# Anthropic Is Asking Wall Street to Finance the Age of Intelligence ### Editorial Walk into a bank and ask for $10 billion\. The first question will usually be simple: what do you own? Anthropic is offering Wall Street a more ambitious answer\. It owns a claim on the future of intelligence\. The company has reportedly told bankers that the market it may eventually address is worth roughly $30 trillion, about 650 times its current revenue\. Several banks are considering lending it approximately $10 billion from their own balance sheets, partly for the interest and fees and partly for the chance to help lead what could become one of the largest initial public offerings ever attempted\. ![](/writing/fintech-zen/anthropic-is-asking-wall-street-to/assets/editorial.jpg) Call it the denominator effect\. When the imagined future becomes large enough, almost any present investment can be made to look small\. Anthropic is hardly alone in selling that future\. Nvidia posted another record quarter this week\. Alibaba and Tencent are spending billions on AI infrastructure\. Technology companies are ordering chips, securing electricity and constructing data centres at a pace that increasingly resembles the building of railways, power grids and telecommunications networks\. And just as railroads once brought together steelmakers, land speculators, banks and governments, the AI boom is creating its own reinforcing coalition\. Chipmakers need model companies\. Model companies need data centres\. Data centres need power\. Power projects need financing\. Banks need transactions\. Investors need growth stories\. Each part of the system becomes evidence for the next\. Nvidia’s sales prove that demand is real\. That demand justifies more data centres\. The data centres support forecasts for greater computing capacity\. Those forecasts raise the expected value of the companies that will use the capacity\. Higher valuations make further borrowing appear safer\. The circle contains genuine demand, genuine technology and genuine money\. It is not imaginary\. But it is still a circle\. Reporting this week connected more than $3 trillion in off\-balance\-sheet AI financing with pressure on government bond yields\. That is the part of the story that should command attention beyond Silicon Valley\. AI is no longer merely competing for engineers, chips and electricity\. It is beginning to compete for the world’s capital\. This is happening while the United States carries more than $40 trillion in national debt and must refinance enormous sums at higher interest rates\. Investors worried about debt, inflation and currency values are returning to gold and bitcoin\. Governments need capital\. Housing needs capital\. Energy systems need capital\. Manufacturers need capital\. Now AI needs capital too, and it is arriving at the bank with perhaps the most expansive business plan ever written: financing the automation of intelligence itself\. There is a powerful case for doing so\. AI systems are improving quickly\. Businesses are adopting them\. Revenue is growing\. If advanced models become embedded in software, medicine, finance, research, manufacturing and education, today’s market may indeed look small\. But a technology can transform the world without allowing every company that built it to keep the transformation’s profits\. Railroads opened continents and bankrupted investors\. Airlines changed travel while producing famously difficult economics\. The internet remade commerce but destroyed many of the companies that financed its first expansion\. Infrastructure often creates enormous social value while competition, commoditisation and debt determine who actually captures it\. AI may follow the same pattern\. It could become essential while the models themselves become commodities\. That possibility is already visible\. Capable systems are becoming cheaper to train and operate\. Z\.ai disclosed this week that it was the laboratory behind Ox Alpha, a strong model produced without frontier\-level pricing\. Open models continue to improve\. Competitors are willing to charge less, distribute their technology more freely or tolerate thinner margins in exchange for adoption\. Anthropic therefore faces an unusual problem\. It must persuade Wall Street that AI will become ubiquitous without allowing Wall Street to conclude that ubiquity will make AI cheap\. Those are not the same argument\. The first says intelligence becomes a utility used everywhere\. The second asks who owns the meter\. If businesses can switch among models, combine several of them or run capable open systems themselves, much of the economic value may migrate away from the model laboratories\. It may settle instead with cloud providers, chipmakers, application companies, energy producers or the businesses that own the customer relationship\. Bill Gates warned this week that the world remains unprepared for AI’s broader risks\. The financial risk deserves attention as well, not because Anthropic is necessarily overvalued or because the AI boom is fictitious, but because technological importance is being treated as evidence of permanent economic control\. History offers no such guarantee\. The $30 trillion figure is useful because it reveals the assumptions required to make the deal appear reasonable\. AI must spread through much of the economy\. Demand must keep rising\. Anthropic must remain near the frontier\. Competitors must not force prices down too far\. And the company must capture enough of the value it creates to repay the people financing it today\. Any one of those things could happen\. Wall Street is preparing to lend as though all of them will\. ### Zen The Candle Goes Out ![](/writing/fintech-zen/anthropic-is-asking-wall-street-to/assets/zen.jpg) Tokusan had spent many years among sacred words\. He had studied the Diamond Sutra until its verses seemed to move within him like his own breath\. He carried thick commentaries wherever he travelled, and when he met monks who spoke carelessly of sudden awakening, he regarded them with suspicion\. What could they know that he had not already read? One evening he came to the dwelling of Ryūtan, whose name meant Dragon Pool\. They spoke for many hours\. Outside, the moon crossed the sky and disappeared behind the mountains\. The lamps burned low\. Still Tokusan questioned, and still the old teacher answered, until at last the room contained more silence than words\. Tokusan rose to leave\. Beyond the doorway there was only darkness\. The path, the trees and the mountains had vanished into it\. “It is dark outside,” Tokusan said\. Ryūtan lit a paper lantern and held it toward him\. Its small flame opened a circle in the night\. Tokusan reached out gratefully\. At that moment, Ryūtan blew out the flame\. Tokusan stood without moving\. The darkness had returned, but it no longer seemed empty\. It pressed against him from every side, immense and alive\. The scriptures he had carried across mountains could name the darkness, explain it and divide it into a hundred meanings\. Yet none of them could take the next step for him\. And Tokusan understood\. His learning had not been false\. It had brought him to the doorway\. But he had mistaken the light by which he studied the path for the path itself\. We also travel by borrowed light\. We carry forecasts, valuations, models and great numbers intended to illuminate the future\. They show us what may be possible\. They give shape to things that have not yet arrived\. Under their glow, distant fortunes can appear almost close enough to touch\. But a lantern reveals only a small circle\. Beyond it, the road continues in darkness\. A market may become vast\. A technology may transform the world\. A company may remain at the frontier, or discover that the frontier has moved while everyone was studying the map\. The future does not become certain merely because we have found an elegant way to describe it\. Ryūtan offered Tokusan no better theory\. He gave him no second lantern\. He simply removed the borrowed light and left the traveller alone with the road\. Then Tokusan took the first step\. Reminder for the week When a forecast, valuation or confident plan feels especially persuasive, pause before accepting the light it offers\. Ask yourself: What can I see clearly because of this idea, and what might its brightness be hiding? Keep the lantern\. Use it well\. But take one step beyond its circle before deciding that you know the road\. ### The Week’s Best Articles #### [Walmart Reverses Course, Will Accept Apple Pay and Google Pay](https://www.connectingthedotsinpayments.com) — Connecting the Dots in Payments \| 2026\-08\-24 After years of steering customers to Walmart Pay and its OnePay app, Walmart said it will accept Apple Pay and Google Pay in stores, with the rollout starting August 24 and reaching all US locations by the end of the year\. The reversal ends one of the longest standoffs in retail payments\. Walmart had kept the mobile wallets out to control the checkout and the customer data, and to avoid the tokenization fees that come with them\. The competitive signal is loud\. If the largest US retailer decides that friction at the register costs more than wallet fees save, smaller holdouts have little reason to keep resisting\. Broad contactless acceptance also strengthens Apple and Google just as agentic and stablecoin checkout experiments crowd in\. #### [Revolut Builds the EURR Stablecoin and Doubles Down on Proprietary AI](https://linas.substack.com/p/fintechpulse1119) — Linas’s Newsletter / Future of Banking \| 2026\-08\-27 Revolut launched EURR, a euro stablecoin aimed at owning the on\-ramp for euro stables in Europe, extending the neobank from banking into the issuance of tokenized money\. Days earlier it unveiled Revolut Research, a dedicated in\-house AI unit, a bet that fintech’s durable moat is proprietary data rather than someone else’s model\. Together the moves sketch an ambition to control both the money and the intelligence that moves it, rather than rent either from partners\. #### [AI’s Borrowing Binge Is Helping Push Up Government Bond Yields](https://www.nytimes.com) — The New York Times \| 2026\-08\-21 The scale of borrowing to build AI infrastructure is helping drive government bond yields higher, tightening the cost of money across the economy\. With more than three trillion dollars in off\-balance\-sheet AI debt reported elsewhere, the sector’s capital appetite is now a macro force rather than a tech footnote\. Investors have responded with a revived debasement trade, moving into gold and bitcoin as a hedge against fiscal and monetary strain\. #### [US\-Canada Trade Talks Collapse as the Fight Turns Personal](https://www.wsj.com) — The Wall Street Journal \| 2026\-08\-24 Negotiations between Washington and Ottawa fell apart over fine print, and the rhetoric escalated alongside the tariffs, with Canada describing itself as effectively at war over trade\. The breakdown fits a wider pattern of allies pushing back, from Canada to Qatar, testing the limits of tariff pressure as a tool\. With the midterms ahead, both governments face domestic incentives to hold firm rather than compromise\. #### [Nvidia to Acquire Hugging Face for $12\.9 Billion](https://ceoreport.beehiiv.com/p/new-post-c0a2) — The Neuron / AINews \| 2026\-08\-27 Nvidia agreed to buy Hugging Face, the open\-source hub often called the GitHub of AI, in a deal valued around 12\.9 billion dollars\. The acquisition places the dominant AI chipmaker at the center of the open\-model community, raising questions about neutrality on a platform that hosts rivals’ work\. It also lands alongside an OpenAI incident retrospective tied to Hugging Face, sharpening scrutiny of how much the ecosystem leans on a single hub\. #### [Visa Joins Singapore’s BLOOM to Bridge Cards and Stablecoins](https://www.connectingthedotsinpayments.com) — Connecting the Dots in Payments \| 2026\-08\-26 Visa has joined BLOOM, a Singapore initiative to connect conventional payments with stablecoins, positioning itself to route tokenized value through existing rails rather than watch new networks replace them\. The move fits Visa’s pattern of embracing tokenized money at the settlement layer while keeping its network in the middle\. Trade coverage this week also asked who can power Visa’s roughly seven billion dollar stablecoin machine following its BVNK relationship\. For an industry watching stablecoin rules harden, the strategy is telling\. Incumbents would rather absorb the technology than be disintermediated by it\. #### [Stripe Signals It Is in No Rush to Go Public](https://www.connectingthedotsinpayments.com) — Connecting the Dots in Payments \| 2026\-08\-27 In a new investor letter, the Collison brothers argued that staying private gives Stripe more flexibility as AI reshapes commerce, cooling speculation about a near\-term listing\. The letter also drew attention for a bolder claim reported elsewhere, that the company sees signs of an AI singularity already underway in its own data\. Stripe’s patience contrasts with Anthropic’s reported IPO ambitions and underscores that the richest fintechs can now set their own timelines\. #### [National Debt Hits $40 Trillion and Onetime Skeptics Sound Alarms](https://www.theatlantic.com) — The Atlantic \| 2026\-08\-26 The US national debt has reached 40 trillion dollars, and economists who long shrugged at the trajectory are now openly worried\. The change in tone matters more than any single figure\. When former doves begin warning, the political space for ignoring deficits narrows\. The anxiety is compounded by rising borrowing costs and a market increasingly attentive to who ultimately funds the gap\. #### [Z\.ai Reveals Itself as the Lab Behind the Viral Ox Alpha Model](https://www.thedeepview.com) — The Deep View \| 2026\-08\-27 China’s Z\.ai unveiled itself as the maker of Ox Alpha, the mystery model that had been quietly topping benchmarks, showing capable AI without frontier pricing\. The reveal feeds a growing argument that Chinese labs may win AI’s efficiency race even if they trail on raw capability\. Cheaper, competent models pressure the pricing assumptions built into Western valuations\. #### [Cisco Gives All 90,000 Employees a Personal AI Agent](https://www.wsj.com) — CIO Journal \| 2026\-08\-27 Cisco granted every one of its roughly 90,000 employees access to a personal AI agent, one of the largest internal agent deployments to date\. The rollout is a bet that agents can handle recurring work at scale, and a test of what happens to productivity and oversight when they do\. Enterprise agent adoption is moving from pilots to company\-wide defaults faster than many expected\. See you next week, \-dg Check out all our concepts at [www\.QSTVE\.com](https://www.qstve.com) ## Sources - [Walmart Reverses Course, Will Accept Apple Pay and Google Pay](https://www.connectingthedotsinpayments.com) - [Revolut Builds the EURR Stablecoin and Doubles Down on Proprietary AI](https://linas.substack.com/p/fintechpulse1119) - [AI’s Borrowing Binge Is Helping Push Up Government Bond Yields](https://www.nytimes.com) - [US\-Canada Trade Talks Collapse as the Fight Turns Personal](https://www.wsj.com) - [Nvidia to Acquire Hugging Face for $12\.9 Billion](https://ceoreport.beehiiv.com/p/new-post-c0a2) - [Visa Joins Singapore’s BLOOM to Bridge Cards and Stablecoins](https://www.connectingthedotsinpayments.com) - [Stripe Signals It Is in No Rush to Go Public](https://www.connectingthedotsinpayments.com) - [National Debt Hits $40 Trillion and Onetime Skeptics Sound Alarms](https://www.theatlantic.com) - [Z\.ai Reveals Itself as the Lab Behind the Viral Ox Alpha Model](https://www.thedeepview.com) - [Cisco Gives All 90,000 Employees a Personal AI Agent](https://www.wsj.com)