Peter Thiel on How Palantir Went From Serving the CIA to Building a $15 Billion Business

We talked with the "godfather of venture capital" about his second, less-noticed tech startup.

PayPal founder Peter Thiel is probably the most popular Silicon Valley entrepreneur in China after Elon Musk. Here, his book Zero to One has made Thiel's name almost synonymous with "starting a company."

After selling PayPal to eBay in 2002, Thiel, now a billionaire, became an active tech investor. He was among the first to bet on Mark Zuckerberg, investing in Facebook, whose IPO remains the largest ever for a tech company. He went on to make a string of other successful investments, organized talented young college students to drop out and start companies, and in March this year became a partner at the Silicon Valley incubator Y Combinator.

But in Thiel's own view, the tech company most likely to become the next Facebook or Google is a far less familiar one: Palantir.

Palantir is the second tech company Peter Thiel co-founded. He is its co-founder, largest shareholder and chairman, and in this role he keeps as low a profile as the company itself.

In its first few years, Palantir behaved in ways that made it almost impossible to guess it could become a valuable business. But after its latest funding round closed at the end of 2014, Palantir was valued at US$15 billion, ten times what PayPal sold for.

Palantir's air of mystery also comes from its investors and customers. In 2004, its first outside money came from In-Q-Tel, a non-profit venture capital firm created by the Central Intelligence Agency, which funds startups as a way to get the new technology the agency needs.

Its early customers were all U.S. intelligence and security agencies: the CIA, the Defense Intelligence Agency (DIA), the FBI, the military and big-city police departments. Nearly every agency you have seen in a counterterrorism film was a customer.

How did this business begin?

After selling PayPal in 2002, Peter Thiel, flush with money, felt lost for a time. Soon after, in 2003, two years after 9/11, the United States was fighting wars in Afghanistan and Iraq at the same time. To stop terrorists, the Transportation Security Administration installed large numbers of new airport screening machines. Invasive scans and pat-downs led critics to call U.S. airport security a "nightmare."

Startups often begin with spotting a problem and looking for a way to solve it. Terrorism was the problem Thiel had in his sights.

Thiel turned to a tool PayPal had used. He put a question to Alex Karp, his old classmate from Stanford: could the methods PayPal used to detect online fraud be used to uncover terrorist networks?

During his visit to Beijing last week, Thiel recalled in an interview with QDaily: "Fraud was a major technical challenge for us at PayPal. We ran into it all the time, and it was hard to solve. We had millions of 'good' users and thousands of 'bad' ones, and we had to work out which transactions were fraudulent, while there was simply too much information to handle. We needed to go through all of it and find the patterns."

PayPal's answer was to hand part of the problem to computers and leave the rest to people.

"No computer was smart enough to point out every problem, and there was far too much information for human reviewers to process. That is how we solved the problem at PayPal," Thiel said. "People and machines working together do far better than either one alone."

This became the methodological foundation for how Palantir designs its products. The name comes from Tolkien's fantasy epic The Lord of the Rings. In the books, the elves created the "palantíri," seeing-stones through which wizards could watch events in distant places. Like the stones in the novel, Palantir gathers every database it can, sorts through the information and draws conclusions, helping customers achieve "true sight."

A palantír in the film The Lord of the Rings

In the summer of 2004, Thiel, Karp and three other co-founders launched Palantir. Thiel mainly helped recruit the team and put up seed money. Once the company was running normally, he handed day-to-day management to Karp and advised the company as chairman.

Counterterrorism was the way in, but from the start Palantir wanted to be a company that analyzes all kinds of data and serves businesses, banks and other institutions.

That choice was not easy. In the first few years, the idea was too complicated and commercial investors showed no interest. Then an investor who had turned Palantir down suggested it approach In-Q-Tel, the CIA's non-profit investment arm, which brought Palantir its first outside investment and its first customer.

"The standard path in Silicon Valley is to sell your product to companies first. Nobody thinks of selling to the government first, because the government is too inefficient and too slow. But we felt that Palantir was solving a more precisely defined problem. It's quite similar to the path Musk took at SpaceX, which also sold its rockets first to the U.S. government (NASA)," Thiel said, explaining why Palantir went to intelligence agencies first.

For its first four years, intelligence agencies were Palantir's only customers.

How Palantir met the needs of counterterrorism

Soon after 9/11, retired Navy Rear Admiral John Poindexter was put in charge of launching the Total Information Awareness (TIA) program. It aimed to pour everything the intelligence agencies collected into one database, then use computers to find the clues hidden in vast amounts of seemingly unrelated information.

In an age of information overload, terrorists inevitably cross paths in different settings and leave traces. What U.S. intelligence agencies needed at the time was a product that could quickly connect one piece of information to another.

The ambitious TIA program was shut down in 2003 over privacy and other concerns, but the need behind it never went away. Palantir filled the gap.

In 2005, Karp showed intelligence officials a prototype for the first time. It worked on essentially the same principle as PayPal's tool, except that the targets changed from online fraudsters to two hypothetical terrorists. With In-Q-Tel's help, Palantir's engineers met intelligence analysts face to face, and from then on, every two weeks, the engineers returned to Washington with a revised version of the product. Starting from scratch, they built a complex search tool that drew conclusions from databases that had belonged to separate intelligence agencies. In most cases, the conclusions might still have been written on paper in pen.

According to The Wall Street Journal, in Palantir's early days, analysts at West Point tried using its software to map Syrian suicide-attack networks from 700 captured documents, including hundreds of personal records the military had recovered in Iraq.

One analyst went through the documents on his own and wrote a report noting that foreign insurgents in Iraq were hiring Syrians to carry out suicide bombings. A second analyst, using Palantir's software, built on this and produced more detail about the Syrian terrorist network, including information on contacts at the top. That analyst concluded that more than one Syrian suicide-attack network was operating in Iraq at the time.

Detailed reports like this led U.S. intelligence agencies to discard many old misconceptions. For example, insurgent groups behind suicide attacks in Iraq relied more on mercenary smuggling networks than on fanatical believers.

Palantir in 2008: unlike the slick spy tools in the movies, it didn't look like much

Today, every U.S. intelligence agency uses Palantir's technology, and they have settled into a cycle: analyze intelligence, strike targets, analyze the new intelligence gathered in the strikes together with what they already have, and launch the next round of strikes based on the results.

The 2011 killing of Osama bin Laden brought U.S. special operations in Afghanistan and Pakistan into public view. But before that, U.S. special forces had already carried out thousands of similar raids in Iraq, Afghanistan and Pakistan. A few units, such as the SEALs who took part in the 2011 raid, went out on operations nearly every day.

As shown in the film Zero Dark Thirty, the special forces carried away not just a body but also bin Laden's computers, documents and discs. None of this intelligence was analyzed in isolation. It went into an ever-larger database, and every analysis drew on the entire pool, whereas human effort alone could have covered only a small fraction of it.

To better support intelligence work in this new era, the U.S. military set up a large data center at Camp Victory, the base surrounding Baghdad airport. Software engineers were also sent to train the special forces taking part in raids, helping them gather information more effectively during operations.

In 2008, the U.S. government began signing formal contracts with Palantir. Between 2007 and 2009, Palantir's business in Washington, D.C. grew quickly from 8 pilot projects to more than 50 formal projects.

Something few have tried: selling to the government first

By 2008, four years after its founding, U.S. intelligence agencies were still Palantir's only customers.

But their business eventually became Palantir's ticket into other markets. Thiel said: "Selling to the government takes more time, and the whole sales process is long, but once you break in and win them over, you have very valuable customers. If the CIA uses our product, big banks may be just as willing to work with us. In the United States, intelligence agencies are very important as reference customers."

In 2010, JPMorgan Chase became one of Palantir's first non-government customers. It used Palantir's technology to catch fraudsters trying to steal customers' accounts. It also used Palantir's products to price distressed properties, drawing on local economic information to avoid causing social disruption in places where it held loans on such properties.

Bank of America, the hedge fund Bridgewater Associates and even the U.S. Securities and Exchange Commission also use its technology, while Hershey uses Palantir's tools to make its chocolate more profitable.

The technology behind these tools is complex, but the principle is simple, and you can think of it the way you think of "big data": it builds massive datasets, holding everything from local rainfall to bank transfer records, and draws conclusions from non-obvious links in the data to solve each customer's problems.

As the business grew, so did the range of problems Palantir took on. Today, government and finance are Palantir's two main product lines. For a company with no sales team at all, the endorsement that comes with government contracts is a great advantage. By its funding round at the end of 2013, 60% of its orders already came from non-government customers.

Without media attention, Palantir quietly grew into a giant

Palantir grew almost entirely without press coverage.

In 2010, before Evelyn Rusli, a columnist at the tech blog TechCrunch, took notice of Palantir, the site, known for closely tracking tech startups and investment news, had run only one article about the company. It had run at least 208 about Foursquare, which was founded in 2009.

Karp later said sardonically in an interview: "Between 2005 and 2009, talking about enterprise software in Silicon Valley was sometimes like putting on a circus show for engineers in Palo Alto."

Indeed, an enterprise software company with 70% of its orders coming from Washington had little chance of catching the eye of investors or the tech press.

Today, however, money is no longer a problem for Palantir.

• In July 2010, Palantir, which by then had 250 engineers, closed a US$90 million Series D round at a valuation of US$735 million.
• In December 2013, Palantir raised a new round at a valuation of US$9 billion. By then its annual revenue had passed US$450 million.
• In November 2014, Palantir raised another US$500 million, bringing its valuation to US$15 billion.

As its valuation soared, Palantir also began buying companies to expand its team and the fields it covers. In February 2013, it acquired the voicemail service Voicegem. In July of the following year, within a single month, it bought Poptip, a social testing service, and Propeller, a mobile app incubator. In February this year, Palantir acquired Fancy That, a startup offering an omnichannel marketing platform.

Palantir has not responded to speculation that it is moving into consumer markets. But it is not hard to imagine that, after payments, finance, insurance and healthcare, this data analytics company will soon expand into retail.

Palantir's future: the world's largest consulting firm?

This is a company under no financial pressure. It has raised close to US$1 billion in total, and expects revenue of at least US$1 billion this year.

At a time when most companies dream of going public, Palantir is in no hurry. Last March, CEO Karp told CNET that, at least for now, the company had no plans to go public any time soon.

Talking with QDaily about Palantir's near-term plans, Peter Thiel said that although Palantir has gained "tremendous traction," its product is still at a very early stage and many problems remain to be solved. "Even though it has brought together a great many talented people to work on important problems that hardly anyone else is working on, there are still many areas where they can do better," he said.

As it sells to more customers, Palantir has begun giving each of them the customization they need, much like those biweekly trips between Washington and Silicon Valley in the early days. Asked about the plan for the next five years, Thiel said Palantir's challenge is "how to effectively integrate the product side with the consulting side."

"Our plan is to scale up from where we are now. It is a complex sales cycle," Thiel said. "Those big orders worth between US$20 million and US$200 million make greater demands on the product."

Cover image from Fortune Live Media