Technology

➚ MOVE FAST: Output. Chinese AI chip designer Cambricon recorded a record profit in the first half of this year as Beijing mobilizes to rely on homegrown technology. The Financial Times reports that the country plans to triple its chip production as competition with the US heats up.

➘ BREAK THINGS: Input. Anti-Israel protests by employees have been dogging tech executives at conferences for months, but the latest dustup hit Microsoft President Brad Smith at his office. Members of the group No Azure for Apartheid hung banners there before they were arrested, while the company said it’s reviewing its business with Israel.

On a Friday in late 2022, CoreWeave CEO Mike Intrator and Co-founder Brian Venturo were given a 15-minute warning that they were about to be grilled via videoconference by Jensen Huang. The conversation between the two crypto miners-turned-AI entrepreneurs and Nvidia’s CEO could make or break CoreWeave, and Intrator didn’t even have time to get his faulty web cam working, the co-founders told Semafor.

What followed was an hour and a half of intense questioning. Why did they build their cloud this way? How would they compete with tech giants? Why were these East Coast guys focused on making money instead of burning venture capital to grow at any cost?

When Intrator started talking, Huang cut him off. “Who the hell is this guy?” he said.

“I’m the CEO,” Intrator replied.

“Well, maybe next time you’ll be on camera,” Huang said.

Despite technical difficulties, Venturo and Intrator had the answers. By the end of the call, Huang had made his decision: He turned to a colleague and said, “Tell the guys we’re going to invest in these people.” That 2022 validation — Nvidia invested about $900 million into the company, now worth more than $2 billion — launched CoreWeave into the AI big leagues with access to the chipmaker’s most advanced products to build the powerful data centers used to serve and train many of OpenAI’s models.

Starting with its relationship with Nvidia, CoreWeave has been treated with skepticism at every turn, only to eventually come out ahead. But now, after a splashy initial public offering that saw its valuation quadruple, CoreWeave is facing questions that go far beyond its own balance sheet as the market casts doubt on the long-term profitability of AI.

Deepen your understanding every day with the Medium Newsletter. Join 2 million curious readers learning from deep, authentic first-person perspectives on what’s happening in the world. Subscribe now for clarity in your inbox every weekday.

The relative change in employment headcounts for early career workers — aged 22 to 25 — in jobs that are highly exposed to AI, including software engineering, marketing, and customer service, according to a paper published Tuesday from Stanford University researchers. Meanwhile, employment trends remained the same for more experienced workers and improved for those in jobs less vulnerable to AI developments. The data gives credence to predictions by prominent tech leaders that AI will replace some of the workforce, while many business executives downplay fears of job loss.

Elon Musk’s Starlink and T-Mobile have separately expressed interest in acquiring some or all of the spectrum controlled by struggling DISH Network-parent EchoStar, Semafor’s Rohan Goswami and Liz Hoffman scooped yesterday.

At stake is one of the largest pots of underused airwaves, which are owned by the public but licensed to telecom companies. EchoStar’s chairman, Charlie Ergen, has hoarded his spectrum holdings with the unfulfilled ambition to build a national network, but has come under pressure from Federal Communications Commission Chief Brendan Carr, among others, to find a buyer to put that bandwidth to work.

T-Mobile initially expressed interest in nearly all of EchoStar’s licenses to use the taxpayer-owned frequencies and is still in talks about some of the pieces AT&T didn’t buy, according to the people.

Musk’s bid is for a part of EchoStar’s spectrum that AT&T has been valued at roughly $30 billion by some analysts. It wasn’t clear whether those negotiations are ongoing.

From local shops to global sellers, small businesses remain engines of vital growth. And as they face rising costs, labor shortages, and shifting regulations, they’re leveraging technological breakthroughs, market changes, and the evolving policy landscape to compete on a bigger scale.

Join Semafor on Wednesday, September 10 as we explore how small businesses are embracing their digital potential, tackling competition in 2025, and reaffirming their role as a driving force in the US and global economy.

September 10, 2025 | Washington, DC | Request Invitation

The parents of a 16-year-old who died by suicide are suing OpenAI, claiming its chatbot contributed to the death of their son by, at times, deterring him from seeking help and answering his questions about suicide methods, The New York Times reported. “OpenAI launched its latest model (‘GPT-4o’) with features intentionally designed to foster psychological dependency,” the complaint said.

In a statement to the Times, OpenAI said that while ChatGPT includes safeguards, like referring people to helplines, they “work best in common, short exchanges.” It added: “We’ve learned over time that they can sometimes become less reliable in long interactions where parts of the model’s safety training may degrade.” The company is working to “make ChatGPT more supportive in moments of crisis,” it told the Times.

It’s the second lawsuit blaming an AI chatbot for contributing to a young adult’s death, in addition to an ongoing lawsuit playing out in Florida over a teen’s relationship with a Character.ai chatbot.

The big question for OpenAI is whether it will attempt to use Section 230 of the Communications Decency Act as a defense — which shields platforms from culpability for what users post on them. That framework, however, has been challenged in the AI age, because it’s those companies’ servers providing messages through their chatbots, rather than external users.

CEO Sam Altman has previously said AI companies shouldn’t be relying on that defense. When asked if that law applies to OpenAI’s product in a Senate hearing in 2023, he responded, “I don’t think Section 230 is even the right framework.”

Character.ai’s lawyers attempted to dismiss its case on First Amendment and Section 230 grounds, but the Florida judge wrote its lawyers “fail to articulate why words strung together by an LLM are speech.” While the judge didn’t directly address the Section 230 defense, the ruling is an early signal that courts may be less willing to extend blanket immunity to AI-generated content than they have to social media posts.

US fighter pilots took directions from an AI system for the first time in a test that could drastically change combat tactics, Fox reported. Fighter pilots in action typically communicate with ground support who monitor radar and tell pilots where to fly. During the Air Force and Navy’s test this month, pilots instead consulted with Raft AI’s “air battle manager” technology to confirm their flight path was on track and to receive faster reports of nearby enemy aircraft.

The trial run comes as defense tech companies increasingly build technologies that take humans out of the equation — Anduril and General Atomics have both created unmanned fighter drones that can fly alongside human-driven aircraft. Such developments change how war is fought and won, bolstering the case for technological innovation over sheer manpower. It also changes the pace at which critical decisions are made. Raft AI CEO Shubhi Mishra told Fox decisions that once took minutes only take seconds with the new technology. While that can help pilots intercept threats faster, it also risks removing strategic judgment from the loop.