OpenAI to Cut Off Model Services to Cursor, Musk Fires Back
According to reports from multiple international outlets, on August 28, OpenAI announced plans to halt model services to Cursor, the AI-powered code editor under SpaceX’s umbrella. OpenAI took to social media to confirm that it has officially notified SpaceX of the termination, with a proposed end date of November 12, 2026. The company explained that this decision stems from “past experiences,” noting that several of Musk’s ventures have a track record of contractual breaches, making it hard to trust that SpaceX would use the technology within the bounds of the service agreement.
Musk didn’t hold back in his response on X, saying he “couldn’t care less” and doubling down on his criticism of OpenAI’s leadership. He accused CEO Sam Altman and President Greg Brockman of being “completely untrustworthy,” reiterating his claim that they “stole an open-source nonprofit.”
Report: For Every $100 AI Companies Earn, Nearly $40 Goes to Cloud Giants
Barclays recently released a research report on the AI industry that’s turning heads. The British bank found that for every $100 in revenue AI model companies bring in, roughly $35 to $40 flows straight to the three major cloud providers—Amazon Web Services (AWS), Microsoft’s Azure, and Google Cloud Platform (GCP)—in the form of inference compute costs. The cloud providers walk away with about $10 to $20 in operating profit from that, which translates to a hefty 35% to 45% operating margin.
Meanwhile, AI labs’ paid inference businesses are seeing their profit margins skyrocket, jumping from low double digits in 2025 to anywhere between 50% and 65% (or even higher) by 2026. That’s a massive 30 to 50 percentage point year-over-year improvement in adjusted gross margins. Barclays analysts caution that while the actual margins might be even higher than what the report estimates, they expect them to gradually cool off as competition among frontier models heats up and compute supply keeps expanding.
Lambda Closes $926 Million Loan to Deploy GPU Infrastructure
On August 27, Lambda, the U.S.-based AI cloud services provider, announced it had completed a $926 million senior secured Term Loan B facility. The funds are earmarked for purchasing and deploying GPU infrastructure to power compute services for an investment-grade client. The loan carries a Baa2 rating from Moody’s, priced at SOFR plus 3 percentage points, and matures on December 31, 2030.
a16z’s Machine Age Fund Raises $1.1 Billion to Bet Big on Hardware
On August 28, Andreessen Horowitz (a16z), the legendary Silicon Valley venture firm, made it official: hardware is now a core focus. The firm has raised $1.1 billion for its Machine Age Fund, which will pour capital into all the computer infrastructure AI depends on—chips, memory, networking, and storage—as well as the full systems that run AI, from data centers to robots to home AI devices. The firm says it’s already backed several hardware companies, including Unconventional AI, Nexthop, Volta, Atoms, and Mind Robotics.
First Batch of ChiNext Computing Infrastructure ETFs Gets Green Light
According to Shanghai Securities News, the first batch of ChiNext computing infrastructure ETFs has officially received approval, as confirmed by industry insiders. The initial wave includes 10 fund managers: E Fund, China Asset Management, Southern Fund, GF Fund, Fullgoal Fund, Guotai Asset Management, Harvest Fund, Tianhong Asset Management, Penghua Fund, and Dacheng Fund.
