Oracle cuts 21,000 AI-era jobs; Microsoft eyes China's DeepSeek

Oracle eliminated about 21,000 roles — roughly 13 percent of its workforce — in the fiscal year that ended May 31, and its own SEC filing said adopting AI was part of the reason. Over the same stretch, capital spending on AI cloud and data centers jumped 162 percent to $55.7 billion, pushing free cash flow to negative $23.7 billion and total debt to $218.7 billion. The episode shows how aggressively the AI buildout is reshaping even a profitable incumbent — and how much of it is running on borrowed money.

Microsoft told Axios it is exploring a self-hosted, fine-tuned version of China's DeepSeek V4 as a lower-cost engine for its new Copilot Cowork agent, possibly within weeks — an optional model run entirely on Azure, with added safeguards to reduce bias. The driver is surging inference costs, as agents make hundreds of model calls per task and OpenAI and Anthropic pull back from flat-rate pricing. In a Stratechery analysis, Ben Thompson casts the move as part of a broader pull toward China: Microsoft is strongly incentivized to use cheap, capable Chinese models, while memory makers Samsung, SK Hynix, and Micron may regret opening the door to Chinese chipmakers.

Anthropic launched Claude Tag, a beta that embeds Claude directly into Slack channels for Enterprise and Team customers. Colleagues mention @Claude to hand off tasks, and a shared Claude identity follows a channel over time, learning the team's work and — with permission — pulling context from other channels and connected tools. Administrators set exactly what each Claude can access, so a legal-focused assistant cannot bleed into engineering. It is Anthropic's clearest push yet into the enterprise-knowledge market staked out by Microsoft, Glean, and Databricks.

Researchers on Alibaba's Qwen team released Qwen-AgentWorld, a pair of open models — 35 billion and 397 billion parameters — that simulate entire software and tool environments in language, predicting how a world changes in response to an agent's actions. The team uses them two ways: as cheap simulators for training agents with reinforcement learning, and as a pre-training foundation that lifts agent scores across seven benchmarks. Strikingly, agents trained inside the simulated world beat those trained in the real environment alone, hinting that model-generated environments could become central to building capable agents.

India's MoEngage, a customer-engagement platform used by more than 1,350 brands across 75 countries, acquired San Francisco startup Aampe in an all-cash deal reported in the tens of millions of dollars. Aampe's technology assigns a dedicated AI agent to each individual customer, tuning messages to one person's behavior instead of to a broad audience segment. The wager is that per-person agents replace the decades-old practice of marketing to segments — and help MoEngage pull enterprise accounts off Salesforce Marketing Cloud and Adobe Experience Cloud.