NVIDIA Nemotron 3 Ultra now available on Amazon SageMaker JumpStart
On Thursday, June 4, 2026, the AI landscape saw a mix of advancements in infrastructure, funding, and innovation. Today's news is a mixed bag, with updates from NVIDIA, Fivetran, Google Cloud, and more.
NVIDIA's Nemotron 3 Ultra is now available on Amazon SageMaker JumpStart, offering 5x faster inference and 30% lower cost for agentic AI workloads. This frontier reasoning model is built for high-throughput reasoning and long-running agent workflows.
Fivetran published a guide to its pricing, explaining how monthly active rows, syncs, and other factors drive data costs. The company also shared a blog post on minimizing churn with AI, detailing how it used customer data to identify trends and prevent future loss.
On the infrastructure side, Google Cloud announced updates to its Managed Service for Apache Spark clusters, aiming to let users run large-scale analytical and data science workloads with maximum efficiency. Meanwhile, OpenAI and Snowflake are collaborating to bring frontier intelligence directly to enterprise data securely and at scale.
Databricks announced a major update to its Agent Bricks Knowledge Assistant, making it 3x faster with parallel test-time scaling using Instructed-Retriever-1. This update also improves the assistant's quality.
In funding news, Supabase raised a $500 million Series F at a $10 billion pre-money valuation, led by GIC. The company also released Multigres v0.1 alpha, an operating system for Postgres that brings Vitess-grade horizontal scaling and high availability.
NVIDIA's Nemotron 3 Ultra is also supported on vLLM, with Day-0 Support announced for the newly released model. Hugging Face published several blog posts, including one on fine-tuning Nemotron 3.5 ASR for specific languages, domains, or accents, and another on EVA-Bench Data 2.0, which covers 3 domains, 121 tools, and 213 scenarios.
OpenAI introduced a new memory system for ChatGPT, allowing it to better remember preferences and keep context fresh across conversations. The company also shared a case study on how Endava is redesigning software delivery around AI agents, using ChatGPT Enterprise and Codex to accelerate workflows.
On a quieter note from Databricks, the company posted a benchmark for its Agent Bricks Knowledge Assistant. Meanwhile, at Google, the company is continuing to improve its Managed Service for Apache Spark clusters.
And one for the security folks: Starburst published a post on understanding dynamic connection passthrough, a key aspect of managing multiple data sources. MotherDuck announced that its Obsidian Vault can now run SQL, allowing users to freeze results as plain markdown tables or local files.
In other news, xAI released a preview of its grok-imagine-video-1.5 model, available via the xAI API. Helion, a fusion startup, closed a $465 million Series G, pushing its valuation to $15.5 billion.
Supabase's valuation has reached $10.5 billion, driven by demand for its Postgres-based platform. The rise of prompt-to-application tools is reshaping backend infrastructure demand, with Supabase cashing in.
Japan's AI stock surge triggered a $2.5 billion foreign sell-off, with investors dumping Japanese equities amid fears of an echo of the dot-com era. Nvidia, Meta, and Schlumberger top the AIDE Index, a new open-source index that provides a structured methodology for benchmarking AI adoption across the S&P 500.
That's the Daily AI Brief for Thursday, June 4, 2026. More tomorrow.
Sources covered
1 story · 24 sources (4 news, 20 company, 0 voices)