, observe how work gets done, and bring AI into production
Process	Define the outcome → Discover → Prototype → Validate → Go to production → Scale
Tools	OpenAI Presence (an AI agent platform), Context Layer (a shared base for company knowledge), process mapping, and code modernization with Codex
Examples	T-Mobile (55% of customer chats handled without handoff to a person), BBVA (80% accuracy in credit risk assessment), modernization of 15,000 systems
Organization	OpenAI's own FDEs work together with its subsidiary, DeployCo (the OpenAI Deployment Company)

Figure 1. The flow of an OpenAI FDE project (based on OpenAI's page)

The FAQ on that page contains two answers that show what kind of role this is.

To the question "Do FDEs provide professional services?", the answer is "No." FDEs are described as engineers who work with professional services partners to solve problems at the frontier of AI.
It also says: "If something can be achieved with existing AI capabilities, an FDE is usually not needed."

In other words, OpenAI's FDEs do not take on just any request. They go into cases where current AI is not yet good enough to run in production, and they report the model weaknesses they find to the product and research teams. Broad rollout to many companies is handled by DeployCo and partners.

2. History: from Palantir to the AI labs

Figure 2. A short history of FDE (compiled from various sources)

2-1. Origins: Palantir (late 2000s to 2010s)

The FDE role was created by Palantir, the US data analytics software company. It is credited to Shyam Sankar, who joined in 2006 as the company's 13th employee and is now its CTO (profile). Palantir originally built software for government and intelligence agencies. Those customers handle a lot of classified information, and their work is highly specialized. The usual approach of "listen to the requirements, take them back to the office, and build" did not solve their problems well.

Palantir split the people who went out to customers into two roles (Palantir blog):

Delta: Forward Deployed Software Engineer. The person who writes code and builds things on site
Echo: Deployment Strategist. The person who owns the problem and the outcome, and moves the customer organization

The official blog explains that "in theory, Echos are more product-manager-like and Deltas are more technical, but in reality both roles are a mix of product management, engineering, and strategy."

Nabeel Qureshi, who worked at Palantir for about eight years, wrote the essay "Reflections on Palantir" (2024), a valuable inside account of how this worked. The main points:

FDEs spent three to four days a week at the customer's office, learning the work of demanding industries such as manufacturing, healthcare, intelligence, and aerospace.
The first major commercial project was with Airbus in Toulouse, France. The team built software that removed bottlenecks in A350 production and significantly increased the production rate.
FDEs prioritized speed and accepted rough, hacky solutions to get something working. The product development team then cleaned these up into stable software that many customers could use.
Many features of Palantir's main product, Foundry, came from the product team automating what FDEs had been doing by hand in the field.

Figure 3. Palantir's "field → product → next field" loop (based on Qureshi's essay)

2-2. "Isn't Palantir just a services company?"

The model of employing many FDEs was criticized for a long time: "That's not a software company, it's a consulting firm." This criticism was loud around the company's 2020 listing as well. Qureshi himself writes that around 2016 it was "partly fair."

Since then, however, the numbers have looked like those of a software company.

Gross margin in 2023 was about 80% (from Qureshi's essay, which compares it with about 32% at Accenture).
In Q2 2026, revenue grew 93% year over year to $1.935 billion, and US commercial revenue grew 149% to $764 million (Palantir earnings release). Gross margin is reported at about 85% (secondary source).

Because improvements made in the field keep getting absorbed into the product, the share of manual services work does not grow much as the customer base grows. That is Palantir's explanation of why it did not become a services company.

2-3. AIP Bootcamp (2023 onward)

In 2023, as generative AI spread, Palantir launched the AIP Bootcamp. Its FY2024 annual report (10-K) describes it as a way to "deliver real workflows on actual customer data in days." It is usually described as an intensive workshop of about five days, in which FDEs and the customer build AI use cases on the customer's own data. Since this leads into full contracts, FDEs also act as the front door for sales.

2-4. Renewed attention in the generative AI era (2025)

There are two main reasons FDEs suddenly drew attention in 2025.

The problem of pilots not turning into results became clear. In its report "The GenAI Divide: State of AI in Business 2025," MIT's NANDA initiative found that about 95% of enterprise generative AI pilots had no measurable impact on profit and loss (Fortune). The cause was attributed not to model quality but to how organizations adopt the technology. Note that this figure comes from early-stage research, and the reported sample sizes differ between news outlets.
Job postings grew sharply. There are reports that FDE postings grew more than 800% from January to September 2025 (Financial Times, via Fast Company) and an analysis showing 1,165% year-over-year growth (Bloomberry). The methods differ, so the numbers are not directly comparable, but they agree on the scale: roughly 8 to 12 times in one year.

OpenAI's FDE team also started with two people around this time and had grown to about 140 by August 2026, according to Colin Jarvis, OpenAI's head of FDE (Engineering Leadership).

2-5. The year of big commitments (2026)

In 2026, companies began creating dedicated organizations and companies and committing large amounts of money.

Date	Company	What happened	Scale
May 4, 2026	Anthropic	Announced an enterprise AI services joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs. Launched in July as Ode with Anthropic	$1.5B ($300M each from Anthropic, Blackstone, and H&F), about 100 engineers
May 11, 2026	OpenAI	Launched the subsidiary OpenAI Deployment Company (DeployCo) and agreed to acquire Tomoro in the UK	Over $4B initial investment, about 150 FDEs and specialists from Tomoro
June 30, 2026	AWS	Created a new FDE organization that sends teams to customers for 45-day engagements	$1B
July 2026	Microsoft	Created the Frontier Company (an internal unit; some see it as a reorganization of its consulting services)	$2.5B, about 6,000 people
Sept 8, 2026	Accenture and Google Cloud	Set up a dedicated Gemini Enterprise unit	Target of 1,000 FDEs
Oct 2, 2026	Anthropic	Launched the Claude Frontier Academy to train partner engineers in a medical-residency style program	$100M, 10,000 engineers by the end of 2027

Sources: OpenAI, CNBC, TechCrunch, The Next Web (AWS), GeekWire (Microsoft), Directions on Microsoft, Anthropic

Investor Tomasz Tunguz estimates the money committed to FDE efforts over the past year at about $9.75 billion across five companies (The FDE Arms Race). OpenAI's October 22 event is part of this trend.

3. What kind of work is it?

3-1. What gap does an FDE fill?

Figure 4. The deployment gap and the role of FDEs (based on OpenAI's explanation)

OpenAI calls this the "deployment gap." No matter how capable AI becomes, it does not produce results unless it is connected to data scattered across the company, unwritten business rules, access control, and the habits of people on the ground. FDEs go inside the customer to close that gap.

3-2. How is it different from existing roles?

To answer "How is this different from a solutions engineer or contract development?", here is a comparison with similar roles.

	Pre-sales solutions engineer	Management consultant	Contract development (person-month)	FDE
Main output	Demos and proposals	A recommendations deck	A system built to spec	A system used in production that produced results
Who defines the problem	The customer	Worked out together	The customer (requirements)	Found together, including whether it is worth solving
Writes code?	A little	No	Yes (with a large team)	Yes (with a small team)
Where responsibility ends	The sale	The report	Delivery and acceptance	Production use and adoption
Where learnings go	Sales materials	Firm know-how	Often stay inside the project	Into the company's product or AI
Basis for fees	Product sales	Time	Headcount × duration	Product or AI usage, fixed price, outcomes

Figure 5. Where FDE sits relative to similar roles (positions are approximate)

The differences come down to two points.

Responsibility until results show up in production Traditional contract development usually ends at "build what was asked for and deliver it," and consulting ends at "make a recommendation." FDEs own the work until the business numbers actually improve.
A path back into the product This is the biggest difference. In contract development, what is learned on a project often stays within that project. FDEs pass what they learn back to the product team at Palantir, or to the research team at OpenAI. For example, at OpenAI a field report that "the model keeps making simple mistakes when converting COBOL to Java" was passed to the research team as a request for more training data in that area (Engineering Leadership).

Put the other way around, an FDE with no path back into the product is just on-site contract development with a new name. This is a useful lens when looking at the Japanese market later.

3-3. What the work looks like day to day

Job postings and interviews paint the following picture.

A lot of time on site: At OpenAI, about half of working time is spent with customers. It varies by region: about 80% in the UAE and Japan, and two to three weeks a month in Europe and the US (Engineering Leadership). Anthropic's postings list 25–50% travel.
Example work: One OpenAI example from Japan is modernizing the code of a customer with many legacy systems. Days alternate between meetings with teams such as payments, writing tests, and building migration code and progress dashboards.
Skills in demand: According to an industry site that aggregates postings, the most requested skills are customer communication, travel and on-site work, and Python. Only 32% of postings ask for LLM expertise (Synon FDE daily, Japanese). What matters more than AI knowledge is the ability to build something that actually works in the customer's environment.
Pay: Anthropic's US postings list $280K–$320K, and one analysis puts senior FDEs at AI labs at $350K–$550K (Tunguz).

3-4. Does this mean "AI writes the code and engineers just talk to customers"?

Some people read FDE that way. It is half right and half not quite.

What's right: Thanks to generative AI and coding agents, the effort needed to build has dropped sharply. As a result, a small team of three to five people can take a single use case all the way to production (see Fujitsu below). With less need for large teams, deciding what to build matters more.
What's not quite right: FDEs are not people who only talk. Palantir's Deltas and OpenAI's FDEs are engineers who write code and ship it to production. The planning and coordination role is often split into a separate role, such as Palantir's Echo or Salesforce's Deployment Strategist.

In short, the job is not "focus on talking to customers." It is to go into the customer's workplace and, together with them, use AI as a tool to build something all the way through.

4. Business models: how does it make money?

Looking at the companies adopting FDEs, there are roughly three ways they make money (this grouping follows an overview by a Japanese FDE consultancy).

Figure 6. Three FDE business models

Model A: Product adoption (Palantir, Salesforce)

FDE work is treated as an upfront investment. Bootcamps and pilots start small and short, and the company does not try to recover FDE costs at the start.
The cost is recovered through software licenses once the product is embedded in the business. Reports describe services as "a means to drive product adoption, not the primary revenue stream" (Yahoo Finance).
For this model to work, field improvements must keep flowing into the product. Otherwise it turns into an unprofitable on-site service.

Salesforce follows the same model, using FDEs to drive adoption of its AI agent product Agentforce. Its postings pair FDEs with Deployment Strategists, closely mirroring Palantir's Delta and Echo split.

Model B: AI lab and cloud deployment companies (OpenAI, Anthropic, Microsoft, AWS, Google)

Figure 7. Money flow in the deployment company model (based on company announcements and reports)

In this model, FDE work generates two kinds of revenue.

Deployment fees: DeployCo and Ode charge customers. Pricing has not been disclosed.
AI and cloud usage fees: Once AI agents run in production, they are used every day, and the usage fees flow continuously to the AI lab or cloud provider. This is most likely the real goal. A pilot consumes very little, but once AI is used across a whole company, usage grows by orders of magnitude.

Another notable feature is the partnership with private equity firms.

DeployCo has 19 investors led by TPG, including consultancies and integrators such as Bain & Company, McKinsey, and Capgemini. According to OpenAI, its partners sponsor more than 2,000 businesses worldwide (OpenAI).
At Ode, Blackstone and H&F refer their portfolio companies as potential customers, but Ode also sells to other companies (CNBC).

For PE firms, this is a way to improve operations at portfolio companies with AI and increase their value. For AI labs, it means access to more than 2,000 potential customers from day one. There are also reports that DeployCo's investors have a guaranteed minimum return of 17.5% a year (Tunguz). This could not be confirmed in official announcements, but if true, it suggests the business is expected to have high margins.

Microsoft, AWS, and Google have their own clouds, so their model is simpler. Workloads that FDEs bring into production translate directly into Azure, AWS, and Google Cloud consumption.

Model C: IT firms and consultancies selling outcomes (Hitachi, Fujitsu, NTT DATA, Accenture)

For IT companies and consultancies that do not have their own AI models, FDE is a way to move away from person-month billing.

Small teams, short cycles: At Fujitsu, three to five FDEs work on a single use case (Weekly BCN, Japanese).
Fixed or outcome-based pricing: Fujitsu says "the price is fixed and we don't show a breakdown. The deal happens if the customer judges that it's worth that amount." NTT DATA describes "a service package priced based on the customer's problem and expected results, not on person-month rates" (same article).
Reselling partner products: Fujitsu combines its services with Palantir's products, and Accenture with products from Anthropic, OpenAI, Google, Workday, and others.

Figure 8. Person-month contracting vs. the FDE model (based on Weekly BCN and other sources)

When generative AI reduces the effort needed to write software, pricing by "hours × rate" runs into a contradiction: the more you use AI, the less you earn. This is likely the main reason large Japanese IT companies are investing in FDE.

Comparing the three models

	A. Product adoption	B. Deployment company	C. Outcome-based
Examples	Palantir, Salesforce	OpenAI DeployCo, Ode, Microsoft, AWS, Google	Hitachi, Fujitsu, NTT DATA, Accenture
FDE cost	Upfront investment (mostly unrecovered)	Partly recovered through deployment fees	Recovered through fixed or outcome-based fees
Real revenue source	Software licenses	AI and cloud usage	Service fees and resale margin
Where learnings go	The company's product	The company's models and product	The customer and partner products
Main risk	Productization falls behind and it becomes a services business	Customer lock-in concerns, PE return demands	It becomes person-month billing with a new name

5. What each company is doing

5-1. Outside Japan

Company	Initiative	Notes
Palantir	FDEs (Delta) and Deployment Strategists (Echo), AIP Bootcamp	The original. Estimated at about 400–500 FDEs, about 12% of all employees (Tunguz)
OpenAI	Its own FDEs (about 140) focus on feeding insights into product and research, while DeployCo (about 200) scales delivery	Headcounts as of August 2026 (Engineering Leadership). Tomoro's customers include Tesco, Virgin Atlantic, and Supercell
Anthropic	FDEs in the Applied AI team, the Ode joint venture (CEO Chris Taylor, co-founder of Fractional AI), and the Claude Frontier Academy	Ode is "Claude-first" but uses other models when needed. The Academy trains engineers at partners such as Accenture, Deloitte, and McKinsey rather than hiring them into Anthropic
Microsoft	Frontier Company ($2.5B, about 6,000 people, led by Rodrigo Kede Lima, former president of Microsoft Asia)	Most staff come from existing teams. Some see it as a reorganization of its consulting services
AWS	FDE organization ($1B) sending teams for 45-day engagements	Customers include the Allen Institute, Cox Automotive, the NBA, the NFL, Ricoh, and Southwest Airlines
Google Cloud	Hiring its own FDEs, plus a 1,000-FDE Gemini Enterprise unit with Accenture	Also partnered with Hitachi on FDE-driven "physical AI" (June 2026)
Salesforce	Pairs Agentforce FDEs with Deployment Strategists	Plans to roughly double its UK FDE headcount to about 95 (ITPro)
Accenture	Anthropic Business Group (about 30,000 people trained on Claude, Dec 2025), a Gemini Enterprise unit with Google, and an FDE program with Workday (Oct 2026)	Acts less as a company promoting "FDE" as its own job title and more as the delivery partner for each AI vendor's FDE motion

Because Accenture does not have its own AI models or a flagship software product, it fits Model C rather than A or B: it builds a dedicated unit for each AI vendor and takes on deployment work. In 2026 it has also started using the term "FDE" in its work with Google and Workday.

5-2. Japan

Company	Initiative	Notes
Hitachi	Plans to build an FDE organization of about 1,000 people in Japan during fiscal 2026. Launched the "Physical AI FDE Team" service on Sept 3, 2026	Uses AI-driven interviews to draw out the know-how of skilled workers and structure it so AI can use it. Partnered with Google Cloud (Nikkei, Japanese, MONOist, Japanese, Google Cloud blog, Japanese)
Fujitsu	Set up an FDE division in April 2026. Became a Palantir Global FDE Partner on Sept 10	Has worked with Palantir since 2020. Senior Palantir FDEs join Fujitsu's teams for set periods. Examples include about ¥500 million a year in savings in maintenance parts management, and about 98% less effort to draft personnel transfer plans at Trusco Nakayama (Fujitsu, Japanese, Weekly BCN, Japanese)
NTT DATA	"DataEnablement Suite," covering everything from finding the problem to implementation and adoption	Sold as an outcome-based service package rather than by person-month. In one case it identified about 130 issues at a regional power company
NEC and others	Among the "four major IT companies" Nikkei reported as expanding FDE	At their July–August 2026 earnings briefings, the companies all spoke of moving away from person-month billing (Nikkei xTECH, Japanese). Details are behind a paywall and could not be confirmed
Dentsu Group	Began developing about 3,000 FDEs and FDMs (Forward Deployed Managers) across four Japanese group companies (Oct 14, 2026)	Grows FDE capacity by training existing staff rather than hiring
AI and SaaS companies	LayerX, Loglass, AI Shift, JAPAN AI, SB OAI Japan, and others have created or are hiring for FDE roles	76 FDE postings at 47 companies in Japan as of Oct 10, 2026 (unofficial count from an industry site)

What stands out in Japan is that large IT companies are adopting FDE as a way to move away from person-month billing (Model C). Outside Japan, AI labs and software companies lead; in Japan, large IT companies are becoming the main players. As with Fujitsu partnering with Palantir and Hitachi with Google Cloud, they also serve as local delivery partners for overseas vendors' FDE motions.

6. What comes next?

6-1. Why this is likely to become a lasting trend

The bottleneck has moved from "model capability" to "how AI is adopted." As the gap between models narrows, what differentiates is how deeply AI is embedded in the business. Ode's chief technologist put it this way: "model selection matters, but it's not where the majority of calories are spent" (TechCrunch).
Building got cheaper, so small teams can reach production. Integration work that used to take a large team months can now be done by a small team that knows how to use AI, in a much shorter time. That makes Palantir's approach reproducible at other companies.
It matches how money flows. AI labs earn from usage, cloud providers from consumption, and IT firms from moving beyond person-month billing. Whatever their position, "being used in production, continuously" is what drives everyone's revenue.

6-2. Things to watch out for

Drifting into a services business Software companies have gross margins in the 70–80% range, while people-based services companies are said to be around 30–40%. If FDE learnings are not fed back into the product, the company becomes a low-margin consultancy. There is persistent criticism that "FDE is just a new name for implementation consultants" (n47).
Lock-in The more FDEs build business processes on a specific vendor's AI, the harder it becomes to switch. For the vendor that is a moat; for the customer it can feel like a toll booth (Tunguz). Customers should check whether what is built and learned stays with them, or only improves the vendor's product.
Competition for talent People who combine strong engineering, business understanding, and interpersonal skills are rare. Anthropic choosing to train 10,000 partner engineers instead of hiring them reflects this shortage.
A new name on old work As discussed in section 3, what matters about FDE is "owning production results" and "feeding learnings back into the product." Simply renaming on-site contract development as FDE meets neither. As the Fujitsu example of "fixed price, no breakdown" shows, whether the contract itself has changed is a good way to tell.

6-3. What this means for engineers

Three things stand out as expectations for engineers.

Being evaluated on outcomes, not output As AI takes over more of the code writing, the value of deciding "what to build and why" and of "keeping it running in production" goes up.
Domain knowledge OpenAI says it deliberately builds FDE teams that mix industry experts with people from other industries.
Turning field lessons into something reusable Whether you can generalize a one-off solution so the next customer can use it is what separates FDE from traditional contract work.

What FDE is ultimately aiming for is this: go into the customer's workplace, use AI to build until results show up, and leave that experience in a form that can be reused next time.

Summary

FDE is a role Palantir created in the late 2000s. It worked by absorbing field solutions into the product and earning through software licenses.
In 2025 it became clear that many generative AI pilots were not producing results, and in 2026 OpenAI, Anthropic, Microsoft, AWS, Google, and others set up dedicated organizations and companies and made major investments. OpenAI's October 22 event is part of that trend.
What sets it apart from traditional contract development is owning production results and having a path to feed learnings back into the product or AI.
The money comes from product licenses (Model A), deployment fees plus AI and cloud usage (Model B), or fixed and outcome-based pricing (Model C). Large Japanese IT companies are using Model C to move away from person-month billing.
Whether something labeled FDE really is FDE becomes easier to judge by looking at the shape of the contract and where the learnings go.

FDE is still taking shape. This article is a first edition and will be updated as the companies involved make their next moves.

References

Company announcements and official materials

OpenAI, "Forward deployed engineering for enterprise AI" https://openai.com/business/forward-deployed-engineering/
OpenAI, "OpenAI launches DeployCo to help businesses build around intelligence" (May 11, 2026) https://openai.com/index/openai-launches-the-deployment-company/
Anthropic, "Claude Frontier Academy" https://anthropic.com/news/claude-frontier-academy
Anthropic, "Anthropic and Accenture partnership" https://anthropic.com/news/anthropic-accenture-partnership
Blackstone press release https://www.blackstone.com/news/press/anthropic-partners-with-blackstone-hellman-friedman-and-goldman-sachs-to-launch-enterprise-ai-services-firm/
Palantir Q2 2026 earnings release https://www.sec.gov/Archives/edgar/data/0001321655/000132165526000039/a2026q2ex991pressrelease.htm
Palantir FY2024 10-K https://www.sec.gov/Archives/edgar/data/1321655/000132165525000022/pltr-20241231.htm
Palantir Blog, "A Day in the Life of a Palantir Deployment Strategist" https://blog.palantir.com/a-day-in-the-life-of-a-palantir-deployment-strategist-951cb59a5a96
Fujitsu press release (Sept 10, 2026, Japanese) https://global.fujitsu/ja-JP/pr/news/2026/09/10-01
Google Cloud blog, Hitachi and Google Cloud on physical AI with FDE (Japanese) https://cloud.google.com/blog/ja/products/gcp/hitachi-and-google-cloud-collaborate-on-the-social-implementation-of-physical-ai-using-fde?hl=ja

News and analysis

Nabeel S. Qureshi, "Reflections on Palantir" https://nabeelqu.substack.com/p/reflections-on-palantir
Engineering Leadership, "Inside OpenAI's Forward Deployed Engineer" https://newsletter.eng-leadership.com/p/inside-openais-forward-deployed-engineer
Tomasz Tunguz, "The FDE Arms Race" https://tomtunguz.com/fde-arms-race/
TechCrunch (Ode with Anthropic) https://techcrunch.com/2026/07/15/anthropic-blackstone-bet-the-next-trillion-dollar-ai-business-is-implementation-not-models/
CNBC (Anthropic joint venture) https://www.cnbc.com/2026/05/04/anthropic-goldman-blackstone-ai-venture.html
GeekWire (Microsoft Frontier Company) https://geekwire.com/2026/microsoft-announces-2-5b-frontier-company-to-embed-ai-engineers-inside-customers
Directions on Microsoft https://www.directionsonmicrosoft.com/microsoft-launches-its-own-forward-deployed-engineering-unit-the-frontier-company/
The Next Web (AWS) https://thenextweb.com/news/aws-1bn-forward-deployed-engineers-enterprise-ai/
ITPro (Salesforce) https://itpro.com/security/salesforce-doubles-down-on-forward-deployed-engineers-in-uk-agentforce-push
Fortune (MIT NANDA, "The GenAI Divide") https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo
Bloomberry, "I analyzed 1,000 forward deployed engineer jobs" https://bloomberry.com/blog/i-analyzed-1000-forward-deployed-engineer-jobs/
n47, "The Forward Deployed Engineer dilemma" https://www.n47.com/insights/the-forward-deployed-engineer-dilemma
Nikkei, "IT大手4社、『FDE』拡充急ぐ" (Japanese) https://www.nikkei.com/article/DGXZQOUC183EE0Y6A810C2000000/
Nikkei xTECH, "IT大手4社、脱・人月へFDE拡充に意欲" (Japanese) https://xtech.nikkei.com/atcl/nxt/column/18/00001/11956/
Weekly BCN, on FDE at Fujitsu and NTT DATA (Japanese) https://www.weeklybcn.com/journal/feature/detail/20260907_217086.html
MONOist, Hitachi HMAX solutions (Japanese) https://monoist.itmedia.co.jp/mn/articles/2609/07/news035_3.html
Synon, "FDE industry daily" (Oct 15, 2026, Japanese) https://synon.co.jp/article/news/2026-10-15-fde-daily
FDE Consulting, "Companies adopting FDE in 2026" (Japanese) https://fde-consulting.jp/blog/fde-companies-japan

Note: Some figures in this article, namely DeployCo's valuation and investors' minimum return, Palantir's FDE headcount, and FDE salaries, come from investor blogs and job posting aggregates and could not be confirmed in official announcements. FDE job growth figures also differ in period and method depending on who compiled them.

DEV Community