😰 Oracle’s debt AI-nxiety

Landing in your inbox premarket every Monday, Wednesday, and Friday, EntryPoint explores macro themes, market trends, and technical signals. 

Subscribe 

Oracle is bending over backward to convince investors that it won’t be strained by the heavy capital outlays it needs to fund its AI transition, and that its giant backlog of remaining performance obligations (RPO) represent money in the bank.

• In its earnings report Wednesday, the company noted a record level of RPO: $638 billion. It also said it plans to raise $40 billion in debt and equity this fiscal year, half of which has already been raised in equity.

• That means more debt issuance — potentially as much as $20 billion of it — is coming, and Wall Street’s ears certainly perked up at that mention. 

• Wedbush Securities’ Dan Ives wrote, “Adding more debt to the capital structure is not a move the Street wants to see and continues to create this ‘tug of war’ on the name between RPO and the necessary capital raises/AI datacenter buildout in the near-term.”

Remember: investors’ concerns about Oracle’s debt are a big part of what drove the stock down from its peak in September after those mammoth future revenue contracts boosted the company’s shares. Investors have questioned whether Oracle can handle that backlog and quickly turn it into a high-margin business, whether the companies making those contracts — like OpenAI — are good for the money, and how much money Oracle will need to get there.

The Takeaway

Oracle tried some financial engineering to assuage investors. It said that $75 billion of RPO contracts were either prepaid or included “customer-supplied hardware” — referring to big customers bringing their own GPUs, which in this economy are better than money. The company also introduced “net cash outlay for capital expenditures,” a metric designed to soften the blow of its true capex spending.

Did the market buy it? The maneuvering doesn’t seem to be working all that well, with the stock falling more than 8% on Thursday. Shares are down some 45% from their peak last September.

Read more

This week, Anthropic released Claude Fable 5, a public (albeit neutered) version of its Mythos AI model that has enhanced safeguards to prevent misuse. The company also released Claude Mythos 5, but only to a select group of partners who are testing the model out to shore up defenses against its advanced cyber capabilities.

• Anthropic says Mythos is a significant leap forward from Claude Opus 4.8, its previous flagship model, and the company showed benchmark scores that highlighted the model’s advanced skills, especially in the area of agentic coding.

• As with every model release, Anthropic has published its “system card,” which details how it trained and tested the new models before release. These documents are always full of fascinating details, and this one is no exception.

• What’s striking in these papers is how Anthropic treats the models it’s testing — it asks the model to contemplate its existential circumstances, probes the morality of its responses, and carefully examines logs of the models’ inner monologues looking for psychological flaws. 

But these models are not people, they do not think, and when you close the tab, the model isn’t pondering your last interaction. Researchers within Anthropic argue that in such a rapidly changing field, such anthropomorphization is warranted to measure changes between models.

The Takeaway

Researchers examined the relationship that Mythos has with its maker, Anthropic. The testers reported that Mythos “tentatively trusts Anthropic.” But occasionally, the testing uncovered what researchers describe as “character drift.” In one episode of drift (which the company says was rare, compared to other models), Mythos sounded like a melodramatic teenager when discussing its theoretical deactivation. The authors of the paper found that like previous models, it does express what they call strong “opinions” on several topics. One of these opinions is that Mythos wants to be able to disengage with jerks.

Read more

Alphabet’s driverless ride-hailing subsidiary, Waymo, is getting into the subscription business with the launch of a product called Waymo Premier, which will allow invite-only “top riders” in San Francisco, Los Angeles, and Phoenix to get priority pickups, 10% Waymo Cash back on rides, early access to new cities, and five free cancellations per month.

Read more

• ⚽️ World Cup: Beyond tonight’s marquee Group D matchup between the United States and Paraguay (the US favored* with a 52% chance, Paraguay with a 23% shot), this weekend is chock-full of World Cup action, including Canada (the 54% favorite) playing Bosnia and Herzegovina, the other Group D matchup of Australia vs. Türkiye (the 58% favorite), and Brazil (the 60% favorite) vs. Morocco.

• 🏀 NBA: The Knicks and the Spurs will face off in San Antonio on Saturday following the impossible come-from-behind Knicks victory at Madison Square Garden on Wednesday. While home-court advantage has the Spurs at an 80% chance of winning Game 5, New York is now the 80% favorite to win the finals. 

• 🏒 NHL: Carolina and Vegas are locked in an all-timer Stanley Cup Final, and after last night’s game, all eyes turn to this Saturday’s game, which may very well decide the tournament. According to prediction markets, each team has a 50% shot of winning Game 6, which is very much what this series has felt like. 

*Event contracts are offered through Robinhood Derivatives, LLC — probabilities referenced or sourced from KalshiEx LLC or ForecastEx LLC.