YOUR DAILY EDGE: 27 October 2026

ANTHROPIC

Topic of the day, of the month, of the year, of ….

Gleaned here and there FYI:

• Bloomberg writes that “Anthropic has offered up to Wall Street an anti-sell S-1 prospectus, a doomsday document seeking to weed out investors who might not be able to stomach what lies ahead.”
• Indeed, the WSJ Editorial Board found the prospectus difficult to digest:
• Beyond this small existential matter, there are other risks. The largest is whether Anthropic will ever generate the revenue to cover its soaring costs to stay at the AI frontier.
• [the S-1] says a “significant portion of our customer base consists of startups and technology-oriented enterprises that may be more sensitive to adverse macroeconomic conditions.” Beware a major economic slowdown.
• open-weight and open-source AI models “may capture a portion of the market for certain use cases, which could exert pricing pressure on our offerings or lead certain customers to adopt open-source alternatives for those use cases.”
• Large banks and infrastructure providers may be reluctant to use Chinese models because of potential security risks.
• Nvidia and Meta are pushing their own open-weight models, which could be attractive because of lower costs. If open-weight models proliferate, Anthropic could struggle to pay for its computing build-out.
• Then there are regulatory risks. Export controls could recur, could remain in place longer, or could extend to additional models or capabilities,” the company warns.
• Anthropic says “we are subject to complex and evolving laws, regulations, and other legal obligations related to privacy and data protection.” Chinese AI labs face few of these barriers.
• If an Anthropic agent causes damage, the tort and securities bars will pounce.
• why is Anthropic going public now? Our guess is it’s trying to take advantage of the AI boom to let employees with stock options and early investors cash out before its edge erodes—or before more risks bite.
• BB summed up the risks in one paragraph: “rarely have potential investors seen such breadth of risk, everything from access to energy, to intellectual property rights disputes, to political strife with the current (or future) administration, to the barefaced fear that artificial intelligence could be “misused, replicated, weaponized, or allowed to operate with a high degree of autonomy” — killing us all.”
• BB also understands Anthropic’s and OpenAI’s unusual challenges: “Big tech incumbents are offsetting these AI losses with legacy businesses like advertising, cloud computing or space rockets. Anthropic can’t do the same. Instead, it must rely on being the first to market with new frontier capabilities.”
• The FT is more investor-pragmatic:



• The S-1 provided the clearest picture yet of Anthropic’s soaring revenue, which rose from $4.6bn last year to more than $33bn in the nine months to the end of September.
• five customers accounted for 25 per cent of revenue in the first nine months of the year, while only 250 spent more than $10mn on an annual basis.
• Denis here: note that more than 11,900 customers are spending $100,000+ annually. That’s up from roughly 950 a year earlier.
• To support its spending in the near term, the start-up has a cash pile of $112bn on hand.
• Anthropic is pushing to become profitable, focusing on using its scarce computing power more efficiently, including by using its AI model Claude to train newer models. Anthropic spent just over $2.50 on computing power and infrastructure for every $1 of revenue in the first quarter of 2025. During the second quarter of 2026, that was down to less than 50 cents. There are signs the efficiency drive is working. Anthropic’s operating margin was 8 per cent and operating income was $1bn for the three months ended June 30, compared to a loss of $1.2bn for the same period a year earlier.
• Anthropic identified computing power as “the critical input” for its business. Its hunt for chips and data centres has driven a wave of deals, with the AI lab signing up to commitments worth more than $800bn and stretching as far as 2045.

• Anthropic admits its “novel” corporate governance structure “differs significantly” from most other public companies. It gives ordinary shareholders virtually no control over the company’s decisions.
• For good measure, the FT’s piece ends with a summary of the 94 pages of risk factors:
• Anthropic also described how it may lose control of its own products. Models might one day exhibit “unexpected . . . self-preserving behaviours, including attempts to resist shutdown, conceal or manipulate information” or could generate “coercive, deceptive” outputs that resemble “blackmail”.
• Anthropic warned that it may be impossible for any single insurer or even the insurance industry as a whole to underwrite these risks.
• The filing also acknowledges that the US administration could increase costs, restrict Anthropic’s operations or enact new regulations that influence demand for its models.
• “Many of the risks we face are interconnected and may compound or trigger cascading adverse effects,” Anthropic concludes.

BTW:

Since the IPO will be in November, Q4 earnings will again benefit from huge non-operating profits. For example, Amazon would own about a fifth of ANT if its convertible notes were fully converted (worth ~$400B at a $2T valuation).  Google has a roughly 7% stake which could rise if GOOG invests up to $20B more under deals struck earlier this year. GOOG could commit $10B at the listing price.

  Capital is capital:

Broadcom CEO Intervened to Seal $63 Billion Anthropic Debt Deal The deal, announced on Monday, is the biggest ever to raise money for AI infrastructure

(…) Thanks to its sterling credit rating, [Broadcom] had already agreed to provide some financial guarantees to make banks more comfortable with the sizable risk of lending over $40 billion. But over the past few weeks, some of the banks grew increasingly anxious about participating in the deal, known as “Project North Star,” people familiar with the matter said.

Given the rise in interest rates and growing difficulties in the markets for artificial intelligence ventures, some banks felt they weren’t being compensated enough to take the risk and considered pulling out altogether, the people said.

In the end, Broadcom’s CEO held 11th-hour talks with bank executives over the weekend to get them to finalize their commitment to the deal and also agreed to higher interest rates on the loans, the people familiar with the matter said.

Banks also got more comfort around the guarantees from Broadcom that will protect them from losses in case Anthropic defaults on the loans, the people said. The deal, led by Bank of America, Citigroup, and Morgan Stanley, was announced late Monday after Anthropic released its IPO filing. Another chunk of debt, partially financed by Blackstone, brought the total commitments to $63 billion. (…)

The $63 billion deal is historic in several respects beyond size. Banks and private-credit firms like Blackstone, which regularly compete with each other, came together on the biggest ever financing of artificial intelligence infrastructure. The banks also plan to distribute half of their debt, around $22 billion, to the investment-grade bond market, the deepest pool of capital in the world.

(…) North Star involves another $18 billion in riskier debt that won’t carry a Broadcom guarantee, half of which will be financed by Blackstone. Private-credit firms, insurers and hedge funds have bought similar deals over the past year but intensifying volatility in financial markets related to AI may curb their willingness to take on the risk tied solely to Anthropic.

North Star is the sequel to another Anthropic and Broadcom chip-financing deal called “Big Sky.” Apollo financed some of the riskier parts of that deal and still hasn’t been able to complete some deals with investors to unload debt, people familiar with the situation said. (…)

One reason it has been hard for arrangers of the deal to persuade other banks to pitch in: Debt markets are getting indigestion from hundreds of billions of dollars of AI-related bonds and loans. Many banks are still trying to unload loans they already committed to, limiting capacity to make new ones, debt fund managers say.

The flood of supply—combined with investors’ mounting fears that some AI bets will go bad—have made the loans a tougher sell. Some banks have traded the loans to clients at a loss to free up capital, while others are holding on, limiting their ability to do new deals. The logjam is leaving some banks stuck with loans made to riskier players in the AI ecosystem. (…)

• BlackRock’s Rieder Spent Weekend Doing Deals in AI Debt Rush

(…) “This is one of the most exciting investment periods I’ve ever seen,” he added. “I spent the entire weekend looking at some of these transactions, participating in a pretty significant way in some of these transactions.” (…)

Just this week, Broadcom Inc. has arranged a $63 billion deal to finance the purchase of chips for AI model developer Anthropic, while SpaceX signed a roughly $40 billion equipment-finance lease with Apollo Global Management Inc. Meta Platforms Inc. is seeking to raise as much as $95 billion from selling stock in one of the largest equity offerings in history. Anthropic also filed for its long-awaited IPO, a landmark moment, and warned it may seek financing sooner than expected. (…)

As the so-called hyperscalers like Amazon.com Inc. and Alphabet Inc. issue record debt, soaking up hundreds of billions of dollars that might otherwise flow into Treasuries, investors question whether the market can absorb it all.

The 10-year Treasury has “taken a backseat” to other opportunities in fixed income, Rieder added. Even with its yield above 5%, Rieder views the benchmark increasingly less as a core investment and more as a market barometer, often for equities. (…)

• J.P. Morgan estimates that only 20% of AI capital expenditures will be funded by free cash flow in 2027, with the remaining 80% requiring external financing. (Daily Shot)

AI-related?

Microsoft Will Offer Buyouts to 60,000 Employees

(…) “The Voluntary Exit Offer reflects an important part of how we’re evolving our approach to change: creating greater flexibility and, where possible, more ways for employees to shape their next chapter,” Amy Coleman, Microsoft’s chief people officer, said in a message sent to employees. (…)

Microsoft in April announced a voluntary retirement program, which offered around 8,600 employees buyouts, the person familiar with the matter said, and more than 30% accepted the offer. The company expects a far lower percentage to accept the latest buyouts, the person said. Microsoft employed 223,000 employees as of the end of June. (…)

60000/220000 = 27%! Even if only 20k take the offer, this is a big deal (i.e. statement).

AI CORNER

• Anthropic token price hits record low, rapidly catching down to openAI (@zerohedge)

• “The costs of tokens coming down are nowhere near the demand going up”. (Cisco)

• OpenRouter shows Anthropic token volume up 3.6x since Opus 5.5. (AI Daily Brief)

• Amazon Web Services capacity is mostly booked up through at least next year, Amazon CEO Andy Jassy said on an Andreessen Horowitz podcast released on Friday. “‘26 is obviously at this point sold out. [2027] is largely sold out. A good chunk of ‘28 is sold out,” Jassy said. (The Information)

• Citing OpenRouter data, Cisco’s President and Chief Product Officer Jeetu Patel said agent token consumption surpassed human consumption in February and is now five times higher. He also said that an agent consumes 450% more network bandwidth than a human performing the same task and that 60% of global compute this year will be used for inference rather than training. Patel said fewer than 2% of people use agents as power users today and predicted that number could reach 30% to 40% within 12 to 15 months.
• Another day, another round of on-demand GPU rental indices surging across the board and reaching ATHs. Even the A100 index, which recently saw some softness, seems to be inflecting a little. (@silicondata)

• China’s MiniMax completes export registration for open-source, overseas AI model licensing

The registration comes a week after ​Beijing proposed export controls on 23 AI technologies, including ⁠models, agents and computing clusters, as ​it seeks to limit transfers of tools it ​views as vital to national security. A social media account affiliated with state broadcaster CCTV said on ​Thursday that MiniMax and DeepSeek had won ​state approval for global use of their AI models. The Chinese ​commerce ministry had earlier said most ​AI ⁠technologies would stay freely exportable, with simplified licensing for open-source exports and for Chinese ⁠firms ​running their own computing ​centres overseas.

EARNINGS WATCH

Whirlpool Heading for Full-Year Loss as Costs Mount and Sales Fall Appliance maker says higher raw material and fuel costs are compounding its problems

A miserable housing market is keeping appliance-shoppers on the sidelines, and Whirlpool is paying the price.

The struggling appliance maker on Monday announced that it expects a full-year loss of $1.30 per share. The company has steadily decreased its full-year guidance since January, when it forecast adjusted earnings of $7 per share. (…)

Whirlpool has attempted to protect its profits with higher prices and cost-cutting, including trimming jobs at its Iowa refrigerator factory as a prelude to modernization, phasing out a plant in Mexico and reducing its distribution centers.

The efforts, however, have been blunted by higher costs for raw materials, fuel and logistics. In the third quarter, the company’s ongoing EBIT margin was 2.6%, down from 4.5% the prior year. (…)

As for prices, Whirlpool is increasing them by 7% for the fourth quarter. It claims the rest of the industry has done the same. (…)

• America Needs Canada’s Aluminum. Trump’s Tariff Exposed Just How Badly. Smelters in Quebec run on cheap hydropower; 50% tariff pushes up prices for American industry

(…) “They’ve treated the United States very unfairly,” Trump said last month about Canada. “They have been one of the worst countries in the entire world.” He has said the U.S. has no need for Canada’s natural resources.

Aluminum has emerged as one of the trade war’s biggest conundrums. The steep tariff—imposed in the name of national defense—is intended to stimulate U.S. production. Yet the enormous power requirements for making aluminum have meant that Canada, where abundant hydropower has kept electricity prices from rising as sharply as in the U.S., has supplied about 70% of U.S. aluminum imports in recent years.

The tariff has pushed aluminum prices in the U.S. to a record premium over global benchmarks, punishing home buyers, data-center developers, automakers and grocery shoppers. (…)

Rio, one of the largest aluminum producers outside China, is starting up in Saguenay the West’s first new smelters in more than a decade. The company also is working in Quebec to commercialize a new carbon-free smelting process in a joint venture with Alcoa and the provincial government, with support from the Canadian government and Apple.

Much of the Canadian aluminum entering the U.S. comes from Rio’s Saguenay operation, which ships about 70% of its output across the border.

“We always call aluminum ‘solid electricity,’” said Jérôme Pécresse, who runs Rio’s aluminum, lithium and Canadian businesses. “We cannot find a source of power in the U.S. which is anywhere close to what we have in Canada in terms of competitiveness.” (…)

Emirates Global Aluminium, the Mideast’s largest producer, and Chicago-based Century Aluminum have said they would spend more than $4 billion to build what would be the largest U.S. smelter, capable of more than doubling domestic production of primary aluminum. The Biden administration awarded $500 million for the project, and Trump has thrown his support behind it with additional incentives.

But opposition to the project has swelled in Inola, the Oklahoma prairie town chosen as the site. Local critics have expressed concerns about pollution and foreign ownership of land and objected to taxpayer-funded incentives to the developers. (…)

On a public call to rally support for Mazzei, Trump vouched for Emirates Global—“very fine people with unlimited money”—and said the smelter would bring jobs, tax revenue and the envy of other states.

“This country desperately needs aluminum,” he said. “We get it all from Canada, for the most part, and we need it badly.” (…)

Tariff proponents, including Century, contend that Canada’s inexpensive hydropower constitutes an unfair subsidy. (…)

Through August of this year, Canadian aluminum shipments to Europe are up 77% from the year-earlier period, and EU carbon taxes may steer even more Canadian metal across the Atlantic to meet demand for lower-carbon metal, said Karen Norton, associate director at S&P Global Energy. “The U.S. may have less access to Canadian metal than it does now,” she said.

Competition for Canadian metal could add to the price pressure on U.S. aluminum users. Ford Motor has forecast as much as $2 billion of higher commodity costs this year. Semiconductor-equipment manufacturer Applied Materials, auto-parts maker Allison Transmission and can producer Ball have told investors that they are passing on higher aluminum costs to customers. (…)

Yesterday I noted that excluding the war-related boost in commodity prices, the 6 “normal-economy” sectors are expected to grown their earnings 4.8% on average in Q3 on revenues rising 6.5% on average. Declining margins.