Anthropic Puts Claude to Work Inside Slack

Anthropic launches Claude Tag, an agentic coworker embedded in Slack channels, while a global chip selloff raises questions about AI infrastructure valuations and Abu Dhabi's MGX raises close to $50 billion for AI deals.

June 24, 2026 | Reading time: 7 minutes | Issue #193

Lead

Anthropic released Claude Tag on June 23, an agentic AI coworker that lives inside Slack. The beta, available to Claude Team and Enterprise customers, lets administrators grant Claude access to selected channels, tools, data sources, and codebases. Anyone in a channel can tag @Claude with a request, and the agent plans and executes tasks asynchronously, posting results in threads.

The product is multiplayer by design. One shared Claude identity operates in each channel, learning context over time and carrying state across conversations. Administrators can scope memories, set spend limits per channel, and review action logs. Anthropic says 65% of its own product team's code is now created by its internal version of Claude Tag, and the pattern is spreading to support tickets, metrics chase-downs, and bug triage.

Claude Tag is the most direct challenge yet to the idea that frontier AI should be accessed through a chat window. Slack is where enterprise work already happens; by inserting Claude there, Anthropic is competing with workflow tools rather than chatbots. The expansion also reframes Anthropic's relationship with the security debate: the company selling the most sophisticated safety story is now also selling the most deeply embedded office agent.

AI Stock Selloff Tests Infrastructure Valuations

A wave of selling hit AI-linked stocks on June 23, sending the Nasdaq down 2.2% and Asian tech markets lower on June 24. Micron fell more than 13%, Samsung and SK Hynix each dropped about 12%, and the tech-heavy Kospi index fell 10%. NPR reported the selloff reflected mounting doubts over whether AI spending is producing matching returns.

The market had already bid memory and chip stocks to extreme multiples. Micron was up close to 800% over the past year on demand for HBM and datacenter memory. Cerebras, which went public in May, reported first-quarter revenue of $193.4 million, up 92% year over year, but forecast core gross margin contraction to 36%-38% and the stock fell about 10%. The reaction shows investors are no longer rewarding growth alone; they want proof of profitability and durable demand.

The episode is a valuation stress test, not a fundamentals collapse. Data center demand remains strong, and Nvidia, Google, and Microsoft continue to expand capital spending. The question is whether marginal players can justify their multiples if capital costs stay high and returns stay speculative.

Capital Flows

Abu Dhabi's MGX raised close to $50 billion to accelerate AI investments, Bloomberg reported on June 23. The fund pulled capital from regional sovereign wealth funds, global pension funds, and large institutional investors. MGX, launched earlier this year as a state-backed AI investment vehicle, is already deploying the capital.

The scale resets expectations for AI finance. MGX is not a venture fund; it is a sovereign-scale allocator that can write checks large enough to shape infrastructure markets. Its capital will flow into chips, data centers, and AI companies, with a geographic focus that likely favors projects tied to the UAE's domestic compute buildout and G42's international deployments. For startups and sovereigns seeking alternatives to U.S. venture capital, MGX is becoming a primary counterparty.

Menlo Ventures raised $3 billion for AI-focused funds, its largest fundraising to date, Bloomberg also reported. The firm's Anthropic stake is currently worth nearly $14 billion. Menlo's haul and MGX's war chest show that LP capital is still concentrating in AI despite public-market volatility. The private and public markets are sending different signals: public investors are questioning multiples; private allocators are still betting on structural transformation.

Compute Watch

Cerebras delivered its first post-IPO earnings and the market wanted more. Revenue nearly doubled to $193.4 million, net loss narrowed to $14 million from $23.9 million, but the forecast gross margin compression triggered an after-hours drop of more than 8%. The company built its brand on wafer-scale chips for AI training and inference, but its first report as a public company shows it is still in the high-investment phase.

The numbers are a reminder that AI silicon economics are unforgiving. Nvidia captures most of the gross margin in the stack. Challengers like Cerebras, Groq, and Upscale AI can win design wins and headlines, but they face a long path to margin expansion. The June 23 selloff hit the entire chip complex, but it hit speculative names hardest.

Eastern Front

ByteDance unveiled Seedance 2.5 in Beijing, a video-generation model that can produce 30-second 4K clips from a single prompt and accept up to 50 reference materials, up from 12 in Seedance 2.0. CNET reported the announcement on June 23. The improvement is in control: more references mean more consistent characters, styles, and motion across generations.

ByteDance's video models have remained competitive with OpenAI's discontinued Sora and Google's Veo 3. Seedance 2.5 also arrives as reports suggest Qualcomm is in talks to design custom chips for ByteDance, based partly on technology from Alphawave Semi, which Qualcomm acquired in 2025. The combination points to a Chinese tech stack developing its own silicon and media-generation layers in parallel. For ByteDance, the immediate application is TikTok and content tooling. For the broader market, it is another signal that Chinese labs are not slowing their generative media development despite U.S. export controls.

India Lens

India's AI infrastructure buildout is attracting patient, sovereign-scale capital. TechCrunch reported on June 17 that CPP Investments, Canada's largest pension investor, committed up to ₹70 billion (about $741 million) to Indian data center operator CtrlS. The deal includes ₹40 billion for an 8.2% equity stake and up to ₹30 billion for a joint venture to develop hyperscale campuses, with CPP owning 48%.

CtrlS operates more than 15 data centers across India and has become a local partner for global cloud demand. The investment follows the May G42-Cerebras deal to deploy a UAE-financed AI supercomputer in India. Both moves reflect the same strategy: India wants AI capacity governed under Indian rules, and sovereign investors want exposure to the fastest-growing large digital market without taking full construction risk. The challenge is execution. Commitments are rising, but power, land, and chip availability remain binding constraints.

Europe

Mistral released OCR 4, a document-intelligence model with bounding boxes, block classification, and inline confidence scores for 170 languages. The model runs in a single container for self-hosted deployments and targets enterprise search, RAG, and retrieval pipelines. Mistral claims independent annotators preferred OCR 4 over rival systems with win rates averaging 72%, and the model scored 85.20 on OlmOCRBench.

The release is a continuation of Mistral's enterprise-first strategy: smaller, focused models that can be deployed on customer infrastructure rather than rented through an API. It is also a European answer to a market dominated by U.S. document-AI players. OCR is not the headline frontier, but it is a high-volume enterprise workflow where deployment location, language coverage, and data residency matter. Mistral is betting those concerns will keep European buyers from defaulting to American cloud providers.

The View

Three threads pulled tight on June 23. Anthropic is turning Claude into a persistent colleague inside the communication layer where work already lives. Abu Dhabi's MGX and Menlo Ventures are pouring tens of billions into AI infrastructure and startups. And chip stocks sold off because investors doubt whether all that spending will convert into returns quickly enough. The tension is between deployment and digestion. AI systems are being embedded deeper into daily workflows and national strategies, but financial markets are starting to demand evidence that the spending produces cash flow, not just capacity. The likely outcome is not a retreat but a reordering: capital will favor the companies that can show usage, margin, and lock-in, while speculative names face a harsher funding environment.

The Miss

NVIDIA launched the BioNeMo Agent Toolkit for scientific discovery on June 23. The toolkit gives AI agents domain-specific tools for protein structure prediction, molecular docking, generative chemistry, and genomic analysis, with adoption from Dassault Systèmes, Databricks, Lilly, Schrödinger, Snowflake, and the UW Medicine Institute for Protein Design. Anthropic and OpenAI are also integrating it. Jensen Huang described frontier models as the brain and BioNeMo as the scientific toolbox.

The announcement was overshadowed by the chip selloff, but it matters more in the long run. AI's most durable value may come not from content generation but from accelerating science. BioNeMo is an early infrastructure layer for that transition. If agentic AI succeeds in biology and chemistry, the economic impact will exceed anything seen in chat interfaces.

Pull Quotes

"Today, 65% of our product team's code is created by our internal version of Claude Tag." — Anthropic, introducing Claude Tag

"The market just continues to oscillate between 'AI is going to be great' and 'this is all one big bubble.'" — Gil Luria, D.A. Davidson, NPR

"Frontier models are the brains. BioNeMo is the scientific toolbox." — Jensen Huang, NVIDIA

"Dispatch is the brain of every home service business." — George Eliadis, Probook, Fortune

Reads & Links

Anthropic introduces Claude Tag — Anthropic on the agentic Slack coworker now in beta for enterprise customers. https://www.anthropic.com/news/introducing-claude-tag

AI stock sell-off shakes Wall Street to Asia — The Guardian on the June 23 rout and bubble debate. https://www.theguardian.com/business/2026/jun/23/ai-stocks-sell-off-us-markets

Abu Dhabi's MGX raises about $50 billion for AI deals — Bloomberg on the sovereign AI fund's capital raise. https://www.bloomberg.com/news/articles/2026-06-23/abu-dhabi-s-mgx-raises-about-50-billion-to-accelerate-ai-deals

Cerebras falls after first earnings report since IPO — CNBC on revenue growth, margin forecast, and post-IPO reaction. https://www.cnbc.com/2026/06/23/cerebras-cbrs-q1-earnings-report-2026.html

ByteDance's Seedance 2.5 makes 30-second 4K videos from a single prompt — CNET on the Chinese video model's expanded reference control. https://www.cnet.com/tech/services-and-software/bytedance-introduces-new-seedance-2-5-video-model/

Canadian pension giant joins race to fund India's AI data center boom — TechCrunch on CPP Investments' up to $741 million CtrlS commitment. https://techcrunch.com/2026/06/17/canadian-pension-giant-joins-race-to-fund-indias-ai-fueled-data-center-boom/

Mistral releases OCR 4 for document intelligence — Mistral on the self-hosted OCR model covering 170 languages. https://mistral.ai/news/ocr-4/

NVIDIA BioNeMo Agent Toolkit — NVIDIA on domain-specific agent tools for scientific discovery. https://nvidianews.nvidia.com/news/nvidia-launches-bionemo-agent-toolkit-giving-ai-agents-the-tools-to-accelerate-scientific-discovery

Capital is still pouring into AI, but markets now want proof the spending buys returns.