- AI KATANA
- Posts
- Anthropic’s $3B Run-Rate
Hello!
Today’s briefing dives into the shifting power dynamics across AI, defense, and enterprise tech. Anthropic has hit a $3 billion annualized revenue milestone less than a year after launching Claude 3, leveraging enterprise demand in legal and healthcare sectors. Meanwhile, IBM has laid off 8,000 employees as it pushes deeper into AI automation—particularly within its own HR department. In the UK, a sweeping £1 billion investment will create a Digital Targeting Web aimed at reshaping battlefield decision-making, though talent shortages may slow deployment. Over in China, Huawei’s CloudMatrix is now officially on par with NVIDIA’s most advanced AI hardware, according to CEO Jensen Huang, who labeled the firm a “formidable” rival. On a more human-centered note, Google DeepMind’s SignGemma is making strides toward true conversational equity by translating sign language with remarkable accuracy and speed. Stay sharp—new tools, breakthroughs, and job shifts are redefining the AI frontier daily.
Sliced just for you:
📈 Anthropic’s $3B Run-Rate
🚨 IBM cuts 8,000 jobs amid AI expansion
🎯 UK’s £1 bn Digital Targeting Web
🥊 NVIDIA’s CEO Confirms Huawei’s “CloudMatrix” AI Cluster Now Competes With Grace Blackwell
🤟 SignGemma Speaks Your Signs
Anthropic told partners its Claude product line is tracking at US $250 million-plus monthly revenue—an annualised $3 billion just nine months after launch. Enterprise adoption of Claude 3 Opus for legal research and healthcare summarisation lifted average contract value five-fold, while Amazon integrations added 7,000 new paying developers. The figure still trails OpenAI’s projected $12 billion 2025 topline but shows that the “safety-first” challenger can convert hype into cash. Analysts say the surge strengthens Anthropic’s hand as it lines up a secondary share sale expected to value the firm at $60 billion.
IBM has laid off approximately 8,000 employees—around 3% of its workforce—as part of a strategic shift toward scaling AI-driven operations, with its own HR department the most impacted. The company has been increasingly integrating AI to automate tasks like data management and routine paperwork, replacing about 200 HR roles with AI agents just last month. CEO Arvind Krishna framed the move as a realignment rather than a downsizing, asserting that overall employment at IBM is growing in areas requiring human creativity and strategic insight, such as software development and sales. Despite the cuts, IBM insists that AI is enhancing rather than wholly replacing roles, freeing employees from repetitive tasks to focus on decision-making.
London’s long-awaited Strategic Defence Review puts algorithmic targeting at its centre. More than £1 billion will fund a “Digital Targeting Web” that blends ISR, cyber, space and open-source feeds to cut kill-chain times from minutes to seconds. The review also green-lights a new Cyber & Electromagnetic Command, expands the attack-sub fleet to 12 boats under AUKUS, and sets aside £1.5 billion for troop housing to stem retention woes. Defence analysts say the Web could be Britain’s answer to the U.S. JADC2, but warn success hinges on recruiting 4,000 additional AI engineers the MoD currently lacks.
NVIDIA CEO Jensen Huang has publicly acknowledged that Huawei’s CloudMatrix AI cluster now rivals the performance of NVIDIA’s most advanced Grace Blackwell systems, marking a pivotal moment in the global AI hardware race. In a rare and candid statement, Huang confirmed that Huawei’s Ascend 910C chip is on par with the H200 accelerator, a flagship in NVIDIA’s lineup, despite ongoing US trade restrictions. This admission signals that Huawei has significantly closed the technological gap with Western AI infrastructure, even surpassing NVIDIA in system scalability. Huang emphasized Huawei’s rapid advancement and described the Chinese tech giant as a “formidable” competitor that can no longer be underestimated, especially within the constrained Chinese market where NVIDIA’s influence is diminishing.
Google DeepMind unveiled SignGemma, a foundation model that translates sign-language video into spoken or written sentences with 96 % BLEU accuracy across ASL, BSL and eight regional dialects. The 6-billion-parameter multimodal network is trained on 50,000 hours of ethically-sourced motion-capture data and uses a dual-stream encoder—one for hand-pose trajectories, one for facial cues—to retain linguistic nuance. Early testers at Singapore’s Institute for the Deaf report real-time latency under 120 ms on a Pixel 9. DeepMind says the model will ship in the Gemma open-weights family later this year and power YouTube autotranslation as well as AR-glasses captions. Accessibility advocates hail it as a leap beyond isolated-word recognition toward full conversational equity.
🛠️ AI tools updates
Opera resurrected the Neon brand as a subscription “agentic browser” that can autonomously read, click and create while you sleep. Powered by a sandboxed LLM and a task-graph planner, Neon can research competitors, draft code, book travel and summarise pay-walled PDFs, then deliver a morning briefing. A built-in Operator panel shows every action for audit, tackling the “black-box” trust gap. Early beta access opens next month with enterprise pricing to follow.
💵 Venture Capital updates
New York-based Prepared closed an $80 million Series C led by General Catalyst with a16z joining, valuing the emergency-response platform at $540 million. Prepared’s multimodal AI ingests live caller video, building schematics and sensor feeds to give dispatchers instant floor-plans and triage suggestions, already adopted by 140 U.S. municipalities. Funds will double headcount and expand into Asia-Pacific fire services.
🫡 Meme of the day
⭐️ Generative AI image of the day
Before you go, check out Paraglider’s ‘Thundercloud’ Video Was AI Smoke & Mirrors.
