The FDE Brief · Monthly

Demand for forward-deployed engineers is now universal. The people are not.

  • An executive-search study put the number of US engineers who can reliably deliver AI ROI at about 2,000, out of roughly 17,000 FDEs in the market, with much of that supply already inside Palantir.
  • The same study found companies planning FDE hires went from 5 to 10% in early 2026 to 70% by the second quarter, with demand projected to rise 2,100% by year end.
  • Palantir reported Q2 revenue of $1.935B, up 93% year over year, with US commercial up 149%: the clearest proof the deployment-led model scales.
  • Gartner expects more than 85% of tech providers to launch FDE programs by the end of 2026, and separately expects seven in ten enterprises to abandon FDE-led agentic projects over cost and missing internal skills.
  • Travelers and Liberty Mutual are building internal FDE teams rather than buying the capability, the first sign that large enterprises may staff the function themselves.

July settled the demand side: every major vendor now sells embedded engineers. August priced the supply side, and the numbers do not reconcile. About 2,000 people in the country can do the work to the standard buyers expect, against demand forecast to rise twentyfold. Palantir's quarter shows what the model earns when it works. Gartner's warning shows what happens when it is staffed by people who cannot do it.

The reads

PalantirPalantir
The talent mathJuly 30, 2026

About 2,000 engineers can actually do the job

An executive-search study by Christian & Timbers, based on 250 C-suite interviews across 180 companies and 300 FDE interviews between January and June, estimates around 2,000 US engineers have the mix of domain knowledge and applied AI experience to reliably deliver enterprise AI returns. The firm was explicit that this is 2,000 total, not 2,000 available. It counted roughly 17,000 FDEs in the US market overall, a large share already at Palantir, and found the share of companies planning FDE hires rose from 5 to 10% in early 2026 to 70% by the second quarter. TechCrunch→

The read

Set the number against July's announcements. AWS said thousands, Microsoft said 6,000, the Pentagon said hundreds, Ode has about 100. All of them recruit from the same pool of roughly 2,000 people who meet the bar, and the study was clear that pool is not sitting available. Two things follow from that arithmetic. Compensation keeps rising, which the comp data already shows. And a share of engagements gets staffed by engineers who do not clear the bar, because the qualified headcount to fill them does not exist. Ode's CEO put it directly in the same piece: very few are capable of building your flagship AI product.

PalantirPalantir
The proofAugust 4, 2026

Palantir grows 93% on the model everyone is copying

Palantir reported Q2 2026 revenue of $1.935B, up 93% year over year. US revenue rose 115% and US commercial revenue rose 149% to $764M; US government revenue rose 90% to $809M. The company closed 220 deals worth $1M or more, including 70 above $10M, ran a 63% adjusted free-cash-flow margin, and raised full-year guidance to about $8.15B. Quartz→

The read

This is the model every arm launched this year is copying, and the quarter shows what it produces at maturity. Palantir has spent twenty years building the bench, the tooling, and the customer relationships behind these numbers, so a 93% growth rate is not a template for a services arm three months old. The deal mix carries more information than the growth rate. 220 contracts above $1M, and 70 above $10M, means the revenue sits in large, long engagements rather than volume.

The forecastAugust 3, 2026

Gartner expects almost every vendor to run an FDE program, and most buyers to walk away

Gartner analyst Alex Coqueiro expects more than 85% of tech providers to have launched forward-deployed engineering programs by the end of 2026 as the main way they deliver AI. Gartner separately projects that seven in ten enterprises will abandon agentic AI projects led by FDE engagements, citing high ongoing cost and insufficient internal skills. Travelers and Liberty Mutual are building their own internal FDE models instead of buying the capability. CIO Dive→

The read

Both forecasts can be right, and together they describe the year ahead. FDE becomes the default way software is delivered, and most of those engagements disappoint, for the reason the talent study explains. The detail worth watching is Travelers and Liberty Mutual staffing the function internally. If large enterprises decide embedded engineering is a capability they own rather than rent, the vendor arms end up serving the middle market, and the enterprise work turns into hiring and training rather than services revenue. Gartner's cost objection points the same way: the recurring expense of an outside pod is one of the two reasons it expects most of these engagements to end.

More from August 20263 more, in brief
  • July 15, 2026

    OpenAI's Partner Network goes live

    A $150M program in three tiers, aiming to certify 300,000 consultants by the end of 2026. OpenAI is building delivery capacity it does not have to employ. OpenAI→

  • July 29, 2026

    A seed-stage vendor productizes the embedded engineer

    Zig.ai launched Enterprise Forward Deployment, putting one of its engineers inside each customer account. The motion has reached companies with no bench to speak of. Yahoo Finance→

  • August 3, 2026

    Enterprises ask for help as deployment complexity mounts

    Demand-side reporting on why buyers are turning to embedded engineers, and why Gartner thinks most of those engagements will not last. CIO Dive→

Zoom out

The demand side is settled; the supply side is arithmetic

The last four months moved fast on one side of the market. OpenAI and Anthropic committed billions in May, AWS and Microsoft answered with about $3.5B in July, Ode opened mid-month, and Gartner now expects almost every technology provider to run an FDE program by December. Nobody is arguing about whether embedded engineering works.

What August added is the constraint. Roughly 2,000 people in the US can do the work to the standard buyers are paying for, against demand forecast to rise 2,100%. That gap does not close with capital, and it explains both the compensation numbers and Gartner's expectation that most FDE-led projects fail. For anyone buying, the question worth asking is no longer whether a provider has a deployment arm. It is who specifically will be in the room, and whether they have shipped this before.

What we’re watching

  • Whether enterprises build or buy. Travelers and Liberty Mutual are staffing internal FDE teams. If that spreads, vendor deployment arms lose the enterprise segment and the work becomes training, not services.
  • The quality gap. With demand far past supply, most engagements will be staffed by engineers who do not meet the bar. Watch whether buyers start asking who is on the team rather than which vendor is on the contract.
  • Gartner's seven in ten. The same analyst house expects near-universal vendor adoption and majority buyer abandonment. Watch which half of that forecast the second half of the year confirms.
  • Whether the role survives automation. The study's own author raised the possibility that in two years the work is automated and the role thins out. Watch whether agent tooling starts absorbing the integration work FDEs do by hand today.
  • Palantir's alumni pool. Much of the qualified supply sits inside one company that the labs have already twice bought teams from. Watch how long that holds.

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.