Anthropic Doubles AI Model Development Speed

Key Takeaways

• Anthropic's AI model development speed has doubled compared to a year ago, confirmed by independent group METR
• Broadcom projects AI chip revenue of $58 billion in fiscal 2026, up 186 percent from $20 billion in fiscal 2025
• Global markets lost $3.5 trillion as AI stocks dropped up to 20% following U.S. midterm elections
• Google's TPUs deliver 15 to 30 times higher performance than older chips for specific AI tasks
• Airbnb struck a $7.1 million deal with an AI lab founded by CEO Brian Chesky and CTO Ahmad Al-Dahle
• Broadcom's long-term growth faces headwinds from high interest rates and potential data center regulations
• Anthropic plans to release its own analysis of the METR report findings soon
• AI stocks have begun recovering after the sharp post-election sell-off
• Airbnb saved $26 million in professional service costs due to increased AI adoption
• METR was given unprecedented access to Anthropic's research and development data

Anthropic speeds up AI development despite calls to slow down

Anthropic is developing AI systems about twice as fast as a year ago, according to a report from the nonprofit Model Evaluation and Threat Research. CEO Dario Amodei called for the industry to slow down in a September essay. Co-founder Sam McCandlish said the findings match his own views and the company plans to release its own analysis soon. The report found that AI itself is helping speed up Anthropic's research.

Outside research group confirms Anthropic's fast AI progress

An independent monitoring group called METR reviewed Anthropic's internal data and found the company is improving its AI models about twice as quickly as last year. METR was given unprecedented access to Anthropic's research and development information. The report noted that the public version of Anthropic's risk report understates the level of acceleration. Anthropic's co-founder Sam McCandlish acknowledged the findings but said the company redacted details for commercial and safety reasons.

Broadcom's AI chip revenue set to surge to $58 billion this year

Broadcom expects its AI chip revenue to jump 186 percent to $58 billion in fiscal 2026, making up 55 percent of its total projected revenue. The company makes customized AI accelerators called XPUs for hyperscalers, unlike Nvidia which makes general-purpose GPUs. Broadcom projects AI chip revenue could reach $230 billion by fiscal 2028. However, high interest rates and new data center regulations could pose challenges for the company.

Broadcom AI growth faces macro and regulatory headwinds

Broadcom's AI chip revenue grew 65 percent to $20 billion in fiscal 2025 and is expected to reach $58 billion in fiscal 2026. The company bundles its XPUs with Ethernet switches, optical interconnect components, and infrastructure software. Texas governor-elect Gina Hinojosa plans to suspend new data center construction, and other states may follow. High interest rates could also force AI companies to cut spending, pressuring Broadcom's long-term growth estimates.

Kogi State backs AI and fact-checking training for communicators

More than 50 public communicators in Kogi State, Nigeria, received training in artificial intelligence, fact-checking, and crisis communication in Lokoja. The two-day workshop was organized by Image Merchants Promotion Limited with support from NITDA and CCC. Governor Ahmed Usman Ododo urged professionals to use AI responsibly and strengthen verification practices. IMPR Managing Director Yushau Shuaib said the training addresses the growing threat of misinformation through deepfakes and fabricated content.

General Counsel Must Track Data Locations Before Approving AI Tools

Before approving AI tools, general counsel need to know exactly where company data is stored. Rehan Hasan of Omnus Law explains that AI models perform better with more data, but that also increases risk. Network segmentation physically enforces data limits, unlike written policies. Boards must ask where data lives and what systems can reach it. Law firms should keep client files on infrastructure physically separated from AI systems.

Global Markets Lose $3.5 Trillion as AI Stocks Drop After U.S. Midterms

Global markets lost $3.5 trillion in value as AI stocks dropped sharply after the U.S. midterm elections. The Democratic Party's election wins raised fears of AI infrastructure delays. AI stocks fell up to 20% in one week. European markets dropped over 10% and Asian markets fell over 15%. The IMF warned the global economy faces recession risk due to the sell-off.

Worldwide AI Stock Sell-Off Halts but U.S. Stocks Still Face Pressure

The worldwide sell-off for AI stocks has stopped, but other U.S. stocks continue to struggle. The broader market remains under pressure even as the sharp decline in AI-related shares comes to a pause. Investors are watching closely for signs of stability in the coming days.

AI Stocks Begin Climbing After Trillions in Market Value Wiped Out

Shares in artificial intelligence and technology firms started to recover after a massive sell-off on Thursday wiped trillions of dollars from stock values. BBC experts answered audience questions about the market developments. The rebound followed the steep losses that had shaken investors earlier in the week.

TPU vs GPU: Choosing the Right AI Chip for Training and Inference

Google's Tensor Processing Unit (TPU) delivers 15 to 30 times higher performance than older chips for certain AI tasks. TPUs are optimized for tensor and matrix math in neural networks on Google Cloud. GPUs, made by Nvidia and AMD, offer broader compatibility with open-source models and CUDA software. The right choice depends on your framework, deployment setup, and workload needs.

AI stocks fall after midterm elections as regulation fears grow

AI stocks dropped after the midterm elections because Democrats won across the country. Professor Larry Straub at Newman University said investors fear Congress will regulate AI and data centers. David Cochran from the Newman University AI Institute for the Common Good said the drop shows investor fear but not a bubble popping. Straub compared the situation to the dot-com bubble burst and told people to wait for the market to recover.

Airbnb reveals AI lab deal with CEO Brian Chesky

Airbnb disclosed a $7.1 million transaction with an AI lab founded and controlled by CEO Brian Chesky and CTO Ahmad Al-Dahle. The deal included an investment option and a license for Airbnb to use some of the lab's intellectual property. A special committee of independent directors reviewed and approved the arrangement. Airbnb also reported a $26 million reduction in professional service costs due to fewer agent contacts as AI use increased, though total operations and support expenses still rose 8 percent to $393 million.

Sources

• Anthropic is reportedly rapidly accelerating AI development amid slowdown calls
• Anthropic is rapidly accelerating AI development amid slowdown calls, outside research group says
• Broadcom's AI Revenue Is Exploding. Here's What Investors Should Watch Next.
• Broadcom's AI Revenue Is Exploding. Here's What Investors Should Watch Next.
• Gov Ododo Backs IMPR, NITDA on AI, Crisis Communication and Fact-Checking Training
• Before Approving AI, General Counsel Must Know Where Data Lives
• Global Markets Lose $3.5T as AI Stocks Plummet Post-U.S. Midterms
• The worldwide sell-off for AI stocks halts, but other US stocks are still struggling
• BBC Audio | Outside Source | AI stocks climb after trillions wiped from market values
• TPU vs GPU: Which AI Chip Wins for Training and Inference
• AI stocks drop after midterm elections
• Airbnb Discloses AI Lab Deal With Investment and Licensing Rights