Nvidia's $40 billion AI equity spree + Nemotron Elastic checkpoint ships

Nvidia has now committed more than $40 billion to equity investments in AI companies in 2026, including a single $30 billion stake in OpenAI, up to $3.2 billion in glassmaker Corning, and up to $2.1 billion in data-center operator IREN. Wedbush analyst Matthew Bryson described the pattern as "squarely circular" — money cycling between chip vendor, model customer, and infrastructure provider. The chipmaker has also closed roughly two dozen private startup rounds plus 67 venture deals across 2025, intensifying scrutiny of how concentrated the AI capital stack has become around a single supplier.

NVIDIA Research published Nemotron Elastic, a post-training method that embeds 30, 23, and 12 billion-parameter nested reasoning models inside a single 30 billion-parameter parent — extractable via zero-shot slicing without further fine-tuning. The recipe achieves a 360-times token reduction over pretraining from scratch, and the 30 billion checkpoint compresses to 18.7 gigabytes under NVFP4 quantization. The 23-to-30 billion configuration advances the accuracy-and-latency Pareto frontier with up to 16 percent higher accuracy and 1.9-times lower latency than the default Nemotron Nano v3 budget control. All three precision variants (BF16, FP8, NVFP4) are available on Hugging Face under nvidia/NVIDIA-Nemotron-Labs-3-Elastic-30B-A3B.

Google updated its Gemini API File Search tool on May 5 with three additions for retrieval-augmented generation (RAG) workflows: multimodal indexing of images and text together using the Gemini Embedding 2 model, custom key-value metadata labels (such as department: Legal) for query-time filtering, and page-level citations that link readers to the exact source location in indexed documents. The update targets both prototypes and production-scale RAG applications, with example use cases in creative agencies, scientific research, and engineering. No pricing change was announced.

Microsoft Research released DELEGATE-52, a benchmark simulating extended document-editing workflows across 52 professional domains including coding, crystallography, and music notation. Across 19 large language models tested, the strongest frontier systems — Gemini 3.1 Pro, Claude 4.6 Opus, and GPT 5.4 — corrupted an average of 25 percent of document content by the end of long sessions, with errors silently accumulating rather than failing loudly. Adding agentic tool use did not improve results, and degradation worsened with larger documents, longer interactions, and distractor files. Authors: Philippe Laban, Tobias Schnabel, and Jennifer Neville.

The Council of the EU and the European Parliament reached provisional agreement on May 7 to delay compliance deadlines for high-risk AI systems under Annex III to December 2, 2027, and for AI in regulated products under Annex I to August 2, 2028. The deal also bans so-called "nudifier" tools and AI-generated child sexual abuse material starting December 2026, while postponing watermarking and synthetic-media disclosure obligations to the same date. The simplification was driven by industry lobbying and a final push from Germany, France, and Italy ahead of the original August 2026 enforcement deadline.