Claude’s crossover moment

Good morning,

A Southwest flight was delayed by an hour, not because of weather, but because of a humanoid named Bebop (his battery exceeded the allowable size). It seems even robots have carry-on issues.

The 30-year Treasury yield topped 5.1% on Friday, the highest in nearly a year, in a sign that bond markets are bracing for higher inflation after last week’s hot prints and surging crude prices. All eyes will be on Nvidia earnings tomorrow. In today’s letter:

Claude’s crossover moment… For the first time, Anthropic has surpassed OpenAI in adoption by U.S. businesses. According to the Ramp AI Index, Anthropic adoption rose 3.8% in April to 34.4% of businesses while OpenAI adoption fell 2.9% to 32.3%. We’ll see if the launch of GPT-5.5, praised in tech circles, and Codex freebies will turn the tide for OpenAI. In this nascent space with low switching costs, there’s no entrenched winner, and the playing field changes fast:

AI’s low-loyalty era is partly a result of low lock-in. For one thing, it’s way less of a headache to switch AI models than it is to switch, say, your CRM, ERP, or HRIS software. Meanwhile, the pace of innovation creates ever-shifting competitive advantages that can sway customers. We’ve seen this back-and-forth of customers turning on either OpenAI or Anthropic due to policy stances, product updates, or changes to billing models (most recently with Claude subscriptions no longer covering third-party usage).

“We have never seen a software industry as dynamic, where newcomers can disrupt market leaders in a matter of months,” wrote Ara Kharazian, Ramp’s lead economist.

The AI providers with lock-in are not pure AI companies. Diversified players like Google and Microsoft have indirect lock-in though their entrenched enterprise products (e.g. Google bundling Gemini into Workspace, Microsoft’s Copilot integration). Ramp data likely undercounts Google adoption (now at 4.5%) because many businesses have Gemini rolled into their Workspace plans rather than paying for it separately.

Without lock-in, leadership is a fling… In traditional software, enterprise adoption often creates long-term commitment. AI model providers have minimal switching costs. That can be good, allowing for surging adoption in a short time span, but it’s also a liability. With two dominant players, this creates a race to the top (offer the most advanced models), but also a race to the bottom, where labs may compete to offer the most freebies. The lack of loyalty also allows for a dark horse scenario, where a lower-priced underdog rises fast.

CPI and PPI came in hot, and Fed rate cut chances are getting shot (try saying that three times fast…). 

Producer prices jumped 6% in April on an annual basis, the largest 12-month spike since December 2022. For the month, wholesale prices rose a seasonally adjusted 1.4%, about 3x higher than the Dow Jones consensus forecast. 

Consumer inflation also came in uncomfortably warm (highest since May 2023) for the same reason: fuel prices. Energy accounted for 40% of the CPI increase, but many things that don’t fall into the category are still affected by it. Most obviously: transportation.

Retail sales rose 0.5% in April, largely due to higher spending at the pump. Several discretionary categories, like furniture and clothes, were down. 

Odds of an interest rate cut this year are looking slim, while bets on rate hikes are rising. Traders are now pricing in a ~47% chance of a hike in December. 

The Fed’s meeting minutes, set to drop tomorrow, should shed light on how the central bank was leaning on April 28 and 29 (before the inflation prints came out). After eight years with Jerome Powell at the helm, investors are looking for signs on how the Fed will move under Kevin Warsh, who officially took over as Chair this month.

Uber and ServiceNow are among the companies that have reportedly already blown through their AI budgets for the year. And the Financial Times reported that Amazon employees are using AI tools for unnecessary tasks to inflate usage scores. 

Employees are adopting AI faster than any previous technology, and a lot of that’s happening through informal buying. Picture: signing up for an individual subscription, then expensing it. New Ramp data shows: 

Unlike fixed seat-based SaaS pricing, the AI billing model is largely based on token usage. That’s leaving many companies unable to predict their bills in any given month. 

But there’s a lot of room for negotiation and cost efficiencies in this still-new category. Here’s your cheatsheet to buying AI.

The CEOs that accompanied President Trump to China have a combined net worth of about $1 trillion

Tuesday, May 19: Google I/O conference begins. Earnings expected from Home Depot and Cava

Wednesday, May 20: FOMC meeting minutes. Earnings expected from Nvidia, Target, TJX, Hasbro, Lowe’s, and Intuit

Thursday, May 21: Housing starts. Weekly jobless claims. Earnings expected from Walmart, Deere, Nio, Ross, Workday, and Zoom

Friday, May 22: University of Michigan consumer sentiment. Earnings expected from BJ’s Wholesale and Booz Allen Hamilton