This is for executives and engineering leaders who've read the headlines about forward deployed engineering (FDE), where engineers work inside the customer's own systems, and want to know what the money means for their own AI plans. Four of the biggest names in AI are paying to put engineers into customer companies, which says a lot about where the difficulty in AI sits.
Below we separate the numbers we can source from our own interpretation. Then we get to the practical question for a company that isn't a giant cloud customer.
How much money is going into forward deployed engineering?
Forrester counts about $9B committed to FDE-style units by four providers between 4 May and 15 July 2026. That's Forrester's tally, and at least one component is Forrester's own estimate. It's the best public summary we have. Nobody has audited it.
| Provider | Commitment (per Forrester) | Note |
|---|---|---|
| OpenAI | $4B+ | Forrester's tally |
| Microsoft | $2.5B | Forrester's tally |
| Anthropic | About $1.5B | Anthropic's services venture; Forrester's own figure |
| AWS | $1B | Forrester's tally |
| Total | About $9B | 4 May to 15 July 2026 |
The source is Forrester's blog on FDEs as training wheels for AI reinvention. Forrester argues that FDEs are temporary by design, and that services firms should sell outcomes, build for handoff, and let their footprint shrink. We cover that in how an FDE engagement should end.
What did CIO.com report about Microsoft and AWS?
CIO.com reported that Microsoft and AWS are putting billions into thousands of forward deployed engineers. It also quoted analysts, including Info-Tech and Carmi Levy, who warn that FDEs supplied by a hyperscaler carry a risk of vendor lock-in for the customer.
You can read the original in CIO.com's report on Microsoft and Amazon's FDE spending. The lock-in point is straightforward. An engineer paid by a platform vendor will naturally build on that platform. That can be fine, and sometimes it's the right call. Just make sure you chose it on purpose.
CIO.com has also pointed out that the title covers two completely different roles: a product builder who decides what to build and writes it, and a platform operator closer to solutions engineering and customer success, covering roughly 5-10 accounts each. When a vendor says it will "send an FDE", ask which one.
What is Anthropic doing differently?
Anthropic announced in March 2026 that it is growing its partner-facing team fivefold to provide dedicated applied AI engineers on live partner deals. Membership of the Claude Partner Network is free and open to any firm. The partner focus stands out, because it sends engineering help to live deals that involve partner firms.
The details are in Anthropic's Claude Partner Network announcement of 12 March 2026. One possible implication for buyers (our reading) is that vendor engineering support may increasingly arrive through service partners. So ask any partner you hire what direct support they get from the model provider, and what they do when the vendor's help runs out.
Is the demand real?
Lightcast job-posting data shows very fast growth. It also shows consultancies filling the role as much as technology companies do, which affects how you should read the title.
- Fortune, citing Lightcast, reports FDE job postings up more than 1,000% in January to August 2026 versus the same period of 2025.
- The same report puts the median advertised salary above $188K, against about $145K for software engineers.
- TechRadar, citing Lightcast, reports about 200 US FDE postings in 2023, about 1,200 in 2025, and about 7,300 in the first nine months of 2026.
- TechRadar adds that three of the top four advertisers are consultancies: Deloitte, Accenture and KPMG.
Postings aren't hires, and the title is used loosely. Still, the direction is clear. The scarce resource in applied AI is engineers who can work inside a customer's systems.
What does this mean for a mid-market company?
The market is pulling senior engineers toward the largest accounts and the vendors' own platforms. If you're a mid-market company without a large cloud commitment, don't assume a vendor engineer will be assigned to you. Plan for who will do the deployment.
What follows is our interpretation, with no sourced statistic behind it. Large cloud and model commitments tend to attract dedicated engineering support, while smaller buyers more often get partner programs, documentation and credits. Advertised salaries above $188K also suggest senior engineers are expensive to hire directly.
That leaves a mid-market company with a few realistic routes. You can hire your own, which is possible but slow and costly at those salary levels, and you still need a first project to point them at. You can use a vendor's program, which works well if your use case sits squarely on their platform and you're comfortable with the lock-in trade-off. Or you can hire an independent FDE firm: a small pod of senior engineers, scoped to one problem, deployed in your cloud account, with handoff planned from the start.
The first-time buyer questions are covered in how to buy forward deployed engineering services. For the trade-offs between the last two routes, see vendor FDE versus independent FDE.
What does a typical FDE timeline look like as a benchmark?
Deloitte describes its FDE pods as about 4-5 people, time-boxed to roughly 3 days to a use case, 3 weeks to a prototype and 3 months to production. Use that as a benchmark when you compare proposals. If a provider can't say when you'll see production, treat the engagement as open-ended.
The figures come from Deloitte's forward deployed engineering page. For comparison, Kaprin places senior engineers in client teams, usually as small pods, on fixed-scope sprints of 4-6 weeks, with production deployment typically within 30-45 days. Firms define scope differently, so compare what's being delivered as well as the calendar. Our week-by-week view of the first 90 days shows what ours looks like.
What should you ask any provider, whoever they are?
Five questions separate a useful FDE engagement from a costly one. They apply equally to a hyperscaler, a model provider, a consultancy or a boutique firm, and the answers are usually clear within one call.
- Who owns the code, the database and the cloud account at the end?
- Which tools will you recommend, and do you earn anything from that recommendation?
- What is the baseline metric, and when will we see it move?
- When does production start, and what will I see every week before then?
- What does handoff look like, and what happens when you leave?
Kaprin's own answers: the client owns 100% of the code, database and IP in the client's cloud account or private VPC; engagements run as fixed-scope sprints of 4-6 weeks with weekly live demos; production deployment typically comes within 30-45 days. The forward deployed engineering overview sets out the model in detail.
Where does that leave you?
Read the $9B figure as a signal about the market. It says the biggest AI providers have concluded that software alone doesn't deliver the result. Your job is to decide who puts engineers in front of your real systems, and on what terms.
If you're weighing those options for a specific workflow, talk to us. We'll tell you honestly whether an independent FDE pod is the right fit or a vendor program would serve you better.
