Palantir Soars 30% in a Day, Erasing $3 Billion in Short Sellers’ Paper Gains

Easy Tech, Read With Rhythm ⑤

Palantir’s comeback after many had written it off

An unfashionable strategy pays off

Palantir Technologies Inc. shares surged nearly 30% in a single session on Aug. 4, marking the company’s second-best day on record. The jump wiped out about $3 billion in paper gains that short sellers had built up betting on the stock’s decline. The catalyst was its second-quarter results. Revenue rose 93% from a year earlier to $1.94 billion, well ahead of market expectations.

Those figures suggest more than a company simply executing well. In the near term, they point to a clear shift in the enterprise artificial intelligence market. Over a longer horizon, they hint that the growth formula Silicon Valley software companies have followed for the past two decades may be starting to wobble.

What particularly stirred investors was not the revenue total itself, but its composition. Palantir’s U.S. commercial business, which serves private-sector clients rather than government agencies, expanded 149% from a year earlier. An even more important figure was total contract value, or TCV, the combined value of newly won contracts during the quarter. That jumped 153% to $2.132 billion. The metric matters because it reflects business already committed for the next several years, not just revenue recognized today. Net revenue retention, or NRR, which measures how much more existing customers spend when they renew, climbed to 157%. Put simply, a customer that spent 100 last year is spending 157 this year.

Building It Instead of Trading Proposals

How did Palantir pull this off? At the center is an unusual sales approach called the AIP, or Artificial Intelligence Platform, Bootcamp. Enterprise software is usually sold through months of back-and-forth proposals, demo videos and contract reviews. Palantir largely skips that process. It brings in a client’s staff and spends five days building a working AI system on the spot using the client’s own data. The company has held more than 1,300 bootcamps so far, and about 75% of participating companies are known to have gone on to sign contracts. That effectively cuts a sales cycle that once took 12 to 18 months to about a week. Rather than explaining the product over and over, Palantir shows customers that it works.

Even winning contracts quickly is not enough if clients do not trust AI with their data. That is where Palantir’s Ontology technology comes in. It organizes data scattered across a company — such as sales ledgers, personnel records and inventory data — into a structured map that AI can understand. Large U.S. companies have grown increasingly concerned that handing sensitive internal data to outside large AI models could eventually turn that information into training material that rivals might also benefit from. Palantir argues that its Ontology, built over the past 20 years, allows companies to use AI without sending their data outside the organization. Chief Executive Officer Alex Karp’s repeated emphasis on “sovereign AI” on earnings calls comes down to that same point. The market’s reaction suggests investors accepted Palantir’s argument that safe AI at scale is hard to achieve without such a layer.

Palantir’s Front-Line Secret Weapon

Who actually builds all this? Palantir’s Forward Deployed Engineers, or FDEs. These engineers work inside client sites alongside customers. Instead of shipping finished software like a package, they effectively move into the customer’s environment and build what is needed for that specific operation.

The software industry has long been skeptical of that model. If one engineer must stay attached to one client, adding 10 customers means hiring 10 more engineers. That is why software-as-a-service, or SaaS, became the dominant formula: build once and let customers install and use it themselves.

Among South Korean engineers, some still question whether FDEs are much different from on-site engineers dispatched by system integration firms. Palantir, however, counts FDE labor costs as customer acquisition cost, or CAC. By sending engineers in before a contract is even signed and producing results first, the company can embed itself deep within a client’s workflow once it gains a foothold. That makes it hard to dislodge later. One widely used metric, the Rule of 40, combines growth and profitability. A score above 40 is generally considered strong. Palantir posted 155 this quarter.

Big Tech Expands FDE-Style Roles

Big Tech companies are also moving aggressively to expand FDE-style roles. OpenAI has committed $4 billion and set up a separate company dedicated to embedding its AI into other businesses. Google Cloud’s chief executive officer posted directly on LinkedIn to recruit for the effort, opening 59 positions at once. Anthropic has teamed up with Blackstone Inc. and Goldman Sachs Group Inc. to create a $1.5 billion joint venture that places its engineers inside financial-sector clients. ServiceNow Inc. and Accenture Plc have already launched similar programs that send engineers to work alongside customers.

Heo Jin, Hankyung.com reporter, hjin@hankyung.com