Why did OpenAI launch a "red alert"? Should NVIDIA also raise a red flag? A visual representation of AI competition

This week, a major news in the field of artificial intelligence (AI) is that OpenAI CEO Sam Altman announced a "red alert" to all employees on Monday, focusing all resources on optimizing ChatGPT to respond to the fierce competition from Google's Gemini. This strategic adjustment reflects profound changes in the AI competitive landscape and reveals the potential threat of Google's self-developed chip TPU to NVIDIA's chip dominance.

Media reports indicate that OpenAI has decided to postpone the development of other products, including advertising services, health and shopping AI agents, and the personal assistant Pulse, reallocating core resources to improve the daily user experience of ChatGPT. Altman stated that OpenAI still needs to enhance the daily experience of ChatGPT, including improving personalization features, speed, and reliability, as well as expanding the range of questions it can answer.

UBS technology analyst Tim Arcuri pointed out in a recent research report that Google's upcoming next-generation TPU chip Ironwood and its TPU ecosystem are posing a substantial challenge to NVIDIA. NVIDIA's stock performance has significantly lagged behind Google's.

Google User Time Surpasses, ChatGPT Daily Active Users Decline

Market data shows that Google is narrowing the gap with OpenAI on multiple dimensions. According to Sensor Tower data, in November, Gemini's monthly downloads reached 100.8 million, while ChatGPT had 67.8 million.

More notably, users are now spending more chat time on Gemini than on ChatGPT or competitors like Claude.

According to statistics from Deedy Das, in the two weeks since the release of Google Gemini 3, the daily unique active users (seven-day average) of ChatGPT have declined by 6%, showing the direct impact of competitive pressure. Although OpenAI still has over 800 million weekly active users, dominating overall chatbot usage, users are shifting towards Google.

Nick Turley, the head of ChatGPT at OpenAI, posted on social media on Monday evening that search is one of the biggest opportunity areas, with ChatGPT currently accounting for about 10% of global search activity and growing rapidly He also stated that the company's focus is to make ChatGPT stronger, continue to grow, and expand its global reach, while making it more intuitive and personalized.

UBS: Google's TPU Chips Pose a Threat to NVIDIA

Behind the competition in AI models, the battle at the chip level is equally fierce. UBS technology analyst Tim Arcuri pointed out in a research report that advancements in Google's TPU chips are changing the market landscape.

According to Arcuri's analysis, Google first disclosed its latest generation TPU chip, Ironwood, in April this year and officially launched it in November. This chip is optimized for large language models (LLM), mixture of experts (MoE), and advanced reasoning, supporting training, fine-tuning, and inference workloads, contrasting with the narrow customization of previous TPUs.

Ironwood has not yet been submitted to MLCommons' MLPerf v5.1 data center training benchmark tests, but given the increased computational resources, FP8 support, and significantly higher bandwidth memory compared to its predecessors, Arcuri expects its single-chip performance to significantly surpass Trillium.

Arcuri noted that Google's previous generation Trillium chip was specifically optimized for inference workloads and had lower HBM capacity (32GB vs 95GB). In contrast, Ironwood has more computational resources, FP8 support, and significantly increased HBM capacity, with expectations that its single-chip performance will greatly exceed that of Trillium. Ironwood also scales TPU to a domain of up to 9,216 TPUs, far exceeding the 8,960 of v5p and the 256 of Trillium.

Arcuri pointed out that this is precisely why NVIDIA's entire ecosystem is significantly lagging behind Google's, as Google is enjoying a surge in attention brought by its TPU products. Koray Kavukcuoglu, Chief Technology Officer of Google DeepMind, stated that by using Google's self-developed custom chips to train AI models, the company has "significantly improved performance."

UBS believes that while Google may consider expanding the TPU ecosystem over time, any such efforts must limit potential encroachment on Google Cloud Platform (GCP) revenues. From this perspective, Meta and Apple are major candidates for internal TPU deployment, as they have large AI projects supporting internal workloads, substantial internal AI clusters, and relatively low dependence on GCP

OpenAI Faces Multiple Competitive Pressures

The background for OpenAI's red alert this time is the pressure from multiple competitors. Google's newly released Gemini AI model surpassed OpenAI in industry benchmark tests last month, driving the stock price of Google's parent company Alphabet to soar. Last week, Alphabet's stock price rose over 14% in a week, and since the release of Gemini 3 two weeks ago, it has accumulated a rise of over 10% in less than two weeks.

Since the release of the image generator Nano Banana in August, Gemini's user base has continued to climb. Google revealed that monthly active users grew from 450 million in July to 650 million in October.

OpenAI is also facing pressure from Anthropic, which is becoming increasingly popular among enterprise clients. Although OpenAI still has over 800 million weekly active users and dominates overall chatbot usage, users are gradually being attracted to Google.

NVIDIA Responds to TPU Challenge

In the face of the rise of Google's TPU, NVIDIA emphasized its strong relationship with Google Cloud Platform in communications with UBS, pointing out that Google uses both TPU and GPU for Gemini inference workloads.

NVIDIA believes that cloud service providers are unlikely to run TPU in their cloud stacks, as significant workload optimization is required to achieve total cost of ownership (TCO) advantages on application-specific integrated circuits (ASICs). NVIDIA also stated that, so far, its performance advantage relative to peers has not diminished.

Looking ahead to 2026, NVIDIA noted that Anthropic's 1 gigawatt (GW) capacity and HUMAIN's expansion of 600,000 units are incremental to its $500 billion order quantity for 2025-2026, providing potential upside.

NVIDIA's CPX chips are targeting advanced programming applications that require context windows of over 1 million tokens. NVIDIA has not officially disclosed the market size but has previously hinted that context window applications account for about 20% of the inference market.

Altman stated last month that OpenAI's data center projects are committed to a total investment of approximately $1.4 trillion over the next eight years. In other words, OpenAI has $1.4 trillion in committed funding to maintain its industry-leading position.

Overall, OpenAI indeed has reasons to feel anxious, but this turmoil is currently still limited to the company itself. As for whether NVIDIA, the world's most valuable company, is also facing a similar "red alert," the market is still closely watching