CxO · Weekly · March 7, 2026

war just reached your cloud region

Iranian drones and other projectiles hit AWS facilities in the UAE, causing fires, power cuts, and outages that took EC2 and RDS offline in parts of the Gulf.[1]

Additional drone attacks in Bahrain and the UAE knocked AWS data centers offline during military strikes for the first time.[2] The broader Iran war is already disrupting shipping and air cargo and is flagged as a direct threat to global chip supplies and AI expansion.[3]

EU embassies in Iran have shifted to Starlink for secure communications, underscoring how quickly operators will abandon fragile terrestrial networks once shells start landing.[4]

ai labs as defense contractors, and the crowd walks

OpenAI agreed to deploy its models on U.S. Department of War classified networks, a move Sam Altman later called “opportunistic and sloppy.”[5] The “Cancel ChatGPT” campaign drove a reported 295% spike in ChatGPT uninstalls and visible protests at OpenAI’s HQ, as users objected to surveillance and weapons use.[6]

In parallel, the Pentagon designated Anthropic a supply‑chain risk after Anthropic refused a major contract on ethical grounds, even as Claude was used to help select over 1,000 targets in the Iran conflict via Palantir’s Maven Smart System.[7]

Claude’s app then surged to the top of Apple’s free charts and Anthropic’s revenue run‑rate neared $20B, with 80% of demand coming from non‑U.S. clients positioning it as the “ethical” counter‑party to OpenAI.[8]

ai infra capex roulette

OpenAI closed an unprecedented $110B round backed by Nvidia, Amazon and SoftBank, lifting annualized revenue to about $25B but pairing it with a projected $157B cash burn through 2028 and only $40B cash on hand.[9]

The company is talking about $500B of AI infrastructure spending, even as critics question whether its profitability pathway is realistic and warn that assumptions underpinning new data center builds may be unstable.[10]

Oracle plans to invest “tens of billions” into AI data centers and GPUs while simultaneously cutting thousands of jobs and cancelling a Texas data center expansion over power‑grid concerns.[11]

Nvidia is burning billions on photonics and optics, including a $2B stake in Lumentum and $4B for U.S. photonics manufacturing, while Satya Nadella calls $100B‑scale data center contracts a risky bet on future innovation.[12]

Against that, Goldman Sachs estimates AI added “basically zero” to U.S. growth last year and Apple’s AI servers reportedly sit unused in warehouses due to weak Apple Intelligence uptake.[13]

ai law: copyright, licensed advice, and nationalization shadows

By refusing to hear a test case, the U.S. Supreme Court effectively confirmed that purely AI‑generated art is not copyrightable, putting such works into the public domain and forcing human authorship to be the hook for protection.[14]

Commenters note this extends beyond images to software and other creative outputs, creating a broad gray zone where AI‑heavy works may be hard to defend as proprietary IP.[15]

New York’s bill to ban AI from answering questions in licensed professions and to hold vendors liable for substantive responses would sharply limit use of chatbots in law, medicine, and finance if enacted.[16]

Alaska’s bill on AI sexual imagery and child social media use shows states are willing to legislate downstream uses of generative models, not just training data.[17]

On the national security side, the Department of War’s unprecedented move to label Anthropic a supply‑chain risk, plus threats to invoke the Defense Production Act to seize its tech, sits alongside Palantir’s CEO publicly speculating that AI may be nationalized.[18]

china’s parallel stack under tariff and conflict pressure

China announced an approximately $144B fund to drive tech self‑reliance and is marshalling senior semiconductor executives for a national effort to build a domestic alternative to ASML’s lithography systems.[19]

DeepSeek is tuning its models to run on Huawei and Cambricon chips, and DeepSeek V4 with image and video generation is due next week, signalling a full Chinese AI stack that can bypass U.S. export controls.[20]

Alibaba’s Qwen 3.5 line now includes a 9B model that reportedly outperforms ChatGPT’s o1 while running locally, even as key leaders exit and speculation mounts about a pivot away from open source.[21]

Xiaomi’s humanoid robots are already hitting a 90.2% success rate in autonomous factory settings, making Chinese robotics a live competitive factor rather than a science‑fair project.[22]

In the background, a U.S. judge just ordered more than $130B in tariff refunds even as Treasury gears up for higher tariffs, and the Iran war is flagged as a threat to global chip supply, all of which raises the cost of being deeply tied to Chinese or Gulf‑region production.[23]