Palantir Can't Satiate Demand: What Breaks in the FDE Org at 100

Palantir reported on Monday, and the number everyone will quote is $984 million: US commercial revenue, up 148 percent in a year, past the government business for the first time. The lines I'd quote are Alex Karp's answers on the call. Asked about the bootcamps, he said the timeline from first interaction to pilot or production "is completely shifted," and then: "What we have to do internally is figure out how to re-resource our company to focus exclusively on those people, because we don't have even the bandwidth to do that." Later, on CEOs who still aren't convinced: "we really cannot satiate demand."

The company that invented the forward deployed engineer, with 686 US commercial customers, each worth 88 percent more on a trailing basis than a year ago, says it's out of people.

The same day, Peter Baumann published a State of Forward Deployed Engineering with the hiring plans in one place. Microsoft's Frontier Company at about 6,000 engineers and $2.5 billion, which Makenzie Holland reported from Judson Althoff's launch post in July. TCS planning up to 8,900, per Ana-Maria Stanciuc at TNW. Anthropic's $100 million to train 10,000. Consultancies saying they need ten times the FDE headcount, in teams of 20 to 100. Against all of it, Baumann cites Gartner: a few thousand practicing FDEs. And this morning Park Min-joo at Seoul Economic Daily called it a race.

This is part 14 of The FDE Transition. Part 13 was the tooling. This one is the last planned item in the arc, the FDE org at 100 and what breaks on the way there. The inventor just said the thing nobody with a 100-FDE hiring plan wants to hear. At the top of the model, demand still outruns the seat.

The short version

An FDE team doesn't scale by adding FDEs. It scales by needing fewer of them per deployment, and the org at 100 is where every shortcut taken at 10 comes due. Five things break between the tenth hire and the hundredth: the handback stops being read, the engineers end up on islands, the maintenance tail eats the team, the reporting line fragments, and the bench appears. Palantir's answer to its own shortage isn't a thousand more FDEs. It's bootcamps that compress discovery, contracts that start at $25 million instead of $250,000, and a platform that keeps what the engineers built. That's the loop this series is about, run at scale, and it's the only version of the org at 100 that isn't a consulting firm.

What the inventor said, and what it didn't say

Karp didn't say Palantir is hiring a thousand FDEs. He said the company has to "re-resource," which is a word for moving people it already has. Dave Glazer's expense line says where the new ones come from: a new grad class and "the most elite technical talent." Ryan Taylor's examples say what each of those people is now worth. An industrial materials company piloted at nearly $3 million in Q2 and converted to a $29 million ACV deal in Q3. A home appliance company went from a $1 million pilot to $15 million. An apparel company that came in through an agent camp signed a $2.5 million pilot and converted to $11.5 million. Gennaro Cuofano at FourWeekMBA added up the four cases: about $8.1 million of pilots into about $76.5 million of ACV, not like-for-like, as he notes.

Read that as an org chart. Customers up 29 percent. Revenue per customer up 88 percent. The engineer count didn't go up 148 percent. The dollars per engineer did. The scaling lever at the inventor was never headcount, it was deal size and the time each deployment takes, and Karp's "instead of being 250K, it's $25 million to start" is the number a 100-FDE plan should be built around.

The org design chapter said size the team to shrink per deployment. The call is what that looks like from the top: compress the walk-up (the bootcamp, "show and don't tell" in Taylor's words), shorten the contract, expand it after. Nothing in that answer is a hiring plan.

Five things that break between ten and a hundred

The handback stops being read. At ten FDEs, one product manager reads every packet. I've been that PM, and it's a Friday afternoon. At a hundred FDEs running two deployments a year each, it's two hundred packets, and nobody reads them. Part 3 made the packet the FDE's main deliverable to product, and the loop chapter gave the PM the editor's job. At scale the editor is a desk, not a person: two or three PMs whose whole job is reading what the field built and writing the generalization eval, with the generalization rate from the scorecard reported per pod. If the packets have no reader, the FDEs find out within a quarter. Then they stop writing them.

The engineers end up on islands. Desmond Loh wrote up three years of an internal FDE practice in August, and the management review found what every scaled team finds. "FDEs often felt like they were on an island," with thin mentorship, unclear handovers, and two units solving the same problem twice. His fix was pods, a manager and a tech lead that an FDE belongs to without leaving the customer, plus quarterly huddles to look at the whole system. Recruiting from Scratch's scale guide puts the first formal manager at four or five FDEs, the point where a senior one spends over 30 percent of the week coordinating, and a director with regional structure at fifteen. Both say the layer arrives late, because the best FDE got promoted into it, and field delivery and people management aren't the same skill.

The tail eats the team. This is the one the hiring plan never has a line for. In Milos Mandic's survey, twelve of 24 FDEs spend more than a tenth of their week keeping delivered systems running, and seven said the systems they build stay with them after handover. I wrote about those seven last week. Loh's version: every quick-win tool became something somebody had to maintain, then app sprawl, then burnout. Now scale it. At 100 FDEs with two-year tails, by year three a third of the org is a support desk with a field title, and it's the third that shipped the most. Month two was the fix for one deployment. At a hundred it's a gate: a deployment that hasn't passed the independence test by day 120 counts against the pod every quarter it stays attached.

The reporting line fragments. Mandic's 24 respondents report into five different places: engineering, the CEO, services, sales, and product. On one team that's a quirk. At a hundred, a new region gets stood up under the regional sales leader because sales asked first, and the handbook's reporting-line rule fails one geography at a time. The tell is a scorecard that differs by region. Pipeline in EMEA, generalization rate in North America, and you have two orgs with one name, with sales owning the promise in one of them.

The bench. Hire in headcount, and demand arrives in deployments. The Gartner warning Mukul Saha gave Matt Ashare at Channel Dive in September, a large team of consultants sitting on the bench bleeding money, is Karp's bandwidth problem in reverse, and the same plan produces both, in different quarters. Loh's practice had a third version: every full-time FDE left a backfill gap on the product team they came from. TCS's up to 8,900 is the biggest headcount number in the market, and the question isn't whether they can hire them. It's what each one is deployed against in month four.

What the org at 100 is for

The five failures have one shape. Each is a place where the org kept doing at a hundred what worked at ten because a person was doing it. One PM read the packets. One senior engineer mentored the rest. One FDE maintained what one FDE shipped. One leader held the reporting line. None of that is a system, and the hundredth hire is where the person runs out.

Palantir's call shows the alternative without calling it one. Bootcamps replaced a person's walk-up with an event. Starting contracts replaced a person's scope negotiation with a floor. And the platform, the thing Shyam Sankar means when he says only Palantir has FDEs and everyone else has sales engineers, replaced a person's handback with a product that keeps the work. Part 2 said the shortage is a product problem. The org at 100 is the same argument with a headcount attached. If the product absorbs what the field builds, each of the hundred needs less time per customer than each of the ten did. If it doesn't, the team at 100 is a consulting firm with a model vendor's logo, and the economics chapter has the margin math.

What this means in the seat, and at the top

If you're the FDE: find out who reads your packet, by name. If the answer is a mailbox, you're on an island whether the org chart says so or not.

If you run product: the editor's job doesn't scale as a person. Staff the desk before the fiftieth FDE, and publish the generalization rate per pod.

If you run engineering: the tail is yours to count. FDE hours on systems shipped more than 120 days ago, per pod, per quarter. If it's rising, the harness isn't absorbing the work, and part 4 says what it owes.

If you're the exec with a 100-FDE plan: write it in deployments per engineer and dollars per deployment. The inventor's response to not having the bandwidth was bootcamps, $25 million starting contracts, and a platform. Not a req.

One thing to try this week: three numbers for your FDE team, however many you have. Deployments per FDE this quarter against last. Share of FDE hours on systems shipped more than 120 days ago. Handback packets a named PM read last quarter. Flat, rising, and zero means you're building the dev shop, and the hundredth hire makes it worse, not better.

Sources: Palantir Technologies, Q3 2026 earnings call, with Alex Karp, Dave Glazer, and Ryan Taylor, transcript via Investing.com (November 2, 2026) · Gennaro Cuofano, "Palantir AI Pilots Became Deals Up to 15x Bigger, Call Shows," FourWeekMBA (November 2, 2026) · Peter Baumann, "State of Forward Deployed Engineering (FDE)," Data Strategy in a Nutshell (November 2, 2026) · Park Min-joo, "U.S. Big Tech Races to Embed AI Engineers Inside Client Firms," Seoul Economic Daily (November 3, 2026), machine-translated from Korean · Makenzie Holland, "Microsoft pours $2.5B into push to embed engineers with customers," CIO Dive (July 2, 2026), on Judson Althoff's launch post · Ana-Maria Stanciuc, "TCS bets on 8,900 AI deployment engineers to defend India's IT services model," TNW (July 13, 2026) · Desmond Loh, "Three Years of Forward Deployed Engineers (FDE)," Medium (August 22, 2026) · Recruiting from Scratch, "How to Build a Forward Deployment Engineering Team at Scale (2026)" (June 25, 2026), author unnamed · Milos Mandic, "State of FDE 2026: The Survey Results," FDE Hub (October 27, 2026) · Matt Ashare, "Are forward deployed engineers flirting with failure?", Channel Dive (September 30, 2026) · Matt Ashare, "Palantir touts forward deployed engineers, reports record revenue growth," Channel Dive (August 5, 2026), with Shyam Sankar quoted

Part of the running argument on AI Product Management: once building is cheap, the scarce work is deciding what gets built for everyone, and an FDE org that grows to a hundred without that decision being somebody's job is a services firm that hired the title.

Related answer: What breaks when a forward deployed engineering team scales?