Research from the University of Michigan, Stanford, Google, and Microsoft sheds light on the underlying reasons why token usage can be difficult to control. Their study found that total token usage can vary by up to 30X on the same task, and that higher token usage does not translate into higher accuracy.\u00a0This highlights the\u00a0difficulty\u00a0of estimating token costs going into a\u00a0project,\u00a0and\u00a0thus understanding the return that\u00a0could\u00a0be generated.\u00a0<\/p>\n
Lastly, research from OpenAI itself shows that higher token consumption does not necessarily translate into higher customer revenue. An OpenAI-led study, which included researchers from Columbia Business School and Wharton, analyzed\u00a0nearly 400\u00a0public U.S. companies in\u00a02024 and\u00a02025. Notably, the researchers found that \u201crevenue per employee is not meaningfully associated with output tokens per employee or messages per active user.\u201d\u00a0\u00a0<\/p>\n
These pieces of evidence highlight several flaws in the\u00a0token-consumption\u00a0model, and\u00a0lend\u00a0credence to\u00a0Palantir\u2019s view that consuming more tokens\u00a0doesn\u2019t\u00a0inherently lead to better business outcomes.\u00a0<\/p>\n
Rising token consumption alone does not tell investors which AI stocks are generating long-lasting value\u00a0\u2013\u00a0but the I\/O Fund\u2019s institutional-grade\u00a0research\u00a0does.\u00a0We publish hundreds of articles per year on the AI trade, hold weekly\u00a0webinars,\u00a0and offer real-time trade alerts.\u00a0\u00a0Subscribe now<\/a>\u00a0to follow the\u00a0I\/O Fund, with\u00a0six stocks\u00a0already up\u00a0more than 100% YTD.\u00a0<\/p>\n Palantir\u2019s Land-and-Expand Model Is Driving Higher-Quality Growth<\/p>\n We can see how these different business models result in\u00a0very\u00a0different\u00a0financial profiles between Palantir and AI labs.\u00a0<\/p>\n Recent reports suggest that OpenAI recently crossed\u00a0$40 billion\u00a0in ARR<\/a>, doubling in approximately seven months. Meanwhile,\u00a0Anthropic\u2019s\u00a0ARR has\u00a0reportedly hit\u00a0$65 billion<\/a>\u00a0in July, increasing by more than 7X in seven months. Assuming these figures are calculated by multiplying monthly revenue by 12, they would imply monthly revenues of\u00a0$3.33 billion\u00a0for OpenAI and\u00a0$5.20 billion\u00a0for Anthropic.\u00a0\u00a0<\/p>\n Particularly in the case of Anthropic, it\u00a0is fully\u00a0possible that no other tech company in history has grown this quickly off of\u00a0what was already a very sizeable revenue base to start the year.\u00a0<\/p>\n Palantir\u2019s revenue\u00a0is growing at\u00a0a\u00a0rapid pace\u00a0as\u00a0well, with sales rising\u00a093% YoY last quarter to\u00a0$1.94 billion.\u00a0However,\u00a0OpenAI and Anthropic are still growing dramatically faster off\u00a0larger revenue bases.\u00a0<\/p>\n Still, the differences in their business models\u00a0can help\u00a0explain\u00a0some\u00a0of the\u00a0gap. When a new enterprise begins using their\u00a0AI, the consumption-based model means that revenue can begin to\u00a0accrue\u00a0almost instantly. This makes the AI lab business model extremely scalable, as token consumption\u00a0often\u00a0simply needs to continue\u00a0to\u00a0rise\u00a0for\u00a0revenue to grow.\u00a0<\/p>\n Contrast this with Palantir, which has a much slower-moving monetization model. Notably, the company deploys forward\u00a0deployed\u00a0engineers (FDEs) into\u00a0organizations\u00a0it works with.\u00a0These FDEs spend\u00a0a significant portion\u00a0of their time embedded on-site with customers and build customized solutions.\u00a0<\/p>\n This should allow FDEs to create solutions that actually generate value, build trust, and then\u00a0expand the relationship into more workflows.\u00a0Notably, this \u201cland-and-expand\u201d strategy is driving the vast majority of Palantir\u2019s commercial sales growth.\u00a0<\/p>\n Palantir\u2019s revenue from commercial customers increased by $495 million, or 110% YoY in Q2. Of the increase, $407 million was from commercial customers existing as of\u00a0the end of 2025.\u00a0In turn, 82% of Palantir\u2019s commercial revenue growth came\u00a0from\u00a0existing\u00a0customers;\u00a0a clear representation of its\u00a0land-and-expand success,\u00a0and\u00a0an indicator of real value creation.\u00a0\u00a0<\/p>\n Palantir\u2019s net dollar retention rate\u00a0of\u00a0157% last quarter\u00a0also\u00a0demonstrates\u00a0this,\u00a0with\u00a0existing customers increasing\u00a0their spend\u00a0by 57% from the prior year.\u00a0Furthermore,\u00a0Palantir\u2019s U.S. commercial remaining deal value (RDV),\u00a0surged 124% YoY to\u00a0$6.24 billion, showing that the company\u00a0is rapidly growing its contracted revenue.\u00a0<\/p>\n Meanwhile, OpenAI and\u00a0Anthropic\u2019s\u00a0ARR figures\u00a0provide little clarity into the quality of their sales growth. They\u00a0do not\u00a0show how\u00a0much\u00a0growth is coming from new customers experimenting\u00a0versus\u00a0existing customers\u00a0spending more, or what percentage of sales are contractually obligated.\u00a0\u00a0<\/p>\n Amid this,\u00a0it is notable that\u00a0labs are actively\u00a0working to replicate<\/a>\u00a0the\u00a0FDE model,\u00a0signaling the merits of Palantir\u2019s\u00a0approach.\u00a0Ode with\u00a0Anthropic\u00a0and OpenAI\u2019s Deployment Company are joint ventures with asset\u00a0management\u00a0companies\u00a0staffed with FDEs\u00a0whose\u00a0purpose\u00a0is\u00a0implementing\u00a0their\u00a0technology within enterprises.\u00a0<\/p>\n On\u00a0$1.94 billion\u00a0in revenue last quarter, Palantir generated $912 million in GAAP operating income and\u00a0$1.216 billion\u00a0in operating cash flow (OCF). Among all tech stocks in the\u00a0S&P 500 Index,\u00a0Palantir\u2019s operating margin ranked in the top 10 at 47%, and its OCF margin was the second highest at 63%.\u00a0<\/p>\n At the same time, AI labs are struggling to even\u00a0turn\u00a0a profit despite having significantly higher revenues. OpenAI\u00a0reportedly generated\u00a0$6.7 billion\u00a0in revenue<\/a>\u00a0during Q2, yet its operating losses widened from\u00a0$9.3 billion\u00a0to\u00a0$12.3 billion, implying a margin of -184%.\u00a0<\/p>\n Meanwhile, Anthropic\u00a0reportedly expected\u00a0to generate\u00a0$10.9 billion\u00a0in revenue<\/a>\u00a0during Q2, and $559 million in adjusted operating profit,\u00a0for a margin of 5.1%.\u00a0\u00a0<\/p>\n While\u00a0Palantir\u2019s\u00a0high-quality\u00a0growth metrics\u00a0indicate\u00a0that\u00a0its\u00a0value-based\u00a0model is\u00a0resonating with customers,\u00a0its expectational profitability profile suggests\u00a0pricing is still strong enough to provide a significant margin.\u00a0<\/p>\n However, there is one\u00a0key\u00a0caveat\u00a0to Palantir\u2019s\u00a0exceptional\u00a0profitability. For more than a decade, the U.S. government\u00a0funded\u00a0the company\u2019s\u00a0R&D,\u00a0helping support\u00a0Palantir\u00a0as it built the products\u00a0it\u00a0has today.\u00a0Palantir\u2019s accumulated deficit still sits at\u00a0$1.63 billion, showing the\u00a0remnants\u00a0of\u00a0its\u00a0unprofitable past. Meanwhile,\u00a0R&D spending has dropped to just 10% of revenue.\u00a0On\u00a0the other hand, OpenAI and Anthropic are still in the middle of\u00a0a\u00a0longer-cycle\u00a0R&D\u00a0investment period.\u00a0\u00a0<\/p>\n Thus, Palantir\u2019s\u00a0margins\u00a0are superior\u00a0today\u00a0partially because the money\u00a0required\u00a0to build its core products has already been spent, while labs\u00a0are still\u00a0building heavily.\u00a0Amid this, it remains to be seen when labs will achieve\u00a0similar profitability, or\u00a0if the opposite occurs,\u00a0as Palantir believes,\u00a0which is that\u00a0general-purpose\u00a0models become commoditized\u00a0and pressure\u00a0lab\u00a0margins long-term.\u00a0<\/p>\n Conclusion<\/p>\n Palantir holds an\u00a0enviable\u00a0position as one of the world\u2019s\u00a0highest margin tech companies, while continuing to grow at an extraordinary clip.\u00a0\u00a0<\/p>\n That position is\u00a0supported by the company\u2019s focus on generating value\u00a0for customers, rather than pushing raw token\u00a0consumption.\u00a0This value\u00a0creation is showing up in Palantir\u2019s financials, with existing customers\u00a0rapidly\u00a0allocating\u00a0more dollars to its platform.\u00a0\u00a0<\/p>\n As it stands today,\u00a0Palantir is the\u00a0only name among these firms\u00a0converting massive AI demand into\u00a0a high-margin\u00a0profit engine.\u00a0That\u2019s\u00a0not to say Palantir\u00a0doesn\u2019t\u00a0have a material headstart, having launched over twenty years ago in 2003.\u00a0\u00a0<\/p>\n Perhaps the\u00a0more important takeaway, then, is not that Palantir has discovered the exact business model others should follow, but that it is\u00a0demonstrating\u00a0just how profitable AI can become over time.\u00a0<\/p>\n Identifying\u00a0which companies are best positioned in the AI trade requires separating noise from actionable insights\u2014and this is exactly what I\/O Fund helps investors do. Subscribe to access the I\/O Fund\u2019s\u00a0Top 20 Stocks report:\u00a0Unlock Full Access<\/a>.\u00a0<\/p>\n![\"Bar](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2026\/09\/Palantir-High-Quality-Growth-Metrics.webp\")
This bar chart highlights three indicators of Palantir\u2019s growth quality and customer retention. The chart shows that 82% of commercial revenue growth came from customers acquired before 2026, demonstrating strong expansion within the existing customer base. Net dollar retention reached 157%, indicating customers increased spending by 57% year over year. U.S. commercial remaining deal value (RDV) grew 124%, reflecting significant growth in the value of signed contracts. Source: I\/O Fund, PalantirPalantir\u2019s Profitability Stands in Sharp Contrast to AI Labs<\/p>\n![\"Chart](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2026\/09\/Operating-Margin-Comparison-OpenAI-Anthropic-Palantir.webp\")
This bar chart compares the operating margins of OpenAI, Anthropic, and Palantir. OpenAI reported an operating margin of -183%, indicating losses substantially exceeding revenue. Anthropic\u2019s projected operating margin is 5.1%, reflecting modest profitability. Palantir reported a 47% operating margin, significantly outperforming both companies. Source: Wall Street Journal, Palantir<\/p>\n