Amazon discusses OpenAI investment as Nvidia integrates CUDA Tile

Amazon is reportedly in discussions to invest up to $50 billion in OpenAI, potentially expanding its computing power deal and integrating OpenAI's models like ChatGPT into its services. This move is part of a larger funding round that could reach $100 billion, with other investors including Nvidia and SoftBank Group. Meanwhile, Nvidia itself is enhancing GPU programming by integrating its CUDA Tile technology as a backend for OpenAI Triton, aiming to optimize performance for new GPU designs like Blackwell.

In China, Alibaba's semiconductor arm, T-Head, has shipped over 100,000 units of its Zhenwu 810E AI chip, which performs similarly to Nvidia's A100. This positions Alibaba ahead of local competitor Cambricon Technologies and helps reduce reliance on foreign chips due to US export controls. Additionally, Moonshot AI, backed by Alibaba, has upgraded its Kimi K2.5 model, claiming superior video-generation and agentic capabilities compared to rival US AI models, intensifying the competition ahead of DeepSeek's new model release.

The Trump administration's Department of Homeland Security (DHS) is rapidly expanding its use of AI, now exploring over 200 AI tools, a 37% increase since July 2025. Immigration and Customs Enforcement (ICE) is a key driver, adding 25 new AI applications, including identity confirmation with biometrics and fraud detection, with three new products sourced from Palantir. On the educational front, DeVry University plans to embed AI literacy and skill-building into every course by the end of 2026, preparing students for an AI-focused job market and providing faculty with specialized AI training.

Despite rapid advancements, AI still faces significant challenges, as demonstrated by "Humanity's Last Exam," a graduate-level test where leading AI models scored only 50%, highlighting a gap in complex reasoning compared to human intelligence. In industry, Ziff Davis launched its AI-driven RetailMeNot Group to personalize online savings, while Chugai Pharmaceutical, despite removing an AI-developed drug candidate (BRY10) from its pipeline, maintains strong faith in its Malexa AI tool for future antibody projects. Healthcare organizations, as noted by Dr. Kari Yacisin, require robust change management for successful AI adoption to ensure improved care without compromising staff well-being.

Key Takeaways

  • Amazon is considering a significant investment of up to $50 billion in OpenAI, potentially integrating ChatGPT and other models into its services.
  • Nvidia has integrated CUDA Tile technology into OpenAI Triton to enhance GPU programming, requiring CUDA 13.1+ and Blackwell GPUs.
  • Alibaba's T-Head has shipped over 100,000 Zhenwu 810E AI chips, competing with Nvidia's A100 and surpassing local rival Cambricon.
  • Alibaba-backed Moonshot AI upgraded its Kimi K2.5 model, claiming superior video-generation and agentic abilities over US rivals, intensifying China's AI competition ahead of DeepSeek's new model.
  • The Trump DHS is testing over 200 AI tools, a 37% increase since July 2025, with ICE using Palantir's AI products for identity verification and fraud detection.
  • "Humanity's Last Exam," a graduate-level test, showed leading AI models scored only 50%, indicating current limitations in complex human-level reasoning.
  • DeVry University will integrate AI literacy and skill-building into all courses by 2026, preparing students for an AI-focused workforce.
  • Ziff Davis launched its AI-driven RetailMeNot Group, combining coupon sites like RetailMeNot to offer personalized online savings.
  • Chugai Pharmaceutical removed an AI-developed drug, BRY10, from its pipeline but continues to invest in its Malexa AI tool for future drug discovery.
  • Successful AI adoption in healthcare requires strong change management, focusing on clear communication and clinician involvement to prevent burnout and bias.

Alibaba's AI chip Zhenwu 810E ships over 100,000 units

Alibaba's semiconductor arm, T-Head, has shipped over 100,000 units of its advanced AI chip, the Zhenwu 810E. This chip is designed for AI training and inference, and its performance is similar to Nvidia's A100. Alibaba has now shipped more chips than its local competitor, Cambricon Technologies. This move helps Alibaba gain a larger share of China's growing AI chip market and reduces reliance on Nvidia due to US export controls.

Alibaba backed Moonshot AI upgrades Kimi K2.5 model

Moonshot AI, a company supported by Alibaba Group Holding Limited, upgraded its main AI model called Kimi K2.5. The company claims this new version has better video-generation and agentic abilities than rival AI models from the US. This upgrade comes as the AI competition in China heats up, especially before DeepSeek's new model is released. Bloomberg reported this news on January 27.

Trump DHS quickly tests over 200 AI tools

The Trump administration's Department of Homeland Security, or DHS, is quickly testing many new artificial intelligence tools. The DHS now explores over 200 AI use cases, which is a nearly 37% increase since July 2025. Immigration and Customs Enforcement, or ICE, is a main reason for this growth, adding 25 new AI uses. These new uses include processing tips, checking mobile device data, confirming identities with biometrics, and finding false identifications. Three of ICE's new AI products come from Palantir, a tech partner for the US government.

AI struggles with Humanity's Last Exam

Researchers from Arizona State University and the University of California, Riverside created a new test called "Humanity's Last Exam." Nearly 1,000 experts from around the world helped create questions that require graduate-level knowledge in many subjects. They only included questions that leading AI models could not answer correctly when the test was designed. The best AI models scored only 50% on this exam, much lower than their scores on other tests. This exam aims to show what AI still cannot do, highlighting the big difference between human and machine intelligence.

Ziff Davis launches AI powered RetailMeNot Group

Ziff Davis, also known as ZD, launched its new AI-driven RetailMeNot Group. This group combines several coupon and deal websites like RetailMeNot, Offers.com, and BlackFriday.com. The company uses AI tools to make online savings easier and more personal for users. This move aims to improve user experience and provide more value for advertisers, driving growth for Ziff Davis. The company expects AI to boost its digital brands, similar to how CNET and IGN have seen advertising revenue grow.

DeVry University adds AI skills to all courses

DeVry University plans to include AI literacy and skill-building in every course by the end of 2026. This new plan aims to prepare students with the technical skills and smart thinking needed for jobs in an AI-focused world. The university will add new AI courses and credentials, and place AI learning assistants in every class. Elise Awwad, DeVry's President and CEO, believes AI skills will become a basic requirement for many jobs. Faculty members are also getting special AI training to help students learn these important skills.

Amazon considers huge AI investment in OpenAI

Amazon is reportedly discussing a large investment of up to $50 billion in OpenAI, along with expanding its deal to provide computing power. OpenAI is also considering allowing Amazon to use its AI models, like ChatGPT, in Amazon's products and services. This potential investment is part of a larger funding round that could reach $100 billion, with other investors like Nvidia and SoftBank Group. While Amazon is also a partner in a UK program to teach AI skills to 10 million workers, some experts and unions worry about AI leading to job cuts. They argue that AI should help humans focus on creativity and judgment, not replace them.

Chugai removes AI developed drug but keeps faith in AI

Chugai Pharmaceutical, a Japanese company, removed BRY10, an antibody for chronic disease, from its drug development plans. BRY10 was the only drug candidate developed with the help of Chugai's AI tool called Malexa. The company decided to stop its development after reviewing existing data, but this decision was not because of the Malexa technology itself. Chugai still believes strongly in AI and plans to use Malexa for many other antibody projects. The company has an "AI Vision" to use AI in all its business processes to create new value.

NVIDIA improves GPU programming with CUDA Tile for OpenAI Triton

NVIDIA has integrated its CUDA Tile technology as a backend for OpenAI Triton, aiming to improve GPU programming. CUDA Tile helps programmers work with data blocks, called tiles, and automatically manages how tasks run on the GPU. This integration allows Triton's Python language to directly use CUDA Tile, keeping high-level instructions and supporting NVIDIA Tensor Cores. This makes programming easier and helps optimize performance for new GPU designs. Developers need CUDA 13.1 or newer and NVIDIA Blackwell GPUs to use this feature.

Healthcare AI success needs good change management

For healthcare organizations to successfully use artificial intelligence, they need strong change management. Dr. Kari Yacisin, a physician from the University of New Mexico, will discuss this at a HIMSS26 session. She explains that health systems must adopt AI in ways that improve care without harming the trust or well-being of doctors and nurses. It is important to focus on clear communication, good rules, and involving clinicians to prevent burnout or bias. Dr. Yacisin emphasizes that adopting AI is mostly about managing change, not just about the technology itself.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

Alibaba AI chip Zhenwu 810E T-Head AI training AI inference Nvidia Cambricon Technologies China AI market US export controls Moonshot AI Kimi K2.5 AI model Video generation Agentic AI China AI competition DeepSeek Trump administration DHS AI tools AI use cases ICE Biometrics Identity confirmation Palantir Government AI AI limitations Humanity's Last Exam AI performance Human intelligence Machine intelligence AI research Ziff Davis RetailMeNot Group AI-driven commerce Online savings Personalized offers Advertising revenue DeVry University AI literacy AI skills AI education Workforce preparation AI courses AI credentials Amazon OpenAI AI investment ChatGPT Computing power SoftBank Group Job cuts AI ethics Workforce training Chugai Pharmaceutical AI drug development Malexa Pharmaceutical AI Drug discovery GPU programming CUDA Tile OpenAI Triton Tensor Cores AI optimization Healthcare AI Change management Health systems AI adoption Clinician well-being Burnout Bias

Comments

Loading...