Top AI Breakthroughs and Industry Moves Including Anthropic’s $30B Revenue and Neural Computer Advance - April 12, 2026

Key points

• •Anthropic is rapidly expanding its AI market presence, with projected annual revenues reaching $30 billion by 2026. Simultaneously, the company faces intense regulatory scrutiny in the UK over its latest AI model, which it deemed too powerful to release publicly due to safety concerns. This highlights the dual pressures of commercial success and ethical responsibility within leading AI firms.[Google News - AI][Google News - AI][Google News - AI]
• •Meta AI and KAUST have pioneered a new Neural Computer architecture that integrates computation, memory, and I/O into one learned neural model. This research represents a significant hardware advancement, potentially redefining AI system design by enabling more efficient neural computation and hardware resource usage.[Marktechpost]
• •Cisco is set to acquire AI cybersecurity startup Astrix for up to $350 million, underlining the growing emphasis on securing AI systems against threats. This acquisition is part of a broader trend of consolidation in the rapidly evolving AI security market.[Google News - AI]
• •NVIDIA launched the MiniMax M2.7 update to its platform, improving scalable agentic workflows tailored for complex AI applications. This release underscores NVIDIA's leadership in providing tools that enable efficient, agent-based AI deployments across diverse scenarios.[Google News - AI]
• •OpenAI introduced a $100 ChatGPT Pro subscription plan featuring defined usage limits, prompting internal clarifications to address user confusion. This reflects continued efforts to monetize AI products while managing customer expectations around service boundaries.[The Decoder]
• •Natter secured $23 million in funding to expand its enterprise AI conversation intelligence platform globally, indicating strong investor confidence in AI-driven business communications and analytics solutions.[Google News - AI]
• •Liquid AI released LFM2.5-VL-450M, a 450-million-parameter vision-language model that supports bounding box prediction, multilingual instruction following, and rapid edge inference under 250 milliseconds. This model demonstrates advances in efficient, multimodal AI capable of real-time applications across languages.[Marktechpost]
• •AI technologies now autonomously designing and conducting thousands of lab experiments signal a breakthrough in automating scientific discovery. This capability raises questions about human preparedness and oversight needed for safe integration of such autonomous AI in research sectors.[Google News - AI]

Relevant articles