Revenue up 96% in the third quarter — and in the same results, a warning from CEO Alex Karp that executives stalling on AI are heading for oblivion.
Palantir Technologies on Monday reported its fastest growth as a public company, and its CEO paired the numbers with some of the bluntest language any large-company chief has aimed at corporate America’s AI holdouts.
Palantir said total revenue reached $2.31 billion in the quarter, a 96% year-over-year increase and a 20% sequential rise, according to Karp’s Q3 letter to shareholders. The engine was the company’s U.S. commercial business, which surged 148% to $984 million. For the first time in the company’s 23-year history, the public sector did not account for the majority of revenue: the government revenue grew 96% to $954 million, according to the letter.
The letter is where the earnings end and the prophecy begins. Karp wrote that “every industry in the world—from banks to insurers to hospitals and even once-storied law firms—will either be remade or set briskly aside,” and that outcomes “will increasingly be binary, with winners and losers, and nothing in between.”
For companies that fail to heed that warning, Karp offered two fates, both grim.
The lonely middle will be little more than a barren wasteland filled with those who were once legacy players who failed to heed the word and paid the ultimate price, he wrote, adding that the crowded bottom will be occupied by those who long ago lost their chance to build on what might once have been an insight or an edge, now marching inexorably toward irrelevance, their fate sealed, the envelope unopened.
The numbers behind the sermon
Karp’s argument rests on the quarter his company just posted. Palantir’s letter cited deal counts and retention figures that, on their face, are unusual for a company of its size — and the letter itself frames them as evidence that the shift Karp describes is already underway, not theoretical.
| Metric | Q3 figure | Comparison stated in the letter |
|---|---|---|
| Total revenue | $2.31 billion | Up 96% year over year; up 20% sequentially |
| U.S. commercial revenue | $984 million | Up 148% year over year |
| U.S. government revenue | $954 million | Up 96%; first quarter in 23 years without a public-sector majority |
| Net dollar retention | 165% | Up 800 basis points from the prior quarter |
Deal flow was, by the letter’s own account, dizzying: the company closed 212 deals of at least $1 million, 107 deals of at least $5 million and 64 deals of at least $10 million in the quarter. Net dollar retention — a measure of how much existing customers expand their spending — hit 165%, up 800 basis points from the prior quarter, which the letter cites as evidence that small pilots are turning into large commitments.
That last figure is the answer to Karp’s own rhetorical problem. A CEO warning that laggards will die is doing marketing, of course — and the letter makes no pretense otherwise, closing with the claim that “there is simply no means of effectively and reliably connecting the language models to large enterprises without the operating model and capabilities we have pioneered.”
What gives the warning weight is that customers are returning for more, faster: retention at 165% means the average existing customer grew its spend by nearly two-thirds over the trailing period.
The LLM “blast radius that tends towards infinity”
Karp does not argue that deploying AI is automatically good. His letter contends the opposite: shortcuts will fail, and the risk is not mediocrity but disaster. Enterprises have spent decades and hundreds of billions of dollars wiring together disparate hardware and software, and he argues that approach collapses when large language models meet a real corporate environment.
In the best case, Alex Karp wrote, these models on their own generate limited value, and in the far more likely outcome, they become agents of chaos, lacking a grounding in the means of interfacing with the labyrinth of datasets, people, business units, use cases, and critical safeguards scattered across the modern enterprise.
That is Karp’s pitch for Palantir’s ontology-based platform in a single image: an autonomous AI agent wandering an enterprise unmoored from its data and decision structures is, in his framing, a liability rather than a productivity tool. The letter positions Palantir’s operating model as the missing grounding layer — the connective tissue between language models and the “labyrinth” of enterprise data.
Karp also described a compression of the sales cycle itself. What may have taken months and years—a slow, coy, and theatrical dance towards a contract—can now take days and weeks, he wrote, adding that some leaders who bet they could “delay the future” have been stunned to find “they cannot get out of town.”
What the customers say the software actually does
Palantir’s own customer materials, published on the company’s site, supply concrete — if company-curated — examples of the demand Karp describes. Walgreens’ VP, Product Pharmacy said automating decisions across the pharmacy operation would otherwise require “about 384 billion decisions every day,” calling human-only orchestration “flat out impossible.”
General Mills’ Sr. Director Supply Chain Digital Transformation reported savings of “about $40,000 a day, which is about $14M annually,” and noted the tool is “really only deployed to part of our network.”
Other cited deployments span industries and sizes. Lowe’s said its AIP-powered system processes “over 100 times more leads with the same amount of resources,” cutting lead processing time by more than 90%. United Airlines reported a $9M annual cost reduction within three months of combining Foundry’s data integration with machine-learning models from Fujitsu.
Sompo’s Chief Technology Officer cited a $60 million profit improvement over three years, with an additional $100M expected. Panasonic Energy North America said its “Ask Atom” training tool cut a three-to-six-month technician learning curve “down to just a few weeks.”
The speed claims recur across nearly every testimonial. AARP’s SVP Digital Strategy & Membership said a prototype launched within 45 days; Lowe’s said it went from POC all the way to production in less than four months. One CIO quoted in the company materials was blunter than most: “we were looking for a partner and an approach to our data strategy, if you will excuse the term, that actually produced s***.”
Wake-up call, or sales pitch, or both
The honest reading of the letter is that it is both. Karp is selling — the letter says so in substance if not in those words, and Palantir’s growth validates the pitch better than any competitor’s counterargument could.
But the letter’s intent, read closely, is less to bury the hesitant than to accelerate them. Karp wrote that “a broad swath of leadership across American industry has awakened to the significance of this moment,” framing the urgency as advice rather than only advertisement.
The markets, institutions and skeptics all get the same message: the window Karp describes is measured in quarters, not years. “The revolution is happening so quickly,” he wrote, “these leaders have been stunned to find that they cannot get out of town.”
What comes next is measurable. Palantir’s next quarterly report will show whether the 165% retention rate and the government-minority revenue shift — a genuine structural change in a company long tied to federal contracts — hold or accelerate.
And the open question the letter itself leaves unresolved is whether “every industry” really remakes on Palantir’s timeline, or whether competitors and enterprises find a different path to whatever the binary outcome Karp describes turns out to be. As the letter puts it: “Whoever has ears ought to hear.”
