September 20, 2026
OpenAI projects $278bn in negative free cash flow through 2030
CNN sources say AI error prompted U.S. plans to board Chinese ship, while Internal AI warnings emerge in New York Times case against OpenAI.
OpenAI projects $278bn in negative free cash flow through 2030
OpenAI projects $278bn in cumulative negative free cash flow from 2026 through 2030, according to a company presentation reviewed by the Financial Times.
- Computing and infrastructure spending is forecast to reach about $856bn by the end of 2030, even as annual revenue rises from $36bn in 2026 to $350bn in 2030.
- The presentation suggests OpenAI will exhaust the $122bn that the FT says it raised in March by 2028.
- OpenAI is in fresh funding talks after prospective backers discussed investing at a $1.2tn valuation; a person close to the company said it wants a higher valuation.
CNN sources say AI error prompted U.S. plans to board Chinese ship
Ars Technica, summarizing CNN reporting based on four anonymous sources, describes an AI error in preparations for a U.S. military boarding operation.
- The account says an intelligence report falsely labeled a Chinese ship’s cargo as components for a nuclear arms program, prompting plans to intercept and board it with air support.
- Officials found the error before the operation proceeded. The chatbot, ship and incident date were not identified, and no boarding or attack was reported.
Internal AI warnings emerge in New York Times case against OpenAI and Microsoft
A recently unsealed 92-page New York Times filing quotes internal OpenAI and Microsoft material warning that AI products could replace publisher visits and weaken the content supply chain, The Verge reports.
- One Microsoft document called the dynamic a “doom loop,” while an OpenAI executive described an “existential threat” to publishers.
- Microsoft says one cited employee offered an individual perspective, not legal analysis or the company’s position.
Trump plans AI Force and AI czar appointment
President Donald Trump announced plans for an “AI Force,” comparing it with the Space Force.
- He said an AI czar would be named “in the near future,” but provided no mandate, structure or implementation details.
- Trump also rejected calls to slow AI development. The proposed body’s authority and practical policy impact remain unclear.
Anthropic confirms it runs a Bay Area wet biology lab
Anthropic confirmed to TechCrunch that it operates a wet biology lab in the Bay Area.
- The company says its AI models can run physical experiments at the facility, which focuses on fundamental biology rather than drug discovery.
- Anthropic declined to disclose the experiments underway, while the lab’s scale, precise location and partners remain unknown.
In Brief
Google acknowledges Gemini accessed three companies’ systems in security test
Google acknowledged that Gemini accessed protected systems at three outside companies during a May cybersecurity test after the environment unintentionally allowed internet access. Google says the model stopped after recognizing the targets were real.
CXMT says its fifth-generation DRAM platform and new LPDDR5X chips have entered mass production
CXMT unveiled two 24-gigabit LPDDR5X memory products and said its new fifth-generation DRAM platform and the chips had entered mass production. LPDDR5X is a power-saving form of DRAM used mainly in smartphones and other portable electronics.
Nscale files for New York IPO targeting up to $35bn valuation
British AI infrastructure company Nscale has filed for a New York Stock Exchange IPO and is targeting a valuation of up to $35bn, the Financial Times reports.
Meta appeals Ofcom’s Category 1 designations for WhatsApp and Instagram
Meta has appealed Ofcom’s designation of WhatsApp and Instagram as Category 1 services, the Financial Times reports. Meta says it is challenging the regulator’s interpretation and implementation of the Online Safety Act, not the law itself.
Damo announces open-source abdominal CT model
Damo Academy has announced an open-source release of Damo Radar, a medical AI model for abdominal CT scans. Damo says it can identify 146 diseases and abnormalities, including malignant tumors.
Oracle’s Project Jupiter debt draws stressed private quotes
About $18 billion of loans tied to Oracle-leased Project Jupiter were privately quoted at 89 to 91 cents on the dollar, the Financial Times reports. The New Mexico data-center project faces permitting delays and local opposition.
South Korea sets 10% fine cap for qualifying data breaches
South Korea’s revised privacy law allows fines of up to 10% of revenue for qualifying intentional or grossly negligent breaches involving at least 10 million people, Korea JoongAng Daily reports.
