Source: https://reinouttebrake.com/stories/ai-is-growing-fast-who-will-pay-for-it - [Home](https://reinouttebrake.com/) - [AI](https://reinouttebrake.com/topic/ai) - AI Is Growing Fast. Who Will Pay for It? # AI Is Growing Fast. Who Will Pay for It? Anthropic, Oracle and Palantir reveal where AI makes money, who funds its expansion and why the gap matters for businesses, workers and customers. 3 Nov 2026 6 min read By [Reinout te Brake](https://reinouttebrake.com/about) Updated 3 Nov 2026 ![Rows of server racks at a Macquarie Telecom data centre in Sydney](https://reinouttebrake.com/assets/editorial/servers-1280.webp) Server racks at a Macquarie Telecom data centre in Sydney. Archival illustrative photograph, not an Anthropic, Oracle or Palantir facility. Photo: Lgate74 / Wikimedia Commons, CC BY 3.0. [Original source](https://commons.wikimedia.org/wiki/File:Dcentre_racks.jpg) · [CC BY 3.0](https://creativecommons.org/licenses/by/3.0) · [Source & licence](https://reinouttebrake.com/data#image-credits) Oracle is raising $20 billion while Palantir reports a quarterly operating margin of 47%. Both are expanding in AI. Their numbers describe very different experiences of the same boom. Alongside them, Claude developer Anthropic has disclosed enormous computing commitments as demand for its models grows. Read separately, these announcements look like financial news for investors. Read together, they raise a question for anyone using AI at work: can the value customers receive support the cost of supplying it? The answer will influence which providers survive, what businesses pay and how much control customers retain over increasingly essential tools. ## Three Companies, Three Different Businesses Anthropic develops models. Oracle supplies cloud infrastructure alongside its established software businesses. Palantir connects data, software and AI to operational decisions inside organisations. These are different businesses with different investment needs. Palantir’s margins cannot establish that training frontier models is profitable. Oracle’s financing requirements cannot establish that applying AI inside a business is uneconomic. Nor do these announcements describe a single transaction chain linking all three companies. The connection is economic: [demand for useful applications](https://reinouttebrake.com/stories/the-biggest-change-in-ai-is-happening-behind-your-chat-window) supports spending on models and computing capacity. The risk is that spending across those layers grows faster than customers’ willingness to pay. ## Anthropic: Growing Demand Meets Long Commitments Anthropic’s [November 2 registration filing](https://www.sec.gov/Archives/edgar/data/2130357/000162828026068676/anthropicpbc-sx1ax1.htm?utm_source=chatgpt.com) reports approximately $17 billion in third-quarter revenue, compared with $1.4 billion a year earlier. Its operating loss was $567 million. The filing also discloses approximately $895.4 billion in purchase commitments and lease agreements, substantially related to computing and infrastructure. About $69.8 billion falls due in the twelve months following September 30, 2026. Much of the remainder extends well beyond 2030. Comparing the entire commitment with one quarter’s revenue would be misleading. These obligations cover many years, and contractual conditions matter. Nevertheless, the mismatch in timing matters. Anthropic is committing to capacity before knowing precisely how demand, model prices and competitive alternatives will develop throughout those years. My reading is that strong adoption has reduced uncertainty about whether customers want these tools. It leaves substantial uncertainty about the margins available as the market matures. ## Oracle: Profitable Does Not Mean Self-Funding Oracle exposes a different part of the problem. According to [Reuters’ reporting on the financing](https://www.schwab.wallst.com/schwab/Prospect/research/stocks/news.asp?doc=3200-nL3N46J11L-10&symbol=MSFT&utm_source=chatgpt.com), it is raising approximately $10 billion through shares and another $10 billion through debt to support its expansion. This is an established, profitable company. Its [fiscal 2026 results](https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Record-Q4-and-FY-2026-Results-Driven-by-Cloud-Infrastructure--Cloud-Applications/?utm_source=chatgpt.com) show approximately $17 billion in net income, but negative free cash flow of $23.7 billion as infrastructure investment absorbed cash. Accounting profit and money available after investment measure different things. A company can report healthy earnings while needing external funding for construction and equipment. At its [October investor day](https://investor.oracle.com/investor-news/news-details/2026/Oracle-Investor-Day-2026-Accelerating-Value-Creation/default.aspx?utm_source=chatgpt.com), Oracle projected positive free cash flow in the second half of fiscal 2028. Customer prepayments, customer-supplied hardware and vendor financing are expected to reduce upfront funding needs. That is a plan with assumptions. Its success depends on delivery, utilisation, payment and the eventual relationship between operating income and investment requirements. ## Palantir: Applications Are Already Making Money Palantir’s [third-quarter results](https://www.sec.gov/Archives/edgar/data/1321655/000132165526000045/a2026q3ex991pressrelease.htm?utm_source=chatgpt.com) provide an important counterweight. Revenue rose 96% to $2.31 billion. Operating income reached approximately $1.08 billion, producing a 47% margin under standard US accounting rules. US commercial revenue reached $984 million, exceeding US government revenue for the first time. The company also generated approximately $1.20 billion in operating cash flow. These figures demonstrate that a business selling software and AI-enabled applications can generate substantial profits today. They do not reveal the return achieved by every customer. Revenue records what buyers pay; it does not measure how much additional profit, reduced waste or improved service those buyers obtain. The distinction matters because renewals ultimately depend on customers finding sufficient reason to keep paying. Supplier profitability is evidence of a functioning business, rather than a complete audit of AI’s economic benefits. ## What The Past Helps Us Understand Palantir was already an established business before the current results. Its [2023 annual figures](https://investors.palantir.com/news-details/2024/Palantir-Reports-Its-Fifth-Consecutive-Qua?utm_source=chatgpt.com) showed $2.23 billion in revenue and $210 million in net income. That history changes the interpretation. Its current performance combines new technological opportunities with capabilities built earlier: customer relationships, software deployment and knowledge of complex organisations. For entrepreneurs, the implication is practical. Access to a capable model is only one ingredient. [Distribution, integration, trust](https://reinouttebrake.com/stories/ai-is-making-games-easier-to-build-getting-players-is-the-hard-part) and an understanding of the customer’s problem can determine who captures value. Oracle adds another familiar lesson: expansion must often be financed before it pays for itself. The longer that interval, the more important [financing terms and reliable customer payments](https://reinouttebrake.com/stories/ai-financing-through-spvs-quietly-shifts-risk-and-debt) become. These observations cannot settle whether AI is overvalued. They explain why useful technology, growing businesses and financial stress can coexist. ## Why The Numbers Cannot Simply Be Added Together There is another trap in reading the boom. When a model provider spends subscription revenue on cloud computing, that expenditure becomes revenue for an infrastructure supplier. Both transactions are real. Adding them together does not establish an equivalent amount of value created for the final customer. That makes independent customer evidence essential. Are businesses producing more with the same resources? Are errors falling? Do savings remain after implementation, supervision and correction? Growing supplier revenue answers a different question from whether an entire industry is becoming more productive. ## How This Reaches Our Businesses And Working Lives For a small company, the consequences appear in ordinary decisions: subscription budgets, recruitment, supplier contracts and promises made to customers. A cheaper AI service might allow a team to offer something previously unaffordable. A price increase or tighter usage limit could weaken the same business case. A tool that accelerates drafting may still require expensive checking. For workers, faster task completion creates choices about workload, service quality and staffing. These financial announcements cannot tell us which choice an employer will make. For business owners, the useful measurement is [the full cost of an accepted result](https://reinouttebrake.com/ai-for-business#accepted-output), including human review. Counting generated documents or completed prompts says little about the money retained afterwards. ## What To Watch Next Three developments deserve attention. First, customer renewals and expansion: whether organisations move beyond experimentation and keep spending because outcomes justify it. Second, supplier economics: whether model providers improve margins and infrastructure operators generate cash after investment. Third, pricing and dependence: whether competition makes useful AI cheaper, or switching costs leave customers exposed to changing terms. Demand could grow rapidly while prices and margins fall. Applications could flourish while particular infrastructure investments disappoint. Several outcomes can happen together. Consider a hypothetical company that saves ten hours each week with an AI assistant. If checking its work takes eight hours, the real gain is two. If the subscription price doubles, that gain may no longer justify the expense. Another company might use the same model to remove a recurring bottleneck and earn considerably more. The technology is identical; the economics are different. That is why evidence from actual workflows will matter more than broad claims about an industry becoming more intelligent. For most entrepreneurs, the opportunity is to solve a customer problem using infrastructure someone else has financed. Keeping data portable and measuring results makes that opportunity easier to evaluate as suppliers and prices change. Would your AI investment still make sense if you had to prove its return today? Filed under [AI](https://reinouttebrake.com/category/ai)[Tech](https://reinouttebrake.com/category/tech) ## Tags - [Anthropic](https://reinouttebrake.com/tag/anthropic) - [Oracle](https://reinouttebrake.com/tag/oracle) - [Palantir](https://reinouttebrake.com/tag/palantir) - [AI infrastructure](https://reinouttebrake.com/tag/ai-infrastructure) - [Enterprise AI](https://reinouttebrake.com/tag/enterprise-ai) - [AI investment](https://reinouttebrake.com/tag/ai-investment) - [Business strategy](https://reinouttebrake.com/tag/business-strategy) Written by [Reinout te Brake](https://reinouttebrake.com/about) Gaming · AI · Tech · Entertainment · Dance Entrepreneur, consultant, investor and board member with 20+ years in tech and gaming, spanning business strategy, development, publishing, growth and investment. [Share on WhatsApp](https://wa.me/?text=AI%20Is%20Growing%20Fast.%20Who%20Will%20Pay%20for%20It%3F%0Ahttps%3A%2F%2Freinouttebrake.com%2Fstories%2Fai-is-growing-fast-who-will-pay-for-it)[Share on LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Freinouttebrake.com%2Fstories%2Fai-is-growing-fast-who-will-pay-for-it)[Share by email](mailto:?subject=AI%20Is%20Growing%20Fast.%20Who%20Will%20Pay%20for%20It%3F&body=https%3A%2F%2Freinouttebrake.com%2Fstories%2Fai-is-growing-fast-who-will-pay-for-it) Cite this article Reinout te Brake. AI Is Growing Fast. Who Will Pay for It? (3 Nov 2026). [https://reinouttebrake.com/stories/ai-is-growing-fast-who-will-pay-for-it](https://reinouttebrake.com/stories/ai-is-growing-fast-who-will-pay-for-it) [Article data (JSON)](https://reinouttebrake.com/data/articles/ai-is-growing-fast-who-will-pay-for-it.json)[Full article text](https://reinouttebrake.com/stories/ai-is-growing-fast-who-will-pay-for-it.txt)[Markdown](https://reinouttebrake.com/stories/ai-is-growing-fast-who-will-pay-for-it.md) Next / Read ## Related reading Follow the subject further. ![Exterior of the Roblox office at 2850 South Delaware Street in San Mateo](https://reinouttebrake.com/assets/editorial/roblox-1280.webp) [Image: BrokenSegue](https://reinouttebrake.com/data#image-credits) Related analysis Artificial Intelligence 2 Nov 2026 ### [AI Is Making Games Easier to Build. Getting Players Is the Hard Part.](https://reinouttebrake.com/stories/ai-is-making-games-easier-to-build-getting-players-is-the-hard-part) Roblox’s AI adoption and Playgama’s game index reveal how game development is changing, what the numbers prove, and why discovery still matters. [Read analysis](https://reinouttebrake.com/stories/ai-is-making-games-easier-to-build-getting-players-is-the-hard-part) ![Rows of server racks at a Macquarie Telecom data centre in Sydney](https://reinouttebrake.com/assets/editorial/servers-1280.webp) [Image: Lgate74](https://reinouttebrake.com/data#image-credits) Related analysis AI 22 Nov 2025 ### [AI Infrastructure Financing: SPVs, Leases and Hidden Exposure](https://reinouttebrake.com/stories/ai-financing-through-spvs-quietly-shifts-risk-and-debt) How SPVs, leases and private credit finance AI infrastructure, using Meta’s Hyperion project to examine where debt and long-term risks sit. [Read analysis](https://reinouttebrake.com/stories/ai-financing-through-spvs-quietly-shifts-risk-and-debt) [AI for Business](https://reinouttebrake.com/ai-for-business) [More in AI](https://reinouttebrake.com/category/ai)[More in Tech](https://reinouttebrake.com/category/tech)[Explore AI](https://reinouttebrake.com/topic/ai)[Explore Tech](https://reinouttebrake.com/topic/tech) [Browse all articles](https://reinouttebrake.com/articles)[Find a specific article](https://reinouttebrake.com/search) [Sources, dates & corrections](https://reinouttebrake.com/data#editorial-policy)