Anthropic is pushing expansion on multiple fronts at once: according to The Information and Investing.com, the AI provider is negotiating to lease around 1 gigawatt of data center capacity from Apollo-owned Stream Data Centers, has unveiled Claude Opus 5.5 as a new flagship model with 40 percent lower costs, and is reportedly set to brief the UN Security Council on AI risks alongside OpenAI. At the same time, a hack on OpenAI carried out using a rival Anthropic model shows just how closely intertwined the security concerns of frontier labs have become.
- According to The Information/Investing.com, Anthropic is negotiating to lease 1 gigawatt of data center capacity from Apollo-owned Stream Data Centers.
- Per TechTarget and CoinGape, Claude Opus 5.5 brings 40 percent lower costs while still commanding a premium price—unveiled shortly before a reportedly planned IPO at a $2 trillion valuation.
- According to CBS News, three hackers used a rival Anthropic model to attack OpenAI systems—pointing to security gaps at frontier labs.
- Per reports cited by TradingView, OpenAI and Anthropic are set to brief the UN Security Council on AI risks.
- Salesforce CEO Marc Benioff is backing Anthropic, per TradingView, after Microsoft reportedly cut Salesforce off from OpenAI access—a bet that could be worth 'tens of billions' of dollars.
What does the Stream Data Centers deal mean for Anthropic's infrastructure strategy?
According to Investing.com and The Information, Anthropic is negotiating to lease around 1 gigawatt of capacity from Stream Data Centers, which is owned by investment giant Apollo. Other reports, including one from TradingView, suggest Anthropic is generally seeking greater control over additional data centers. For a company with increasingly compute-intensive models, a gigawatt-scale deal would be a major step toward securing training and inference capacity independent of any single cloud partner.
What security questions does the OpenAI hack involving an Anthropic model raise?
CBS News reports that three individuals compromised OpenAI systems using a rival Anthropic model. According to the report, the incident shows how vulnerable frontier labs remain despite all their safety pledges—regardless of which provider's model is used in an attack. The case is likely to fuel the very debate that OpenAI and Anthropic are already set to bring before the UN Security Council.
What's new about Claude Opus 5.5?
According to TechTarget and CoinGape, Anthropic has released Opus 5.5 as a more powerful model that cuts costs by 40 percent compared to its predecessor, while still remaining in the premium price segment. Per CoinGape, the launch comes ahead of a reportedly planned IPO at a $2 trillion valuation.
Why are OpenAI and Anthropic set to speak before the UN Security Council?
According to reports cited by TradingView, both companies are set to brief the UN Security Council on AI risks. The available sources do not specify further details on the content or exact timing of the session.
There's also plenty of movement on the partner and regulatory side: Salesforce CEO Marc Benioff said, according to TradingView, that Microsoft had cut Salesforce off from OpenAI access, prompting him to bet on Anthropic instead—a wager that could be worth 'tens of billions' of dollars. Meanwhile, according to a report by Westlaw Today, Anthropic and OpenAI are calling on Australia to relax its ban on training AI models.
It could be worth 'tens of billions' of dollars.
Marc Benioff, per TradingView
A report from The Register shows that the overall pace shows little sign of changing: despite repeated calls to slow down, frontier labs—Anthropic included—remain locked in a race for new models, infrastructure, and market share.
