AI Furnace Newsroom: Anthropic Goes Head to Head with the Pentagon, Block's Workforce Cut, OpenAI Raises $110B, Walmart Praises AI Integration, IBM Stock Shock

Welcome to this week’s AI Furnace Newsroom

In today’s insights we cover:

1. Anthropic Refuses Pentagon’s Final Offer as AI Models Start Becoming Defence Infrastructure

2. Anthropic Accuses Chinese Labs of Stealing Claude Blueprint

3. Block Cuts 4,000 Jobs as it Bets on AI-driven Productivity

4. OpenAI Raises $110B as it Locks in Compute and Global Reach

5. Walmart Praises AI Integration as a Key part of its Latest Earnings Report

6. IBM’s Stock Shock Shows Thanks to Anthropic

Read time: 5 mins

💡 Furnace Insights

Anthropic

Anthropic Refuses Pentagon’s Final Offer as AI Models Start Becoming Defence Infrastructure

Last week, Anthropic decided to walk away from U.S. defense work rather than accept Pentagon contract language that would allow “any lawful use” of Claude without clearly carving out (i) mass domestic surveillance or (ii) fully autonomous weapons. Dario Amodei, CEO of Anthropic, said he could not comply as these two points violated people’s democratic right to maintain privacy and Claude simply isn’t good enough to kill the right person without risk of major error. 

The Pentagon pushed back publicly and threatened to classify Anthropic as a supply-chain risk or be compelled under the Defense Production Act (meaning the government would not work with any company that works with or has affiliations to Claude). However, even amid the standoff after the Pentagon cut ties with Anthropic, reporting says Anthropic tools were still used in connection with U.S. strikes on Iran, suggesting that Claude is already entangled in defense workflows in ways that are hard to unwind quickly. 

OpenAI on the other hand agreed to work with the Pentagon, accepting a framework where “lawful use” decides how AI can be deployed, and that difference hit a nerve online and led to OpenAI unsubscribe rates hitting all time highs from its user base. The New York Times also reports that the Pentagon signed a deal with xAI and is getting close to another with Google. We are entering a time where AI companies stop being tech platforms and start becoming defense infrastructure, and this grey area draws a hard line to balance - who really decides how AI gets used once governments rely on it?

Anthropic

Anthropic Accuses Chinese Labs of Stealing Claude Blueprint

Anthropic claims that three of the top Chinese AI labs carried out a coordinated effort to steal the blueprints of Claude to train their own models. They supposedly used thousands of fraudulent accounts and millions of prompt exchanges to extract Claude’s behavior, essentially copying capability via a method called “distillation” rather than building from scratch. Anthropic published a detailed write-up on how it detected and attributed the activity and is framing the issue as both commercial and national-security relevant.  

This means that “model moats” are getting thinner in practice. If behavior can be siphoned at scale, labs will respond by tightening access (verification, rate limits, anomaly detection), which also raises friction for normal developers. If your product depends on a single model provider, assume policy can change fast after an abuse wave. So it’s worth having fallbacks (multi-model routing, cached responses, or a second vendor) even if you prefer one model today. 

Block

Block Cuts 4,000 Jobs as it Bets on AI-driven Productivity

Jack Dorsey publicly tweeted his internal company announcement that Block will be laying off more than 4,000 staff, shrinking a 10,000-person workforce by close to half, with an emphasis on moving to AI-driven productivity. Markets liked the story: shares jumped sharply after the announcement.

What’s unique in this story is that Block was already reporting strong financial performance, and Dorsey didn’t frame this like a bad quarter. This sends a signal that investors expect more companies to restructure around smaller “core teams” that supervise automation, with the biggest pressure landing on coordination-heavy work (status reporting, routing, internal ops) unless teams can prove they’re directly tied to shipping and revenue. 

OpenAI

OpenAI Raises $110B And Locks In Compute And Global Reach

OpenAI announced a $110B funding round at a $730B pre-money valuation, with major checks from Amazon, NVIDIA, and SoftBank, and expanded infrastructure partnerships alongside the financing.

ChatGPT averages more than 900 million weekly users, with over 9 million paying business customers using AI across daily workflows. Codex has also grown to 1.6 million weekly users as developers increasingly rely on AI to build software faster.

Through this investment and the respective partnerships, OpenAI is buying years of compute and locking in distribution through Amazon, while NVIDIA keeps the whole stack tied to its next-gen chips.

Walmart

Walmart Praises AI Integration as a Key part of its Latest Earnings Report

Walmart’s leadership credited technology and AI for improving customer experience and operations, highlighting reduced friction, better decision-making, and tighter inventory visibility while also pointing to growing automation across distribution. 

This result shows that retail AI winners will be the companies that turn AI into inventory accuracy, faster fulfillment, and fewer out-of-stocks, because that’s where margin lives. Walmart’s edge is scale: when it automates even a slice of routing, replenishment, and merchandising decisions, the compounding effect is huge. Walmart claims that 60% of stores receive some freight from automated distribution centers. And it’s hard for smaller competitors to match without similar data and logistics muscle.

IBM

IBM’s Stock Shock Thanks To Anthropic

IBM shares had their worst day in more than 25 years after Anthropic touted Claude Code’s ability to help modernize COBOL, the language still used across financial and government systems, triggering fears that AI could compress consulting-heavy modernization work. IBM quickly pushed back, arguing the value isn’t COBOL itself but the mainframe platform and its integrated hardware-software stack.  

Markets are now hypersensitive to stories where agents “replace” expensive legacy work. But modernization isn’t just code translation. It is dependencies, compliance, performance guarantees, and decades of integration. AI will absolutely speed up parts of this, yet the winners may be the platforms that package AI into safe, auditable migration paths (including IBM itself). 

📈 AI Venture Deals of the Week

• Flox Intelligence, an AI-powered platform for revenue forecasting and intelligence, raised its $3M Seed.

• SolveAI, an enterprise AI software solution startup, raised $45M in its Series A round.

• Firmable, an AI-powered sales platform, raised its $14M Series A round.

• Guidde, an AI digital adoption and training platform for both humans and AI agents, raised its $50M Series B.

• Sensera Systems, an industrial IoT and AI monitoring platform, raised its $27M Series B.

• Profound, an AI marketing and brand visibility platform, raised its $96M Series C.

• Wayve, an autonomous driving and embodied AI company, raised its $1.2B Series D.

⚒️ New AI Product Launches You Don’t Want to Miss 

• Google released Nano Banana 2, their latest advanced AI image generation and editing model.

• Lightfield, an AI-native CRM that automatically captures and understands customer interactions to organize and drive sales workflows.

• PromptURLs, a tool that instantly turns any AI prompt into a shareable, pre-filled URL for major chatbot platforms.

• IronClaw, a secure AI agent runtime that runs autonomous agents in sandboxed environments while protecting credentials and sensitive data.

• MindStudio, a no-code platform for designing, building, and deploying powerful AI agents and automations.

• WorkOS, a developer platform that provides APIs to quickly add enterprise authentication, user management, and directory services to apps.

Upcoming Events 📅

• AIF Founder Supper Club @ Private Room - Tuesday, March 24

Interested in meeting the Who’s Who of AI? Join the AI Hot 100 Summit in NYC

The AI Hot 100 Summit is back for its third edition on May 6-7th, 2026 in New York City. The only AI Summit built for real connections where AI Visionaries meet Industry Leaders. Join over 500 AI Executive Leaders, Founders and Investors for a jam packed 2 days of enterprise AI lightning talks, innovation showcases, and curated networking.

Get ready to meet and greet speakers and attendees from OpenAI, NVIDIA, Haleon plc, Legora, Anthropic, Cursor, Eleven Labs, Adobe, Walmart, Insight Partners & more.

🗓️ May 6-7, 2026

📍New York City