Blog Post

Palantir: The worst-kept $9B secret in Silicon Valley

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Businessweek is reporting that “secretive” data-analysis company Palantir is raising another $100-million-plus round of funding (which would put it above $700 million total) at a $9 billion valuation. This comes just several weeks after the company raised $196 million in capital at what Businessweek claims was a $6 billion valuation. That’s a lot of money, but Palantir’s business — and real estate investments — require it.

I say “secretive,” because although there is an aura of mystery around the company, its customer base — most notably a goodly number of intelligence, law enforcement and banking institutions — is probably among its biggest secrets. CEO Alex Karp was the subject of an in-depth profile in Forbes earlier this year, and has been interviewed on television by Charlie Rose. The company is an omnipresent figure in Palo Alto, too, buying up building after building downtown, some staffed with their own roving security guards.

Palantir-hoodied employees, as someone noted in my last trip to the city, can dominate popular local establishments such as Philz Coffee, and the housing subsidies they receive are helping drive up rent prices. I envisioned a West Side Story sort of environment between Palantir and Facebook before the latter left town, with each company’s employees each decked out in their corporate logos. There were no switchblades involved, but I have heard of awkward staredowns in crosswalks.

The company isn’t too shy about showing off how its technology works, either — and that’s where all the money comes into play. I saw a talk by Palantir engineer Ari Gesher in August where he showed off how the product works and how it’s built. This isn’t analytics like running a Python script over some web logs; it’s analytics involving a complex data architecture that includes numerous connected Oracle, Hadoop and NoSQL data stores. It’s a high-touch, complicated process that the Forbes article says costs about $1 million per installation (I’m guessing that doesn’t include support). Palantir isn’t yet profitable, but Forbes estimates it will do about $450 million in revenue in 2013.

The slide below, from a talk Gesher gave in July, illustrates how a typical Palantir installation looks. He’ll also be speaking at our Structure Data conference in March, so attendees can hear a lot more about the company’s technology and vision there.


However, a high price tag is worth it if the Palantir platform works, and by all accounts it does. In a startup world where so much effort is placed on automating pattern detection and building network graphs, Palantir’s software is markedly different. It’s not about machine learning as much as it’s about taking lots of data from lots of places and making analyzable by the people who know the problem spaces best. It has famously been used to thwart terrorism plots and solve crimes, and is also used for everything from fraud detection in financial services to disaster relief.

A demonstration of Palantir's tools for connecting entities and actions.
A demonstration of Palantir’s tools for connecting entities and actions.
Using Palantir to track flooding during Superstorm Sandy.
Using Palantir to map flooding and relief efforts during Superstorm Sandy.

So, yes, it appears Palantir is awash in capital and investors are valuing it at 20 times its revenue. It might serve as an inspiration to the hordes of other startups trying to capitalize on industry’s desire for all things big data, but often focusing on horizontally applicable areas like BI or general-purpose machine learning. Find a valuable application that companies will gladly pay for, build something that works — well — and customers’ and investors’ money might start pouring in.

6 Responses to “Palantir: The worst-kept $9B secret in Silicon Valley”

    • David Anderson

      This is a real issue. $450MM revenue, and how much of that came from taxpayers, for what public benefit? There are a lot of people who wouldn’t be happy with honest answers here.

  1. A couple notes.

    * UI is terrible, they better find somebody to redesign interface
    * Distributed DB on top of HDFS is either HBase or Accumulo or something home-brewed. If its HBase – its good, otherwise – it is not.
    * $450M revenue and still unprofitable? Do they build next generation nuclear reactor?

    • Sebastian Stadil

      Enterprise sales is expensive, and government is even worse. I suspect a lot of that revenue is paid back to channel partners in some form.

      Thanks for the notes, insightful.

    • Palantir is in good company. Look at AMZN revenue, yet it STILL isn’t profitable! Valuation at 20 times revenue is not so high, compared to S.V. neighbors that are also tech start-up’s.

      I think a lot of Palantir’s work is done internally. Trade-off between security versus portability, ease of use. If they had good, experienced people doing the work, it would be less problematic. Idle thought: I wonder how stable their code is…

      Palantir is oddly diffuse. They do defense contracting, fancy business intelligence with lots of data but no ML, disaster response (nice map, lots of jaggies!), fraud detection and a financial services (trading?) platform. I am really curious, and puzzled, about the latter. It is expensive; what is the target market? Martin Marietta and Lockheed were into many related businesses, but Palantir is spread thin, given their size (small) and inexperience.