Reflections on Palantir

For all you backend engineers: It’s basically Grafana with a bunch of support engineers in the backend cleaning up the data source (like a splunk index) that feeds it.

Palantir does UI and visualization well but needs an inordinate amount of field support engineers to groom the dirty disparate data that governments do a poor job cleaning (either due to incompetence, field conditions, or both).

The amount of manual labor doesn’t justify its market price, but because governments rarely change their vendors, there is significant lock in that probably supports some amount of their market cap.

I imagine back in the LBJ and Nixon days IBM would've been doing similar classified work.

> needs an inordinate amount of field support engineers

Hey now, they're forward-deployed engineers. Nothing like Oracle or SAP consultants.

Leaving this here for people interested in what the software actually is.

https://www.palantir.com/docs/

Touché

But they have 80% margin, according to the article... so those engineers are generating a lot of revenue per capita.

Do they dig latrines too?

“Forward deployed” sounds like they’re in a FOB out in the sticks somewhere.

> they have 80% margin, according to the article

I have a pet theory about private equity: they're in the business of laundering boring jobs for college graduates. Few kids dream of graduating college to work at a chemicals plant in Baton Rouge. But working for Accenture in New York or Atlanta, now that's sexy. Even if you spend your entire work week *checks notes* working at a chemicals plant in Baton Rouge. (Investment banking is similar, though the transaction orientation makes the division of labour a little more sensible.)

Palantir pays less for its consultants (sorry, FDEs) than Bain et al. Few in their generation dreamed of graduating college to work at a soulless corporate consultancy. But a tech company, now that's sexy.

More pointedly: It's remarkable how an ostensibly 80% GM business only barely became profitable last year. Palantir's Q2 '24 cash flows from operations at 40% of revenues looks closer to the mark [1]. (Palantir's cost of revenue "primarily includes salaries, stock-based compensation expense, and benefits for personnel involved in performing [operations & maintenance] and professional services, as well as field service representatives, third-party cloud hosting services, travel costs, allocated overhead, and other direct costs" [2].)

[1] https://www.sec.gov/ix?doc=/Archives/edgar/data/0001321655/0...

[2] https://www.sec.gov/ix?doc=/Archives/edgar/data/0001321655/0...

I like this theory! And I don't think it's a cynical one either—this "laundering" could actually be really useful.

The worker gets the status and security of a tech/consulting job, while having more variety than actually working at the chemical plant, not being at the whims of their org chart, and also just the reframing probably makes it more enjoyable anyway. All the while, the important work is getting done.

I don't think it's cynical at all! I do think it's a decision-delaying choice, however, in that it treats one's work as a series of electives. The person working at the plant, gaining seniority and building deep connections is on their way to industry expertise. It's trading wealth and power for prestige. (It makes sense it's like catnip to our graduates from elite schools.)

This is such a hilarious oversimplification.

>inordinate amount of field support engineers to groom the dirty disparate data that governments do a poor job cleaning

Getting clean data seems like a universal need, but the job is still difficult, under-appreciated and underpaid. How come?

Hilarious if true. Still hilarious if not.

I think the other side of the chemical plant job is that the salary is higher than a college grad would get from the plant itself out of college.

Nobody at the chemical would ever pay a college grad VP^h^h consultant salary to work there.

(I did stuff like this out of college - got paid hourly ~ 3x normal employee salary at non-sexy location)

Matt Levine had a funny similar take recently:

"You could have a model of Harvard Business School that is like:

    1. Harvard Business School teaches you skills that would make you good at running a company.
    2. There are lots of companies that could use those skills.
    3. But you don’t want to run those companies, because they make, like, ball bearings.
    4. You want to run a fancy company; you want to run a hedge fund or a tech startup or something.
    5. Meanwhile, the people currently running the ball bearings company would not be all that excited about you, a fresh-faced business school graduate who has never run anything, coming in to run their company, even if you did learn a lot of useful skills at Harvard.
    6. Therefore various industries exist whose principal business is laundering ball bearings companies into opportunities that appeal to Harvard Business School graduates. You wrap the ball bearings company in a name like “private equity” and suddenly it is legible to the Harvard students, so they flock to it.
    7. Those industries are also in the business of getting the ball bearings companies to accept the Harvard Business School graduates, which in practice means not so much “make the ball bearings company excited about its new Harvard CEO” but rather “buy the ball bearings company and install new management.”

Source: https://archive.is/8IUCA#selection-1795.0-1869.303

I like the pet theory!

just quibbling on profitability. it's not ostensibly 80%, it's 80%. gross margin != "net profit" != cash flow positive, thanks to GAAP.

Compare the margins (gross, operating, net) here [0]. Observe the historical changes in cash on hand (i.e. cash flow) here [1].

They have been accruing cash-on-hand on a YoY basis since 2021Q4.

80% gross margins on 2.5B TTM revenue is really impressive.

For comparisons, Cloudflare sits around 77% (on 1.5B TTM Revenue), Salesforce around 75% (36.5B TTM revenue), Datadog around 80% (2.4B TTM revenue).

It does remain to be seen on whether they can translate that into meaningful operating margin over time. But they're well on their way [1]

[0]: https://macrotrends.net/stocks/charts/PLTR/palantir-technolo...

[1]: https://macrotrends.net/stocks/charts/PLTR/palantir-technolo...

Grafana is a better approximation. I used to say back in the day that Peter Thiel complains about no flying cars but is making a data ingesting platform with a Chart Js frontend.

And yet at the same time I can totally see how it would feel like science fiction to a cop :

https://logicmag.io/commons/enter-the-dragnet/

(And a very different kind of science fiction for a non-cop.)