This week, a major development in the artificial intelligence (AI) sector emerged as OpenAI CEO Sam Altman announced a "red alert" to employees, signaling a full-scale resource reallocation toward optimizing ChatGPT to counter fierce competition from Google's Gemini. This strategic shift reflects profound changes in the AI landscape and highlights the potential threat posed by Google's in-house TPU chips to NVIDIA's dominance in the chip market.

Reports indicate that OpenAI has postponed development of other products, including AI-driven advertising, health, shopping agents, and its personal assistant Pulse, to focus on enhancing ChatGPT's daily usability. Altman emphasized the need to improve ChatGPT's personalized features, speed, reliability, and question-answering capabilities.

UBS tech analyst Tim Arcuri noted in a recent report that Google's upcoming Ironwood TPU chips and its TPU ecosystem present a tangible challenge to NVIDIA. NVIDIA's stock performance has notably lagged behind Google's.

**Google Gains Ground as ChatGPT's Daily Active Users Decline** Market data reveals Google is closing the gap with OpenAI across multiple fronts. Sensor Tower figures show Gemini garnered 100.8 million monthly downloads in November, compared to ChatGPT's 67.8 million. More strikingly, users now spend more time on Gemini than on ChatGPT or rival chatbots like Claude.

Per Deedy Das, ChatGPT's daily active users (7-day average) dropped 6% in the two weeks following Gemini 3's launch, signaling direct competitive pressure. While OpenAI still dominates with over 800 million weekly active users, Google is steadily attracting its user base.

Nick Turley, OpenAI's ChatGPT lead, stated on social media that search remains a key growth area, with ChatGPT accounting for ~10% of global search activity and expanding rapidly. He added that the focus is on making ChatGPT more powerful, accessible, and personalized.

**UBS: Google's TPU Chips Threaten NVIDIA** Behind the AI model race lies an equally intense chip battle. UBS's Arcuri highlighted that Google's TPU advancements are reshaping the market. Ironwood, unveiled in April and launched in November, is optimized for large language models (LLMs), mixture-of-experts (MoE), and advanced reasoning—marking a departure from earlier narrowly customized TPUs.

Though Ironwood hasn't yet undergone MLPerf v5.1 benchmarking, Arcuri projects its single-chip performance will significantly surpass Trillium, thanks to enhanced compute resources, FP8 support, and high-bandwidth memory (HBM). Ironwood also scales TPU clusters up to 9,216 units, far exceeding Trillium's 256.

Arcuri attributes NVIDIA's ecosystem underperformance to Google's TPU-driven momentum. Google DeepMind CTO Koray Kavukcuoglu noted that custom chips have "dramatically improved performance" for AI training.

UBS suggests Google may expand its TPU ecosystem cautiously to avoid cannibalizing Google Cloud Platform (GCP) revenue. Meta and Apple are seen as potential adopters, given their large-scale AI projects and minimal GCP reliance.

**OpenAI Faces Mounting Competition** OpenAI's "red alert" comes amid pressure from rivals. Google's Gemini AI recently outperformed OpenAI in benchmarks, propelling Alphabet's stock up 14% last week and 10% since Gemini 3's debut two weeks ago. Gemini's user base grew from 450 million in July to 650 million in October, buoyed by its Nano Banana image generator launch in August.

Anthropic is also gaining traction among enterprise clients, further pressuring OpenAI despite its 800M+ weekly active users.

**NVIDIA Responds to TPU Challenge** NVIDIA, in discussions with UBS, underscored its strong ties with GCP, noting Google uses both TPUs and GPUs for Gemini inference workloads. It argued cloud providers are unlikely to adopt TPUs widely due to the extensive optimization needed for ASIC-based total cost of ownership (TCO) advantages. NVIDIA maintains its performance lead remains intact.

Looking to 2026, NVIDIA cited Anthropic's 1GW capacity and HUMAIN's 600K-unit expansion as incremental growth drivers beyond its $500B 2025–2026 order book. Its CPX chips target advanced programming applications requiring 1M+ token context windows, potentially addressing ~20% of the inference market.

Altman recently revealed OpenAI's $1.4 trillion committed investment in data centers over eight years to sustain its industry leadership.

While OpenAI has reason for concern, the upheaval remains internal for now. Whether NVIDIA—the world's most valuable company—faces a similar "red alert" is under close market watch.