Palantir Pricelist (page 27) [pdf] (opens in new tab)

I am not quite sure if even this community quite understands what Palantir is and does. Things aren't quite what they appear to be.

I should have also added to my original post, this is their GSA schedule. For these specific line-items, the government considers this the "prenegotiated lowest price". This simplifies purchasing. So if somebody in the government wants buy another core (item 132-33) they just ring up Palantir or their GSA schedule VAR (reseller) and ask for a 132-33 volume:1 and they know that it'll cost $141k and be good for 1 year of O&M.

The downside of being on the GSA schedule is that it means that there are limits on your ability to change prices year to year to reflect a change in business environment. Say next year you want to drop the price by 50%. That's not allowed, since it would basically mean that you weren't selling your software at the lowest possible prenegotiated price. Similarly if you want to increase it, you're limited in that as well. I think the +- variance can't exceed 10% in a single 12 month period, but I might be wrong. The other downside is that this basically provides your prices in public so your competitors can see what you charge, which can be critical information in contract bids.

There's always a loophole. For example, convince your customer to not buy off of the GSA and sell them a "package" at some discount off of GSA. That "package" counts as a different SKU and the GSA doesn't apply. So for example, you could make a Palantir "starter kit" composed of

   2 core license (normally $282k)
   1 extra year of O&M maintenance for both cores (normally $56k)
   500 hours of Ninja Services @ $150/hr (normally $98k)
   500 hours of CONUS FSR Service @ $120/hr (normally $73k)

Any government purchaser would fall over themselves to try to make this happen since it's a >30% discount off of the normal GSA schedule prices (which they can refer to and are governed by various rules). This looks fantastic on their yearly performance eval "-saved government $160k on purchase of Palantir package deal"

This is the crack model. When the O&M runs out in two years and they come asking for more, well, there's no deal anymore, and the GSA schedule prices on O&M and people have gone up 20% in that time.

edit keep in mind that this company has raised almost a billion dollars and is valued at something like $9 billion dollars.

They've done a fantastic job at disrupting the link-chart market (which is surprisingly robust) and making it seem like customers are buying something else, but at 2-3x the price of the competition. They hide their sales guys as "forward deployed engineers" and obfuscate their sales process to the point where it doesn't seem like you've ever dealt with the kind of sketchy enterprise sales goon you'd deal with from any of their competitors.

They're very smart.

That system definitely has problems (it's terrible and sucks in many places), but it's a system of much larger capability and complexity than Palantir.

Comparing Palantir to it is like comparing a wheel to a an entire transport system. Palantir could comfortably fit in as a single capabilty in DCGS-A (and probably do a better job than the stuff currently filling in that role), but it could never replace it.

It's basically just a federated search, mediocre map and halfway decent link-analysis tool. But ask it to do anything even remotely outside of those three things and you're basically dead in the water as it doesn't offer anything relevant to all the other millions of things DCGS-A provides. DCGS-A is more than just some analytic tools.

The reason it's not part of DCGS-A is very political and complex and as much Palantir's doing as the Army's, but the answer is that the components that are on DCGS-A that basically do what Palantir does were selected because they're cheaper over the long run, even if a bit clunky. For example, on the client side, there's presently a forward looking mandate from big Army to drop flash and java clients (because they're an IT administrative menace). Palantir's front end is Java.

i2 was selling their software at some fraction of Palantir's and in the DoD space it's basically as ubiquitous as Microsoft Office. Palantir is everywhere, but it doesn't end up being used nearly as much.

My company spends tens of millions per year on some really awful enterprise software unfortunately.

This is actually one of the reasons that doing business with the government is less lucrative than the commercial industry (for products and services which are comparable, many are things only sold to the government, so there is no hesitation about pricing them through the roof).

As you can imagine, one of the ways that duplicitous federal contractors get around the idea of having to sell to the government for less than their normal prices is to structure their products/services as different (and therefore not apples to apples comparable).

I still would bet that the price that they charge Goldman Sachs (just as an example, I don't actually know which investment banks are using their product) is higher than what they charge for any individual government client.

Also, I think it's probably worth noting that the maximum addressable customer base for their products for federal agencies is way smaller than the equivalent number of financial customers. There's really only a handful of agencies that have this capability (I'd guess more than a handful, but probably less than a dozen).

Anyway, from their Gotham page https://www.palantir.com/palantir-gotham/

"Working closely with the customer, our engineers integrate and map all of the relevant source data—regardless of type or volume—into a single, coherent model.

...

Once the model has been created, data flows continuously from its sources into the Palantir Gotham platform.

...

They can search across all of their data sources at once, visualize relationships, explore divergent hypotheses, discover unknown connections, "

Pretty much my 2009 pitch, and in the hindsight i see that my main weakness was that i couldn't even imagine $150K/core. Man, it is imagination what separates losers from winners! :)

Unify search, build semantic links.

It was a very manually intensive tool based on what I saw. Very little automation, more like those pictures and strings you see in crime shows on TV...but on a computer.