Microsoft Made $10 Billion From OpenAI in Just Three Months

Morning, {{first name | folks}}! Microsoft’s OpenAI business hit $10 billion in one quarter, NVIDIA brought transcription offline, and Anthropic gave companies new ways to track Claude usage and spending. Plus, Google’s latest cybersecurity model and a fresh wave of attacks on South Korea’s financial sector. Let’s get into it!

Microsoft generated $10 billion from its business with OpenAI in the quarter ending September 30, according to its latest filing. That represents nearly 11% of Microsoft’s $92.3 billion in quarterly revenue, although the company hasn’t broken down exactly where the money came from.

The figure highlights just how significant OpenAI has become to Microsoft’s business. The two companies remain closely linked through cloud infrastructure and commercial agreements, even as OpenAI expands its partnerships with other cloud providers and Microsoft develops its own AI offerings.

NVIDIA has released Nemotron 3.5 Transcribe, an open-weight model that transcribes recordings up to an hour long in 40 languages. It can also translate speech, identify different speakers and add word-level timestamps.

Because it can run locally, developers can build transcription tools without sending audio to a cloud service. The model also supports both verbatim transcripts and cleaned-up text, giving developers more flexibility when building tools for interviews, meetings and multilingual conversations.

Anthropic has rolled out three Claude Enterprise features: reports that break down the work teams produce with Claude and what it costs, in-app surveys to gather employee feedback, and shared monthly budgets for teams. The reports can also flag recurring problems and tasks that could be turned into reusable workflows.

Companies can now look beyond their monthly bill to see which tasks Claude helps with, where employees get stuck and what other teams could learn from their work. All three features are now generally available to Claude Enterprise customers.

Google has introduced Gemini 4 Carbon, a model built for complex coding and cybersecurity work. It scored 88.2% on SEC-bench Pro for finding software vulnerabilities and 71% on CWE-bench for patching them, according to Google.

Google is initially giving access to a select group of cyber defenders before expanding to developers and businesses. The reason is clear: a model that can uncover and exploit software weaknesses needs careful handling, even when its goal is to help security teams fix them.

South Korea’s financial sector is facing another round of cyberattacks. Daishin Securities confirmed that personal data from 38,490 customers was exposed, while Mirae Asset and crypto exchange Bithumb have also reported security incidents.

The attacks come only weeks after similar incidents hit the country’s financial industry. Investigators suspect AI-powered tools were used in earlier attacks, but it’s not yet clear whether the latest breaches are connected.