Nvidia spends $20B, robotaxis get smarter & more (December 26, 2025)

Good morning!

The AI end-of-year news cycle is officially unhinged. Mega-acquisitions, self-driving cars getting smarter, ads creeping into chatbots, and researchers proving that one word can break far more than it should.

Let’s get into it.

Want to engage in conversation about AI news? Join the Byte-Sized Reddit Community

🚀 Big Tech Power Moves

🧠 Nvidia Goes Shopping, Again , With $20B

Nvidia is buying AI chip startup Groq for roughly $20 billion, marking its largest acquisition ever. The deal locks in Groq’s ultra-fast inference hardware as competition heats up around real-time AI workloads. This is Nvidia tightening its grip on the full AI stack, from training to inference.

🚕 Waymo Tests Gemini Inside Robotaxis

Waymo is testing Gemini as an in-car AI assistant for its autonomous vehicles. Passengers can ask questions, get explanations, and interact naturally during rides. This is less about convenience and more about making self-driving cars feel trustworthy.

🤝 Coforge Buys Encora in $2.35B Deal

India’s Coforge will acquire US-based Encora in a $2.35 billion transaction. The move expands Coforge’s AI, data, and product engineering footprint globally. Services firms are clearly racing to bulk up before AI reshapes enterprise demand.

🌍 AI, Power, and Control

💬 Sponsored Content Comes to ChatGPT

OpenAI is testing ads and sponsored content inside ChatGPT responses. This signals a shift toward monetization beyond subscriptions. The big question is how far ads can go before trust starts to erode. If not carefully labeled and constrained, sponsored answers risk blurring the line between helpful assistance and paid influence.

🏭 China’s Industrial Output Keeps Rising

China expects industrial output to grow nearly 6 percent in 2025, driven in part by automation and AI adoption. Manufacturing remains a key testing ground for large-scale AI deployment. Efficiency, not flashiness, is the real story here.

🕹️ One Word Can Hijack a Robot Army

Chinese researchers showed how a single malicious command could compromise coordinated robotic systems. It’s a reminder that prompt-level vulnerabilities are not theoretical. AI safety is now a physical-world problem.

🛠️ Tools of the Day

→ NBot – Build autonomous AI agents that handle workflows end to end.
→ Tubeletter – Turn YouTube content into clean newsletters automatically.
→ CrowdSynthetic AI – AI-powered crowd safety and risk prediction platform.

⚡ Quick Hits

→ The year data centers became the main character of AI infrastructure.
→ Can AI fix the operating room, startups are betting yes.
→ UK schools warn about AI misuse in classrooms.
→ BBC explores AI voice cloning risks and regulation.
→ A Zelda puzzle proves AI can plan six moves ahead.

🧩 TLDR

Nvidia just made its biggest bet yet, Waymo is humanizing robotaxis with Gemini, services firms are consolidating fast, and AI risks are spilling into the physical world. Ads are coming to chatbots, data centers are the new battleground, and one bad prompt can break a lot more than a demo.

See you tomorrow,
David