Marissa Mayer raised $8 million for her new AI startup. OpenAI raised $11 billion. That gap tells the real story—and so does the $20 million she burned at Sunshine, which managed just 1,000 downloads across multiple products over seven years.
Astra AI just won Slovenia's Startup of the Year. The numbers look great: 170,000 users, expansion into Germany, rave reviews. But their claimed 50 billion monthly tokens raises a question no one's asking: can a €24/month subscription cover those API bills?
Zhipu AI's GLM-4.7 benchmark chart excludes its strongest competitor. The data table tells a different story. But the real signal isn't the missing bar—it's a €30 annual subscription designed to snap into tools Western labs built.
Google revealed Ironwood, its seventh-generation AI chip, designed to run AI applications faster and more efficiently. The new processor aims to handle the heavy lifting behind AI chatbots and similar tools.
Each Ironwood chip packs 192GB of memory and can work in massive clusters of up to 9,216 chips. Google says it doubles the performance-per-watt compared to its previous chip, Trillium.
The chip specializes in "inference" - running existing AI models rather than training new ones. This focus matters as companies rush to deploy AI applications at scale.
Google built Ironwood specifically for cloud customers who need to run large language models efficiently. The chip includes a specialized core for recommendation systems and ranking tasks.
The launch highlights Google's push to compete with Nvidia, which dominates the AI chip market. Unlike Nvidia's products, Google's chips are only available through its cloud service.
Google plans to integrate Ironwood into its AI Hypercomputer system later this year, though it hasn't named the manufacturer producing the chips.
Why this matters:
Marcus Schuler
Astra AI just won Slovenia's Startup of the Year. The numbers look great: 170,000 users, expansion into Germany, rave reviews. But their claimed 50 billion monthly tokens raises a question no one's asking: can a €24/month subscription cover those API bills?
Zhipu AI's GLM-4.7 benchmark chart excludes its strongest competitor. The data table tells a different story. But the real signal isn't the missing bar—it's a €30 annual subscription designed to snap into tools Western labs built.
Apple buried RDMA over Thunderbolt in a beta update. Now four Mac Studios can run trillion-parameter AI models at conversational speed. The cost: $50,000. Equivalent NVIDIA hardware: $780,000. No press release. No keynote. Just a checkbox in recovery mode.
Cursor just paid a premium for a code review startup to fix what it calls the new bottleneck in software development. But a July 2025 study using Cursor's own tools found AI made experienced developers 19% slower. The math gets uncomfortable from there.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
