AI Tech Daily March 25, 2026 | Rick-Brick
microsoft nvidia openai open-source sora
| Source: Mastodon | Original article
OpenAI announced on March 25 that it will retire Sora, its generative‑video model launched just six months ago with a high‑profile $1 billion partnership with Disney. The decision comes after a modest uptake and mounting competition from specialised tools such as Runway and Adobe Firefly. By pulling the plug, OpenAI signals a strategic retreat from video generation to double‑down on its core language and multimodal models, including the GPT‑5 suite unveiled earlier this month. The shutdown will affect developers who have begun building Sora‑powered workflows, and it may accelerate consolidation in the nascent AI‑video market as startups scramble for the vacated niche.
At the same time, NVIDIA pledged a suite of its GPU‑accelerated networking and inference libraries to the Cloud Native Computing Foundation (CNCF). The donation is intended to create a fully open‑source stack for training and serving large models in cloud‑native environments. For Nordic firms that rely on Kubernetes‑based infrastructure, the move could lower entry barriers and speed up deployment of sophisticated AI services without the need for proprietary licences.
Microsoft revealed a new data‑center efficiency layer that leverages AI to optimise cooling, power distribution and workload placement in real time. The technology promises up to 30 percent energy savings, a figure that aligns with Europe’s tightening sustainability mandates and could make large‑scale AI compute more affordable for enterprises across the region.
What to watch next: OpenAI may repurpose Sora’s research assets for future multimodal offerings, while the CNCF community will soon publish reference implementations of NVIDIA’s contributions, testing their scalability on public clouds. Microsoft plans a limited rollout of its efficiency engine to Azure hyperscale sites later this year, and regulators are expected to scrutinise the environmental claims. The convergence of these three announcements could reshape how Nordic companies build, run and power AI workloads in 2026 and beyond.
Sources
Back to AIPULSEN