Daily TEA – Chips, Scams, Satellites, and Stethoscopes
Etched, Huawei, Crypto Scams, Starlink, MedGemma
Hello, dear TEA-mates, here’s what you need to know today.
1.🧱 AI Chip Startup Etched Raises $500 Million to Take On Nvidia
AI chip startup Etched has raised about $500 million in a new funding round led by investors including Stripes, valuing the company at roughly $5 billion as it races to challenge Nvidia in the booming AI processor market. The San Jose–based company is developing a specialized chip called Sohu, designed specifically for transformer models, and is working with TSMC’s Emerging Businesses Group on manufacturing. With this latest round, Etched has secured close to $1 billion in total funding, backed by prominent investors such as Peter Thiel and several venture firms betting on alternatives to Nvidia’s dominant accelerators. (Read More)
🫖 TEA For Thought: I saw one of their demos in person and was amazed by these Harvard dropouts—their passion, intelligence, and commitment to building a customized GPU is phenomenal, and it is impressive to see how quickly they are growing; it will be interesting to see how markets respond.
2.🐉 Zhipu AI Trains State-of-the-Art Model Entirely on Huawei Chips
Chinese firm Zhipu AI has trained a new state-of-the-art multimodal image generation model entirely on Huawei’s Ascend Atlas 800T A2 processors, demonstrating that competitive AI systems can be built without access to Western chips. The Beijing-based company used Huawei’s MindSpore framework to handle the full pipeline from data preprocessing to large-scale training, relying on engineering techniques such as dynamic graph multi-level pipelining and multi-stream parallelism to optimize performance. Zhipu was previously added to a U.S. Commerce Department entity list over alleged ties to China’s military, a move that cut it off from Nvidia’s H100 and A100 GPUs and pushed it deeper into China’s domestic chip ecosystem. (Read More)
🫖 TEA For Thought: Zhipu began at Peking University focusing on intelligence work, and given Huawei’s deep alignment with the CCP and PLA, this collaboration underscores the strategic dimension of Chinese AI—raising the question of when the West will fully recognize that the CCP, not the Chinese people, represents a profound threat to humanity.
3.🎣 Chainalysis Warns of Explosive Growth in Crypto Scams and Phishing-as-a-Service
Chainalysis reports that crypto scam inflows surged in 2025, with the average scam payment jumping from $782 in 2024 to $2,764 in 2025, a 253 percent year-over-year increase. Impersonation scams grew by 1,400 percent, as fraudsters increasingly rely on AI tools, sophisticated SMS phishing services, and complex cross-border money-laundering networks to target victims. The report highlights China-based cybercrime operations and multiple “crime-as-a-service” groups selling phishing kits and iMessage/RCS phishing campaigns, showing how professionalized infrastructure is making scams more scalable and harder to detect. (Read More)
🫖 TEA For Thought: As AI becomes more advanced, phishing-as-a-service will only expand—growing more sophisticated, harder to spot, and more difficult to prevent.
4.📡 Starlink Helps Iranians Circumvent Severe Internet Blackout
Starlink is providing complimentary satellite internet service in Iran amid a severe government-imposed communications blackout during nationwide protests, offering one of the few channels for images and videos to reach the outside world. Thousands of Starlink terminals have reportedly been smuggled into the country since 2022, after U.S. authorities allowed technology firms to bypass some sanctions to support Iranians’ access to communication tools. Iranian authorities are attempting to jam Starlink signals and seize terminals, escalating efforts to control information flows even as satellite connectivity helps protesters document events and coordinate under repression. (Read More)
🫖 TEA For Thought: Technologies like Starlink are often underestimated, yet they can shift power toward citizens under dictatorship by restoring information flows—intelligence wins wars, and this is a powerful example.
5.🩺 Google Unveils MedGemma 1.5 and MedASR for Next-Generation Medical AI
Google Research has introduced MedGemma 1.5, an updated open multimodal model for medical image interpretation, and MedASR, a new medical speech-to-text system aimed at improving clinical transcription. MedGemma 1.5 expands support from 2D medical images to high-dimensional data such as 3D CT and MRI volumes and whole-slide histopathology, delivering internal benchmark gains of 3 percentage points for CT and 14 for MRI disease-related findings over the previous version. Google describes MedGemma 1.5 as the first publicly released open multimodal large language model capable of interpreting high-dimensional medical data, while encouraging developers to fine-tune the models on their own datasets for real-world applications. (Read More)
🫖 TEA For Thought: With OpenAI’s health tools, Anthropic’s health efforts, and now Google’s MedGemma and MedASR, general AI models are racing into verticals like healthcare—raising questions about what margins and opportunities remain for specialized vertical players.
Prompt Tip of the Day: The Schedule Manager
"We just covered [Topic]. Create a 'Spaced Repetition Schedule' for me.
Give me 3 review questions to ask myself in 24 hours, 3 days, and 1 week. Format this as a CSV file I can import into my Google Calendar."TEAHEE Moment
Stay sharp, stay informed. See you tomorrow.
If you enjoyed this, follow along for more on social: Twitter/X.