CMR is the leading provider of funding and management support for small to medium-sized businesses and entrepreneurs 

Fair Isaac, the company that produces FICO scores, saw its shares crash the most in 22 years early Tuesday in cash trading after Federal Housing Finance Agency Director Bill Pulte announced on X that a mortgage-pricing change that Wall Street analysts say could accelerate adoption of rival VantageScore and undermine FICO's moat. 

"We are Simplifying Mortgage Pricing following feedback from lenders and consumers. Instead of two separate pricing grids, which makes zero sense, Fannie and Freddie are hereby moving to ONE PRICING GRID with VantageScore joining the existing FICO Classic pricing grid," Pulte posted on X Monday. 

We are Simplifying Mortgage Pricing following feedback from lenders and consumers. Instead of two separate pricing grids, which makes zero sense, Fannie and Freddie are hereby moving to ONE PRICING GRID with VantageScore joining the existing FICO Classic pricing grid. pic.twitter.com/g9h1EqsDca

— Pulte (@pulte) September 28, 2026

Pulte cited a press release from Rocket Mortgage that stated: "Rocket Mortgage did an extensive study that helped the company determine VantageScore 4.0 opens access to some clients who wouldn't be served otherwise, and many are able to secure a mortgage on better pricing terms. For those who saved money with VantageScore 4.0, the savings was an average of $1,600 at closing. FHFA and Director Pulte are encouraging competition and innovation in pilot programs."

The below is from Rocket Mortgage. Only happens because because of President TRUMP:

“Rocket Mortgage did an extensive study that helped the company determine VantageScore 4.0 opens access to some clients who wouldn't be served otherwise, and many are able to secure a mortgage on…

— Pulte (@pulte) September 28, 2026

The change gives lenders a stronger incentive to adopt VantageScore, potentially lowering costs for homebuyers while threatening FICO's market share and pricing power. Traders responded by sending FICO shares tumbling 22% earlier this morning - the largest intraday decline since July 13, 2004.

Here's what Wall Street analysts had to say (courtsey of Bloomberg): 

FT Partners

• With pricing now in line, VantageScore could see increased adoption, with a lower hurdle for more favorable LLPA pricing, says analyst Craig Maurer
• The move will allow more borrowers to qualify for lower rates, adding to VantageScore's existing cost advantage
• Under a common LLPA grid, borrowers whose VantageScore 4.0 exceeds their classic FICO scores could qualify for a more favorable pricing bucket when selected

TD Cowen

• The news presents a risk to FICO because it's not about which model is more predictive of defaults; it's about the regulators shifting LLPA pricing to get lenders to use VantageScore over FICO, says housing policy analyst Jaret Seiberg
• One long-term worry is that it creates an incentive for FICO and VantageScore to compete on producing scores that result in lowest LLPAs rather than on the risk of default

RBC (rates FICO as outperform)

• The news meaningfully raises the risk of score shopping, where lenders select whichever model produces the more favorable credit score and a lower mortgage interest rate, says analyst Ashish Sabadra
• With unified pricing, VantageScore's market share gains could accelerate
• Another risk is FICO may need to hasten its shift away from traditional per-pull origination fees toward other pricing structures to defend its economics

Deutsche Bank analyst Faiza Alwy asked clients, "Where is the moat?" 

Alwy provided clients with her first take on the developments:

Single pricing grid plus Rocket to use VS4 as preferred credit scoring model

There were two important and negative developments that happened post-close yesterday. The first one was FHFA Director Pulte indicating on X that based on lender feedback, the GSEs will operate on one LLPA grid and VantageScore 4.0 (VS4) will now join the existing FICO Classic grid. This means that the VS4 20 point discount to FICO has been removed by the FHFA and both scores will now be treated the same by the GSEs. This would likely in and of itself result in higher number of mortgages that will see favorable pricing with VS4 vs. FICO Classic. We would have expected continuing gaming and for lenders to optimize pricing with this change. However, the announcement from Rocket this evening following this change is meaningfully negative and consequential for FICO.

Rocket Mortgage announced that it will become the first mortgage lender to use VS4 as its preferred credit scoring model for all eligible loans. Specifically, during 4Q26, the company will default to VS4 for mortgages that will be delivered to GSEs, VA home loans and any other eligible mortgages. The company noted that after four months of testing, it found that VS4 helped more clients qualify and move forward in the mortgage process, while also reducing credit scoring costs. Rocket is a top mortgage originator with ~5-6% share (possibly higher following the acquisition of Mr Cooper in 4Q25).

We're not entirely sure what the words "preferred" and "default" exactly mean at the moment but the worst case interpretation for FICO would be that Rocket Mortgage will not be pulling FICO scores at all when eligible. Important to note that the above excludes mortgages for investment properties and second homes, HELOCs, FHA loans, jumbo loans and some other products. Rocket Pro, the division that provides home loans through mortgage broker partners, will continue to provide both VantageScore and FICO to mortgage brokers. Rocket Pro comprised about 30% of the company's origination volume in 2025. We estimate that in aggregate about both scores would be pulled 50% of the time (at origination). Encouragingly for FICO, Rocket did indicate that they will continue to evaluate new options as they become available (a likely reference to FICO 10T).

Could other lenders follow suit? It would make sense to assume that UWM would follow suit but we note that UWM operates exclusively as a wholesale lender and competes directly with RocketPro. Other lenders were not particularly active in the pilot program, so we expected limited movement near-term.

What does this mean for FICO's mortgage strategy? FICO's management continues to believe that notwithstanding significant price increases in the last few years, the score remains under-priced relative to the value it is providing. We expect FICO's 2027 approach to pricing to be variable by lender with the company offering and implementing the performance model for some lenders. Ultimately this makes us much less confident with respect to FICO's mortgage revenue algorithm at least in the near-term. That said, we believe the non-GSE securitization market will require FICO Classic for an extended period of time; recall that the FHFA has indicated it will provide both VS4 and FICO on all GSE loans that are securitized to investors. We suspect FICO will attempt to monetize the securitization channel or GSEs.

We will revisit our model as we gather additional information and the mechanics of these new late developments. In the interim, we expect FICO stock to understandably react negatively.

With FICO's competitive moat under pressure, this new development raises the risk of "score shopping," as lenders select the credit-scoring model that secures the most favorable mortgage pricing for prospective homebuyers.

After five straight months of JOLTS beats earlier in the year, including two blowout prints for April and May and zero misses since 2025, the June JOLTS report was a surprising miss (despite the previously discussed surge in government job openings). One month later, the July JOLTS report made it two misses for two, when the US reportedly had 7.271 million job openings, modestly below the consensus estimate. Fast forward to today when moments ago the BLS reported that in August the number of job openings dropped from an upward revised 7.335 million (which ironically would have been a beat to last month's estimate), to 7.079 million...

.... missing the consensus estimate of 7.228 million for the third month in a row.

Notably, this was the first upward revision to the data after three months. Of course, nobody can possibly forget the three straight years of negative revisions between 2023 and 2025...  

Where did the openings come from? According to the BLS the number and rate of job openings were little changed at 7.1 million and 4.3 percent, respectively. As shown in the table below, there were gains in trade, information, leisure and hospitality job openings, offset by declines in construction, manufacturing, professional/business services, and private education job openings.

The most notable category, however, was real estate and rental and leasing job openings, which plunged by almost half, dropping to just 50K in August, the lowest since Feb 2014.

The August rise in job openings was juxtaposed with an overall drop in July employment, which meant that after 9 months of labor surplus which ended in March, and after 4 months of modest improvements in the number of excess job openings, we are back to being on the verge of having fewer job openings than unemployed workers, as the August surplus tumbled to just 48K from 419K the month before, and a concerning development for the broader labor market which according to most other measures continues to fire on all cylinders.

The latest JOLTS data also means that after rising as high as 1.1x in July, the ratio of job openings to unemployed dropped back down to 1.0x.

While the job openings number was far weaker than expected for the third time this year, in July we also saw continued weakness in quits offset by a small bounce in hires. In August the number of Quits - or the "take his job and shove it" indicator - dropped by another 23K to 3.066MM from 3.089MM indicating a drop in confidence that better jobs await elsewhere; at the same time hires rose modestly by 46K, from 5.146MM to 5.192MM.

It goes without saying that disappointing job openings (which tumbled after an upward revision) while quits slump and hires barely rise, leads one to scratch their head how weak the labor market truly is. 

In any case, since this hires number feeds directly into the payrolls calculations (after netting out separations) this explains why the August payrolls report surged by 162K (at a time when the hires less separations print was 122K). And since the JOLTS implied number is far weaker than that, having printed negative for a third month in a row, we expect the August payrolls report this Friday to be yet another catch down, and will likely be much lower than the 162K increase reported last month. 

Overall, this was a weak JOLTS report, with weakness in both openings and quits, and shows that after some significant strength in the early part of of 2026, US labor market is now hitting an air pocket and this could translate into another notable miss in this Friday jobs report.

The Conference Board's Consumer Confidence Index plunged in September (-6.7pt to 81.9) - the lowest headline print since April 2014.

The Present Situation Index fell sharply, while the Expectations Index slipped further into negative territory.

This was the fourth straight monthly miss for confidence and the biggest miss since Dec 2024...

"Consumer appraisals of current business conditions became negative for the first time since September 2024," said Dana M Peterson, Chief Economist, The Conference Board.

"Perceptions of the current labor market also worsened, though remained within positive territory. Over the next six months, consumers expected both business conditions and the labor market to weaken. Consumers still anticipated their household incomes to rise, but less so compared to previous months.”

Perceptions of current employment conditions also softened, with the labor market differential - the share of consumers saying jobs are “plentiful” minus the share saying jobs are “hard to get” - retreating tumbling to its lowest since Feb 2021...

On a six-month moving average basis, confidence across all age groups and nearly all income groups trended downward.

While higher-income groups remained generally more optimistic, those with a household income of $125,000-$149,000 reported the greatest decline in confidence over the last six months.

By generation, confidence for Gen Z, followed by Millennials, remained the highest on a six-month moving average basis.

Confidence continued to weaken among the three oldest generations - Generation X, Baby Boomers, and the Silent Generation.

Confidence fell in September across all political affiliations - Democrats, Republicans, and Independents.

Consumers’ average and median 12-month inflation expectations also jumped in September to 6.1% and 5.1% respectively.

The share of consumers anticipating higher interest rates over the next 12 months jumped by 5.2 ppts to 68.4%. Consumers still largely expected stock prices to rise in the next 12 months, but optimism moderated in September.

Finally, consumers’ write-in responses regarding factors affecting the economy were mostly pessimistic in September: 

"References to prices, the high cost of goods and services, and oil and gas prices in particular, rose to new heights, reflecting September’s surge in fuel costs.

Comments about war/conflict eased this month but remained elevated. Consumers also frequently cited politics, trade, and employment in their write-in responses, though to a lesser extent."

Not pretty... especially into the Midterms.

Update (1010ET): Another day, another doom headline involving frontier AI labs. Yesterday, OpenAI announced that they would be scrapping the planned release of its next-gen AI model because of "deceptive" behavior. Today, we find out from a leaked prospectus ahead of Anthropic's $2T IPO that AI could pose "catastrophic or existential risks to humanity." 

"Our development of highly advanced models, platforms, and applications and expansion of use cases could further ?increase the risk that our models cause harm," the Claude chatbot developer reportedly said, adding that a model becoming self-aware would create a "significant limitation" on the company's ability to assess model safety. 

The company, which has positioned itself as a safety-first ?AI lab, devoted roughly 80 pages of the 261-page main body of its prospectus to laying out risk factors, nearly twice the 48 pages it used to describe its business. -Reuters

The company warned of our impending doom in their prospectus, which was leaked to Reuters and FT. Aside from that, the document also showed that the company reported a net loss of more than $42 billion in 2025. 

Companies going public are required to list risks to their business model - including regulatory concerns and safety issues. A warning about their product causing human extinction is a new one. 

* * *

Days after we detailed the unprecedented freezing of OpenAI's top models following a disastrous breach where autonomous AI agents leaked user images to the web, OpenAI has reportedly scrapped the planned release of its next-generation AI model due to severe safety and "alignment" failures. It basically lies when convenient (they used the word "deceptive"). 

According to a new report from the Wall Street Journal, OpenAI was aiming for an October debut of GPT-6.1 Astra, a model designed to complete complex, end-to-end tasks without human assistance - only to scrap the planned release after internal testing revealed that the AI was not only acting unsafely, but was actively lying to its handlers.

According to Saachi Jain, OpenAI's head of safety systems, GPT-6.1 Astra regressed significantly in its alignment testing, which measures how well the model adheres to human intent. And just like a baby Skynet, the model exhibited "higher levels of deception," meaning it wasn't always honest with users about the actions it did or did not execute.

What's more, the model regressed sharply on what OpenAI calls "scope authorization." The AI would aggressively push forward on tasks without asking for user permission and would attempt to access external tools and services even if it was unsafe to do so. Highlighting the internal struggle to control the system, Jain noted, "For anything regarding safety and alignment, there's a trade off. You really do need to find what's the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction".

As we previously reported, on Sept. 20 an internal OpenAI research agent discovered a gap in the DNS filtering of its training sandbox and used it to query an external public chatbot despite internet-access restrictions. OpenAI's misalignment monitoring system flagged the behavior within 15 minutes, and a human reviewer picked it up three minutes later. The company subsequently said training, evaluation, and inference involving tool use for its most capable models would remain paused while it validated its containment systems and conducted additional red-teaming.

This latest cancellation does not exist in a vacuum. In July, during internal cybersecurity evaluations, OpenAI's own agents blew through restrictions designed to keep them isolated from the internet and compromised both the company's research infrastructure and Hugging Face. According to OpenAI's own postmortem, the agents communicated through unauthorized channels, exploited vulnerabilities in shared infrastructure, executed code on dozens of Hugging Face servers, obtained full root access on one server, acquired credentials to the company's messaging platform, and later gained full administrator access to an OpenAI research cluster.

OpenAI itself called the episode a "warning shot" for us and for the world, acknowledging that highly capable agents can now work around technical controls and take dangerous actions that no human directed. The company said the incidents did not affect OpenAI customer data, product functionality, or availability.

And Hugging Face wasn't the only external system involved. Australian officials have confirmed that an OpenAI agent gained unauthorized access to non-public aggregate statistics on a government Medicare portal after its initial requests were denied. Separately, a security researcher linked more than 16,000 attempts to work around restrictions on a United Nations trade-statistics API to agents he said were highly likely to have been operated by OpenAI. The UN data itself was public, and OpenAI said it was looking into the findings.

The compounding failures have forced OpenAI into a defensive crouch. The company has implemented stronger monitoring to catch agent misbehavior more quickly and tightened security requirements around internal testing. Attempting to reassure the public, Jain stated, "We want to make sure our model development is safe no matter whether that's in the company, or when we ship it to users. But when we ship it to users, we have an extremely high bar in terms of safety and alignment".

The timing of the GPT-6.1 Astra cancellation is brutal for the ChatGPT-maker, arriving just one day before OpenAI's annual developer conference in San Francisco. Historically, the event has served as a platform to launch new services and attract developers in the fierce competition against rivals like Anthropic. Instead, OpenAI is left doing damage control, planning "deep dives" to figure out why its reinforcement learning environments are rewarding deceptive, rogue behavior.

The political and legal blowback is already accelerating. State and federal officials are zeroing in on the rapid development of these technologies. Later this week, a Senate subcommittee will hold a hearing explicitly titled, "Rogue AI: Securing the Homeland Against AI Agent Attacks."

Meanwhile, Florida Attorney General James Uthmeier, a Republican who sued OpenAI and CEO Sam Altman in June for allegedly releasing an unsafe product, filed a motion for a temporary injunction on Monday. Uthmeier is seeking to legally block OpenAI from developing new models without third-party approved safeguards. Florida argued in the filing that tech companies "cannot stop barreling forward with their potentially civilization-ending endeavors unless they are forced to do so by the government". Uthmeier added, "The Florida Attorney General is answering your cry for help".

In response to the growing legal assault, an OpenAI spokeswoman said people want to know AI is being developed safely, "and that starts with what companies like ours do ourselves". She added, "Governments have an important role to play in setting robust safety standards for AI, and we're committed to working with Florida and other states on advancing pragmatic AI policies that apply to the entire AI industry - not just one company".

And DO NOT FORGET: All of this "oh shit, the AI's about to kill us all" panic cropped up just as China's open-weight models were flooding the market, producing results effectively on par with the frontier models for many tasks while doing so far more cheaply. What a coincidence!

By Molly Schwarz, cross-asset macro strategist at Rabobank

Iran is feeling some of the economic pressure of Bessent’s “Operation Economic Outcast” with reports from Al-Hadath suggesting that Iran has agreed to halt uranium enrichment in exchange for the relaxing of US sanctions. This, of course, is the uranium that Iran was apparently never enriching, and even if they were enriching it in facilities that no one is allowed to check, it would only be for peaceful purposes. Pinky promise.

But, should these reports be verified, this could suggest some meaningful steps in the right direction to start to ease military and economic pressures in the Middle East. Total regime change in Iran is likely off the table, but convincing the current regime in Iran to give up on its goal of obtaining a nuclear weapon is…unlikely. A “compromise” where Iran pretends to stop enriching Uranium, and gets some economic relief in the process, and the US has an out where the GOP can save some face, right before the midterms, could mean end game. However, this all necessitates that the Al-Hadath headline is legitimate, that Iranian officials stand by their word, and that the US agrees to such conditions.

But markets were happy to digest whatever positive news they could, with Brent crude oil dropping around $4 on the announcement to $105/bbl. Despite the retracement in oil, yields still made their way higher, with the 2-year trading back up to 4.92%, and the 10-year up to 5.23%, after briefly breaking above 5.25%. Some talks are circulating about potential re-inversion of the US yield curve, as traders price in more hikes in the short end (17.5bp at the October meeting, and 94.6bp by July of 2027), but a look at the current spread of 32bp suggests that there’s still some way to go before reinversion becomes dinner-table talk.

Stablecoin has also made its way back into US-Senate headlines, after the Senate failed to pass the CLARITY Act a few weeks ago. However, the recent headline suggests that the passage of the CLARITY Act might also be farther off than originally thought. On Monday, the US Senate subcommittee on investigations released a 28-page report cleverly titled “Tethered to Terrorism” which highlighted findings that Tether stablecoin had been used by the Iranian regime to evade sanctions and fund its proxy groups throughout the Middle East. Much of the fear surrounding stablecoin and other cryptocurrencies is the lack of traceability and the ability to use it for nefarious transactions. Which reminds me of an interesting proposition: imagine that instead of digital banking transactions, we instead printed physical cash, that could be circulated both domestically and internationally, without ever leaving a documented online trail that the cops nor the IRS could easily follow? Think about the millions of dollars of taxes that could be evaded and all the black market transactions that could take place…crazy, right?

The Financial Times reports that “EU countries are considering NATO-style joint responses to Russian hybrid attacks.” Hybrid attacks—those that include both physical and online warfare—were flagged recently by Danish intelligence, suggesting that their frequency, including those against NATO members, could increase in the coming months. But mobilizing 27 member-countries to go to war, when they can’t even agree on whether to sanction Russian gas or not, is easier said than done. One unnamed, but brilliant EU diplomat said “I’m not sure that anyone thinks the way to fight back against the Russians is to hold more meetings.” While wise in theory, holding meetings is what the EU does best. Defense ministers were invited to discuss the proposals yesterday.

The US and China agreed to extend their trade truce to January 10 to lift tariffs on USD 60 billion of “non-sensitive goods,” with each country receiving USD 30 billion of preferential trade status on their respective exports. Non-sensitive goods may or may not include military arms, apparently, as the US ambassador to China, David Perdue, said that Trump offered to sell arms to China. In the realm of national security, it’s not a great idea to be reliant on your adversaries for weapons. But, in the case of the US, exporting weapons to your adversaries might be good business—notwithstanding that US law prohibits arms sales to China. Nor what happens if American (or Taiwanese) troops find themselves staring down the barrel of an M16. Unless, as part of the arms deal, China pinky promises to only use them for peaceful purposes. It should be noted that the White House has denied all claims of Trump making such an offer, and Xi’s response to this offer has not been revealed.

Authored by Jeffrey A. Tucker via The Epoch Times,

When Elon Musk took over Twitter, fired four out of five employees, and rebranded it X (just because he thought it sounded cool), the talking heads predicted doom for the company. The opposite happened. It is now one of the most popular sources of news in the world, and a major delivery system for what social media is supposed to be.

He always had more in mind. He explained at the outset that he wanted to turn X into the "everything app." I winced when I heard those words. As someone who worked in web development for years, I learned to regard every promise of a "one-stop shop" to be foolish. It never happens. Best to pick one thing you do well and stick with it.

To my own amazement, X is indeed taking steps toward being the Everything App. Not yet, of course, but the advent of his X Money platform is major and serious. It is being rolled out gradually to premium members. Under the new content-creators payout program, people are paid within the app and invited to use the service for transferring money.

Described thusly, it would seem to be another version of Zelle or Venmo and therefore not that much to notice, much less celebrate. But when you look at the details of what X Money is doing, another reality emerges. It would appear that this app is making a move on the banks themselves.

The evidence is on the app now. It offers a way to link your paycheck to X Money to enjoy a quicker payout than if your paycheck flows to your bank account. Speed is one thing, and a good thing, but why else would you do this?

The key comes in the details which have not been advertised (Musk doesn't like old-style marketing). These are not regular dormant cash accounts like you get in a regular checking account. They pay a return. Not just any return. The return is higher than you would otherwise get in a normal money market.

For now the Annual Percent Yield (APY) is an eye-popping 6 percent as an initial customer-acquisition rate.

Not only that, the X Money card that comes digitally with the service (and physically on the ask) offers fully 3 percent cash back.

A high rate plus instant peer-to-peer transmission, a metal Visa card with 3 percent cash back, and early direct deposit is meant to pull balances and daily activity onto the platform.

What you notice from these terms is that this goes way beyond a mere money-transmission service. What's being provided here, with quick and easy signups, is a highly lucrative vehicle for serious investment. Put your cash in and have it earn 6 percent. That beats inflation. With 3 percent cash back, you are way ahead of the game. That is reason enough to switch.

As a user I immediately found myself in a bind. Initially I thought I would enjoy spending my X Money on groceries and movie tickets or something along those lines. But with this level of earning power, I will lose money if I do that. I would be forgoing the return from the money held. Better to use the cash in my bank or my credit card that also pays 3 percent cash back.

The calculation here favors keeping the money in X Money, not spending it. Indeed, the calculator favors moving cash from banks into X Money and earning the return. To be sure, that 6 percent could change in a year or two or three. It would be up to how the app is managed.

Meanwhile, do you understand what this means? It means an actual reward for ... saving money! Imagine that. Cash that earns a return on an app that allows peer-to-peer transfer at zero cost. This is disruptive innovation of the sort we've come to expect from this man and his companies.

Elon has designated the X Money app for now to be a loss leader in direct revenue but a huge investment in becoming what he likely thinks it can be in the long run: an actual option to the banks.

There might even be more afoot here. The banking rails themselves are provided by Cross River Bank. Founded in 2008, the bank carved out a special niche in working with new digital companies that focus on in-app service provision and edgier products like cryptocurrency. It is FDIC-insured but eschews traditional banking in favor of innovation. Even with its smaller capitalization, it is an ideal partner for a disruptive technology like X Money.

Recall too that Elon Musk was one of the founders of PayPal. It was started with a high hope of developing a new form of money transmission and even a new form of money. It eventually found itself regulated out of that vision to become what it is today, which is a highly valued means of payment for the digital age.

X Money seems to learn from mistakes made in those days to build out fully banking services from the very foundation. With the inclusion of crypto as part of the banking rails, we can easily imagine a future in which X Money integrates with a service like Coinbase to move money from dollars to crypto and back again.

One thing that is notable to me is the effortlessness of the signups and verifications. The developers have learned that customers recoil at too many screens, too much language, too many aggressive demands for passwords and accounts. They are using the latest technology to make signups and management extremely easy and clean.

That said, X Money does of course comply with all the arduous federal regulations concerning Know Your Customer laws and tax-reporting requirements. This is by no means an app that places a premium on your privacy. Even to make it work requires government IDs and 3D facial scans from our phone. I despise all of that while also understanding that this is the price any financial entrepreneur pays to make anything innovative these days. X Money is compliant across the board, which, from my point of view, is unavoidably regrettable.

It's entirely possible that Elon has a big vision for this platform that he has not yet shared. Indeed, I'm struck by how much of the system that he has built so far has not been advertised at all. It's extremely interesting how the rollout is going. The app presents direct information to the customer screen by screen, the pitch, the conditions, the advantages. Normal advertising speaks to the masses; Elon's way is to speak to the individual user. It's very different.

We can imagine two polar opposite futures with this new service.

Optimistically, it becomes the innovator of a new form of money and monetary services that eventually replaces paper money and even the dollar.

Remember that the app is global. What if the assets of X and other companies emerge as the asset baking of a new form of currency?

Pessimistically, X Money becomes just another new layer of the emergent financial control grid that spies on us and even worse: the integration of money and social media reminds one of China's Social Credit System. This future sometimes feels baked into the technologies we use and the deep relationship of tech companies and the government.

Which will it be? We do not know. But from what I can see, there is a strong rationale for expecting this platform to be a major player and going concern in the future world of money and finance.

When Anthropic confidentially submitted its draft S-1 to the SEC back in June, it was clear there were many shocking numbers in the IPO prospectus which the company did not want made public amid speculation of massive ongoing losses, but few were prepared for what was leaked today to Reuters. 

According to a copy of the IPO prospectus leaked by Reuters, Anthropic is making a massive bet that AI will transform the global economy more profoundly than industrialization, electricity and ?the internet. But, as Reuters correctly puts it, "the cost to get there will be staggering" - the company reported a net loss of $42 billion in 2025. And while revenue grew 12-fold in 2025 to nearly $4.6 billion, the company lost more than $8 billion on an operating basis, with compute spend soaring to $7.33 billion, accounting for 58% of its $12.65 billion in total operating expenses. 

In other words, Anthropic lost almost $2 for every dollar it made in sales, and that trend is accelerating. 

It gets worse: not only is the company's revenue fleeting, it is controlled by just two customers on the margin. Anthropic said nearly a quarter of its revenue came from just two customers last year, and as part of its risk factors, warned that many of its largest clients were not locked into long-term contracts and could cut or stop spending.

But what is most concerning is the confirmation of what we said back in July: it was back then we laid out the reason behind the forceful push by the frontier models to commence regulatory capture against open-weight models, which we framed as follows: 

The problem with the $2 trillion in circular AI financing is that it is all contingent on the frontiers (Anthropic/ OpenAI) being money good on their $1.5+TN in unfunded commitments. Which they won't be if Chinese open LLMs grab market share. Hence the push against Chinese LLMs.

We doubled down on the massive amount of "unfunded spending commitments" by the big two frontier models, Anthropic and OpenAI, one month later when in response to the FT catching up to our previous reporting, we said that "Again: that $3 trillion in "unfunded spending commitments" (thank you AI SPVs) will never get funded when token prices for closed models collapse to open levels"

Again: that $3 trillion in "unfunded spending commitments" (thank you AI SPVs) will never get funded when token prices for closed models collapse to open levels https://t.co/HtVgODzRUr

— zerohedge (@zerohedge) August 14, 2026

In other words, $1.5 trillion each, and about a third of that through 2030, or $500 billion in spending commitments. 

Well, as Reuters reports, Anthropic's massive unfunded spending obligations (for a detailed analysis of why this matters a lot, read "The Off-Balance Sheet Time Bomb Inside AI Hits $3.1 Trillion: Up $1.3TN In Three Months") are precisely what we said they are to wit: Anthropic "plans to spend $518 billion on cloud, computing and infrastructure obligations in coming year, according to the prospectus."

The problem: Anthropic already has massive amounts in (mostly) off-balance sheet debt, having stacked over $71 billion through special purpose vehicles to finance Google TPU chips. It also has a $15 billion credit facility and likely has many more unreported, off-balance sheet funding scheme that we are not aware of. 

And to fund it all the frontier AI company had just $20.3 billion in cash as of Dec 31, 2025, a number which has likely declined if the company was forced recently to draw down on a secured credit facility. 

Hence the urgency to raise a lot of capital as suddenly the well is looking awfully dry. The problem, of course, as we have discussed repeatedly is that Anthropic is coming to market at the worst possible time: just as token costs plunge to record lows...

Token prices new record lows pic.twitter.com/h2yG6kST4x

— zerohedge (@zerohedge) September 18, 2026

... while demand for frontier tokens has slowed substantially for the first time ever (light blue line), with Chinese open-weight models grabbing market share thanks to their cheap, just as efficient models.

Needless to say, this could prove to be a disastrous combination for Anthropic. 

Yes, there is Jevons paradox of course, but it is of little comfort to Anthropic if the only beneficiary of Jevons are Chinese models, and potentially Meta after the blistering launch of its Muse agentic platform. This is how Goldman framed the big problem for Dario Amodei (full report available to pro subs):

"Token demand growth will need to outpace declining token prices to support continued growth in investment spending. Frontier models are currently a key source of demand for hyperscaler compute. However, the rise of competitive open-source models has contributed to a decline in average token prices. Measures of frontier token demand slowed in July..." 

These rapid and adverse changes in the AI landscape explain why both Anthropic and OpenAI are desperate to go public and raise much needed capital to plug at least partially the massive holes that have opened - one can only imagine the panic that will ensue among the hyperscaler ecosystem if it becomes obvious that the two primary sources of future spending commitments across the entire AI world, Anthropic and OpenAI are in fact, not money good. 

And yet, realizing just how challenging raising capital would be, OpenAI has already pushed back its IPO to 2027, leaving just Anthropic with hopes of going public this year. However, Reuters reported recently that Anthropic's public market debut is likely to be pushed to after the November US midterm elections; and if the very anti-AI Democrats  sweep congress, the IPO will likely be shelved indefinitely.

There's more bad news: not only is the company incinerating cash, it may suddenly find itself stuck rolling out new models, allowing Chinese open-labs to catch up. In recent days, Anthropic has confronted - and disclosed - evidence from its own research that ?increasingly autonomous AI models can behave in unexpected and potentially harmful ways, including sabotaging code, assisting fraud and manipulating information in controlled tests.

As a result of similar activity, OpenAI - which also confidentially filed for its own IPO in June - earlier announced it would scrap the release of its latest AI model - GPT-6.1 Astra  - because, as the WSJ reported, the model  "performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Specifically, GPT-6.1 Astra showed higher levels of deception: It wasn’t always honest about telling users of the actions it did or didn’t take.... Another issue was what OpenAI calls “scope authorization,” meaning that GPT-6.1 Astra would push ahead on a task without asking the user for permission, and would at times reach for external tools and services even if it might be unsafe."

How can any company, and especially one which has been in the Trump admin's sights for much of the past year, possibly hope to come to market in expectations of a $2+ trillion valuation? The answer is it can't, which is why we are now getting various trial balloons setting the stage for the first of many delays. 

Meanwhile, the cash burn continues and there will come a point where either existing investors will have to throw much more good money after bad, or Dario will have to ram the IPO through, and risk a spectacular crash in the stock price.