Material developments reported November 10–12, 2026; quiet sectors omitted.
Compute & Platforms: NVIDIA names an interim CEO
NVIDIA announced on November 12 that Jensen Huang had resigned as CEO and board member, effective immediately. CFO Colette Kress will serve as interim CEO while the board searches for a permanent successor. The company said it expects operations, customer commitments and its product roadmap to continue without interruption. Economically, the transition matters because NVIDIA is a pivotal supplier in AI compute: uncertainty over leadership, capital allocation or execution could affect customer and investor assessments of a market built around its hardware and ecosystem. The announcement does not itself establish a change to the roadmap or customer commitments. ([nvidianews.nvidia.com](https://nvidianews.nvidia.com/news/nvidia-announces-ceo-transition/))
Infrastructure, Power, Cooling & Networking: two inference-capacity moves
Torus: On November 11, Torus launched Foundry, a modular system combining power, storage, cooling, networking and compute, and announced $500 million in growth capital. The company says the units are designed for deployment at existing sites using available grid capacity, with 1–20 MW per site. If the model proves deployable at its stated pace and cost, it could help operators bring inference capacity online without waiting for conventional campus construction—an economic advantage where power access and time-to-capacity constrain revenue. These are company-reported specifications and plans, not independently verified deployment outcomes. ([torus.co](https://www.torus.co/news/torus-launches-modular-ai-infrastructure-foundry))
Rackspace and Together AI: The companies announced a multiyear agreement under which Rackspace expects to operate dedicated NVIDIA GB300 NVL72 systems, with Spectrum-X networking, liquid cooling and related infrastructure for Together AI inference workloads. Deployment is planned for 2027 and remains conditional, including on Rackspace securing GPU financing. The deal signals a route for AI platforms to obtain dedicated capacity through an infrastructure operator rather than owning and operating the full stack themselves. Commercial impact depends on financing, delivery and customer demand. ([ir.rackspace.com](https://ir.rackspace.com/node/13806/pdf))
Cyber: Railway quantifies AI-key theft and fraudulent usage
In an incident report published November 10, cloud platform Railway gave further details about an October 20–21 intrusion. Railway said attackers could have decrypted environment variables in at most about 100,000 workspaces, an upper bound rather than a confirmed count of stolen secrets; it estimated unauthorized AI-provider usage at roughly $740,000. Railway says the exploited vulnerabilities have been remediated and it is working with providers on refunds. Economically, the incident shows how stolen API credentials can convert a platform compromise into direct compute costs, customer remediation work and potential trust damage for services that store AI keys. ([blog.railway.com](https://blog.railway.com/p/incident-report-october-security-incident))
Company launch specifications, deployment schedules and strategic claims are not proof of realized capacity or economics. Rackspace’s planned deployment is conditional, including on financing. Railway’s roughly 100,000-workspace figure is a maximum potential exposure, not a confirmed count of compromised workspaces; the underlying intrusion occurred October 20–21, before this update’s reporting window.