On the heels of its blockbuster IPO, SpaceX announced that it plans to acquire AI coding startup Cursor in an all-stock deal worth $60 billion by the third quarter of 2026. The deal would double the net worths of Cursor’s four young billionaire cofounders, Michael Truell, 25, Aman Sanger, 25, Sualeh Asif, 25, and Arvid Lunnemark, 26. Forbes estimates they will be worth $2.7 billion each.
They aren’t the only ones poised for a windfall from the deal. Early investors including Andreessen Horowitz (which reportedly owns a roughly 10% stake, worth $6 billion) and Thrive (which owns about 7%, worth $4.2 billion, per a source familiar) stand to win big from the deal too.
Cursor has come a long way in a short span of time. Founded in 2022 by four MIT friends and 30 Under 30 alumni, the startup started out as a coding tool for developers. After AI behemoth Anthropic launched Claude Code, Cursor found itself on the defensive, shifting to “war time” mode, as Forbes reported in March.
It appears to have worked. In early June, Cursor crossed $4 billion in annualized revenue, fending off stiff competition from both Anthropic and OpenAI, Forbes reported. Its revenue grew from $2 billion in February to $3 billion in late April. The uptick in revenue is in part due its new product Cloud Agents, which works on complex programming tasks for hours in the background.
Cursor first teamed up with SpaceX in April, when the rocket maker obtained the right to acquire it for $60 billion, or pay $1.5 billion in breakup fees and $8.5 billion in computing resources if the deal didn’t go through. SpaceX, which also acquired xAI in February, had been struggling to improve the capabilities of its models as AI researchers left en masse. But it has a ton of compute, thanks to its gigantic Colossus supercomputer. For the past few months, the two companies have been jointly training a new AI model that will be released in Cursor and xAI’s Grok, SpaceX said in a post.
Now let’s get into the headlines.
BIG PLAYS
On Friday, Anthropic abruptly disabled its new AI model, Fable 5, after the U.S. government issued an order to ban foreign nationals from accessing it, citing national security concerns. The directive came after government officials learned of a way to jailbreak the powerful model’s safeguards. Fable 5, a more secure version of Anthropic’s Mythos family of models, had been in restricted access for months and was launched just days earlier to millions of people. A group of tech leaders including Amazon CEO Andy Jassy had flagged concerns over the model’s security risks to senior Trump officials last week, Reuters reported.
“You have to make a judgment call on these things,” Anthropic’s Chief Commercial Officer Paul Smith told Forbes just hours before the order was issued. “The safest you can be is to not let people use something. And then it’s totally safe. But then how is that helping the mission?”
Also notable: OpenAI’s spending reached $34 billion last year amid a neck-in-neck race with rival Anthropic to dominate the AI market, the Financial Times reported. The giant’s costs far outweigh the $13 billion in revenue it booked in 2025.
SHOW ME THE MONEY
SpaceX’s historic IPO made scores of stakeholders ultra-wealthy. The rocket maker and AI company started trading just before noon on Friday at $150 per share, implying an eye-popping $2 trillion valuation. CEO Elon Musk became the world’s first trillionaire. Early investors like Peter Thiel’s Founders Fund and Antonio Gracias’ Valor Equity Partners’ stakes in the company are worth $67 billion and $71 billion, respectively. As of Tuesday afternoon, SpaceX’s market cap had skyrocketed to $2.8 trillion, surpassing Amazon as the world’s fifth largest company by market value.
AI DEAL OF THE WEEK
Jeff Bezos’ AI venture Project Prometheus raised $12 billion in funding at a $41 billion valuation. Bezos runs the company as co-CEO with Vik Bajaj, a cofounder of Alphabet’s life sciences research lab Verily and a Stanford University professor. The nascent startup is building AI tools to help engineers design and manufacture physical products faster. It plans to use the funding to buy up compute, according to CNBC.
DEEP DIVE
If you’re interested in renting an apartment in one of Equity Residential’s 300 properties, chances are you’ll soon be chatting with Ella to set up an apartment tour or answer questions about a lease.
But Ella isn’t human. It’s an AI assistant that answers the phones and responds to hundreds of emails around the clock. It’s still sending quick replies after all the humans have gone home, when most inquiries typically come in.
Ella is so helpful that some people don’t realize they’re talking to a bot.
“Customers were calling in asking for Ella and saying, ‘We just love her work ethic.’ They wanted to make sure she was going to get her commission,” says Kristin Hupfer, a senior vice president of customer experience at Equity Residential. The firm discloses that Ella is an AI chatbot the first time it communicates with a person through email, phone or text, Hupfer says.
Ella has been a gamechanger for the Chicago-based property manager, which owns buildings in New York, San Francisco and Seattle. Back in 2018, the firm’s staff struggled to keep up with the influx of hundreds of requests from prospective tenants each week. That in turn meant losing customers to rivals and keeping units idle. Then in 2019, it started working with New York-based EliseAI, the developer of Ella. Now the bot handles 1.5 million texts, emails and phone calls every year, allowing Equity Residential to save $20 million in payroll costs, Hupfer says (no layoffs, she clarifies, just not replacing staff who left). The real estate firm owns two buildings in Jersey City that don’t need a human staff member at all because they can be managed from a nearby community and Ella handles all the administrative tasks.
Today one in six apartments in the U.S. and 90 percent of the country’s largest property managers use EliseAI’s tools to respond to questions about a unit, renew leases and triage maintenance requests. It can even use smart locks to let renters into an apartment for a tour, or determine that a request to fix a broken A/C unit in the summer should be prioritized.
Read the full story on Forbes.
MODEL BEHAVIOR
Anthropic’s AI models are incredibly powerful. That is, unless you’re an AI researcher developing frontier large language models that could eventually compete against them. Anthropic disclosed last week that its Fable 5 and Mythos 5 models deliberately become less helpful if they detect another AI research lab using them. Rather than outright refusing to produce an answer, the models secretly modify user prompts to change its own responses.
