OpenAI releases new ChatGPT model after White House delay
OpenAI has released its latest ChatGPT model after the launch was held up by White House cybersecurity concerns. The new model marks a significant upgrade focused on general performance. The company also released a separate tool designed specifically for workplace productivity, signaling a push into enterprise AI. Meanwhile, ChatGPT is set to challenge Google by integrating into the browser via a Chrome feature.
OpenAI is shutting down its Atlas browser project but continues to expand its ambitions in the browser space. The company is pushing forward with a new AI tool that can work both autonomously and collaboratively with users. OpenAI's copyright battle with news organizations may have taken a critical turn, with one analyst calling it a potential fatal misstep.
Nvidia is becoming a victim of the high-performance computing market it helped create, as hardware becomes too expensive for many buyers. In a quirky response, the company launched a set of commemorative trading cards featuring iconic moments from its history. The move highlights the growing affordability crisis in the GPU market, where demand far outstrips supply at current price levels.
Chinese memory chipmaker CXMT is set to launch a $4.3 billion IPO, which would rank as the second-largest on Shanghai's Star Market behind SMIC. Meanwhile, Zhipu AI and chipmaker Iluvatar CoreX are pursuing Hong Kong listings, seeking to raise a combined HK$38.5 billion. China's PCB manufacturers are also investing hundreds of millions in high-end plants to support AI-driven demand, with more than 20 companies announcing aggressive expansion.
China is making a strategic shift by allowing its AI companies to purchase Nvidia's H200 chips, a move analysts say is intended to temporarily ease the AI training bottleneck. However, Beijing remains committed to long-term tech self-sufficiency. A 'token economy' is emerging as AI usage surges across China, and domestic AI labs like DeepSeek and Zhipu are increasingly looking beyond models to develop their own chip capabilities.