Happy Thursday! Last week Snowflake reinvented itself as a full AI services platform while Microsoft launched seven AI models, an agentic platform, and new Surface hardware. Anthropic filed for IPO and launched Opus 4.8.

NEW AI PLATFORM LAUNCH

Snowflake Reinvents Itself as Full AI Services Platform

Snowflake used its annual Summit conference in San Francisco to announce a sweeping expansion of AI services, rebranding its flagship products Cortex Code and Snowflake Intelligence as CoCo and CoWork. The moves follow a strong quarterly earnings report showing 33% revenue growth, as CEO Sridhar Ramaswamy positioned the company's data-plus-AI combination as its core competitive advantage. The announcements signal a strategic shift from data analytics warehouse to fully managed AI services provider. New features include expanded Apache Iceberg interoperability, an AI governance layer called Cortex Sense, and deepened partnerships with Anthropic and OpenAI, including a $200 million multiyear deal with the latter.

AI STARTUP IPO

Anthropic Files Confidentially for Massive IPO

Anthropic, the AI lab behind Claude, has filed confidentially with the SEC for an initial public offering. The company, valued at nearly $1 trillion, has not yet set share counts or pricing, with the IPO subject to market conditions. The filing comes days after Anthropic closed a $65 billion Series H round at a $965 billion valuation. The move sets up a potential IPO showdown with rival OpenAI, which is also preparing to go public. Anthropic's revenue run-rate has surged to $47 billion, up from $9 billion at end of 2025, fueled by enterprise growth and its upcoming Mythos model.

NEW GENERATIVE AI LAUNCH

Anthropic Launches Opus 4.8 With Dynamic Workflows Feature

Anthropic released Opus 4.8 on Thursday, just 41 days after Opus 4.7, an unusually fast upgrade cycle likely driven by user disappointment with the previous version and competitive pressure from OpenAI and Google. The model features improved handling of uncertain data, with testers noting it proactively flags input and output issues rather than leaving errors for users to catch. Alongside the model, Anthropic launched Dynamic Workflows in research preview, enabling complex tasks across hundreds of parallel subagents for large-scale operations like codebase migrations. The company also hinted its held-back Mythos model, delayed over cybersecurity concerns, could reach customers within weeks.

NEW GENERATIVE AI LAUNCH

OpenAI Adds Codex Plugins for Nontechnical Users

OpenAI has released new features for its Codex automation tool targeting nontechnical users, who now make up 20% of its 5 million weekly users. The update includes six plugins covering design, finance, sales, and data science, plus a new Sites tool for creating shareable single-page websites. The plugins automate tasks like building web interfaces, generating marketing assets, conducting financial due diligence, and analyzing sales data. The Sites feature and plugin collection are rolling out in limited availability, with more plugins planned for legal, consulting, and corporate finance use cases.

NEW GENERATIVE AI LAUNCH

Microsoft Launches Seven AI Models and Agentic Platform

Microsoft unveiled a major AI expansion at its Build developer conference, including Microsoft IQ, a unified intelligence layer connecting AI agents to organizational data. The platform includes Work IQ, Fabric IQ, Foundry IQ, and new Web IQ, which returns grounding information 2.5 times faster than competitors. The company's Superintelligence Team also released seven new in-house models, led by MAI-Thinking-1, a 35-billion-parameter reasoning model claiming to rival Anthropic's Opus 4.6 on coding tasks. Additional releases include image, transcription, voice, and coding models, plus Scout, a new always-on personal agent built for autonomous task handling.

NEW AI ASSISTANT LAUNCH

Microsoft Launches Scout, Its OpenClaw-Inspired AI Assistant

Microsoft has unveiled Scout, an always-on AI assistant built on the OpenClaw framework and integrated into the Microsoft 365 ecosystem. Users personalize their Scout instance with a custom name and train it over time, with the assistant developing persistent memories and skills that adapt to individual work habits. Available through Microsoft's Frontier program with a GitHub Copilot subscription, Scout operates across desktop and web, connecting to inboxes and calendars. It includes a policy conformance system with audit trails to prevent unsupervised AI misbehavior — a concern raised by earlier OpenClaw incidents. Scout was among several AI products announced at Microsoft's annual Build conference.

NEW AI HARDWARE LAUNCH

Microsoft Unveils Surface RTX Spark AI Dev Box

Microsoft announced the Surface RTX Spark Dev Box, a desktop supercomputer built with Nvidia's RTX Spark processor delivering up to 1 petaflop of AI compute, 128GB unified memory, and support for 120 billion-parameter models. The Windows-based machine comes preconfigured with Visual Studio Code, GitHub Copilot, and full CUDA support, launching later this year in the US. Alongside the hardware, Microsoft introduced Microsoft Execution Containers for sandboxing AI agents within Windows, a native GitHub Copilot desktop app for agentic development, and general availability of Discovery, an enterprise AI platform for scientific research workflows on Azure.

NEW AI HARDWARE LAUNCH

Nvidia Packs Trillion-Parameter AI Into Desktop Workstation

Nvidia has unveiled the DGX Station for Windows, a deskside AI supercomputer built around the GB300 Grace Blackwell Ultra Desktop Superchip. The system delivers 20 petaflops of FP4 performance, up to 748GB of memory, and can run frontier AI models of up to 1 trillion parameters locally. Developed with Microsoft and launching in Q4 2026, the workstation targets enterprise teams currently forced to push heavy AI workloads to Linux-based cloud infrastructure. It supports Nvidia OpenShell, a secure agentic runtime using Windows security primitives to isolate AI agents, and can be clustered with other units for larger workloads. Partners include Dell, HP, and ASUS.

NEW GENERATIVE AI LAUNCH

Perplexity AI Debuts Smart Local-Cloud AI Router

Perplexity AI unveiled a hybrid local-server inference orchestrator at Computex 2026, claiming it is the first system to autonomously decide in real time which AI tasks run on a user's device and which are sent to cloud models. CEO Aravind Srinivas demonstrated the technology alongside Intel CEO Lip-Bu Tan during Intel's keynote. The system uses Intel Core Ultra Series 3 chips to determine what data stays local and what can be shared with cloud-based frontier models. Perplexity says the key innovation is not local AI itself, but the automatic, task-by-task routing decision balancing privacy, accuracy, and cost.

AI STARTUP FUNDING

DeepSeek Nears $7.4bn Raise at $59bn Valuation

Chinese AI firm DeepSeek is close to finalising a $7.4bn funding round led by Tencent and battery giant CATL, with backing from China's state-funded AI investment vehicle. The round values DeepSeek at between $52bn and $59bn, with founder Liang Wenfeng personally contributing $2.94bn. The raise comes amid a global AI funding surge, with Anthropic recently valued at $965bn after its latest private round and filing for an IPO, while OpenAI, valued at $852bn, also plans to go public. DeepSeek rose to prominence after its cost-efficient R1 model disrupted Silicon Valley last year.

Upcoming Data & AI Conferences

Databricks Data + AI Summit 2026

Mon, Jun 15 - Thu, Jun 18, 2026 | San Francisco, CA

Curated with AI by EMC2 AI

Brought to you by Estevan McCalley

Share this issue

Follow Estevan

Like this? Subscribe for the weekly AI roundup.