Why did OpenAI launch a "Red Alert"? Is Nvidia also about to sound the alarm? Illustrated AI competition

```

A major news event in the field of artificial intelligence (AI) this week is that OpenAI CEO Sam Altman announced to all employees on Monday the launch of a "red alert," aiming to focus all resources on optimizing ChatGPT to respond to fierce competition from Google Gemini. This strategic shift reflects profound changes in the landscape of AI competition and reveals the potential threat posed by Google's self-developed TPU chips to Nvidia's dominance in chips.

According to media reports, OpenAI has decided to postpone the development of other products, including advertising business, health and shopping AI agents, and the personal assistant Pulse, reallocating core resources to improving the daily user experience of ChatGPT. Altman said OpenAI still needs to improve ChatGPT's daily experience, including enhancing personalization, speed, and reliability, as well as expanding the scope of questions it can answer.

UBS tech analyst Tim Arcuri pointed out in his latest research report that Google's new-generation TPU chip Ironwood and its TPU ecosystem are posing a substantial challenge to Nvidia. Nvidia's stock performance has already lagged significantly behind Google's.

Google User Time Surpasses, ChatGPT Daily Active Users Decline

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

Even more noteworthy is that users now spend more time chatting on Gemini than on competing chatbots such as ChatGPT or Claude.

According to statistics from Deedy Das, in the two weeks since the launch of Google Gemini 3, ChatGPT’s number of daily unique active users (7-day average) has declined by 6%, showing the direct impact of competitive pressure. Although OpenAI still has more than 800 million weekly active users and holds a dominant position in overall chatbot usage, users are moving over to Google.

Nick Turley, the head of ChatGPT at OpenAI, posted on social media Monday night that search is one of the largest opportunity areas and that ChatGPT now accounts for approximately 10% of global search activity and is growing rapidly.

He also said that the company’s priority is to make ChatGPT more powerful, continue growing, expand global access, and make it more intuitive and personalized.

UBS: Google TPU Chips Pose Threat to Nvidia

Behind the AI model competition, the rivalry at the chip level is equally fierce. UBS tech analyst Tim Arcuri pointed out in his research report that advances in Google 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 models (MoE), and advanced reasoning, supporting training, fine-tuning, and inference workloads—a contrast to the narrow customization of previous TPUs.

Ironwood has not yet been submitted to MLCommons’ MLPerf v5.1 data center training benchmarks, but given its more computing resources, FP8 support, and vastly higher high-bandwidth memory than previous generations, Arcuri expects its single-chip performance to significantly surpass Trillium.

Arcuri noted that Google’s previous generation Trillium chip was optimized specifically for inference workloads and had lower HBM capacity (32GB vs. 95GB). In contrast, Ironwood has more computing resources, FP8 support, and a significantly increased HBM capacity, so its single-chip performance is expected to vastly outperform Trillium. Ironwood also expands TPU scale to a domain of up to 9,216 TPUs, far surpassing v5p's 8,960 and Trillium’s 256.

Arcuri pointed out that this is why Nvidia’s entire ecosystem is performing noticeably behind Google, and Google is enjoying a surge in attention brought by its TPU products. Koray Kavukcuoglu, CTO of Google DeepMind, said that by training AI models with Google’s self-developed custom chips, the company has “significantly improved performance.”

UBS believes that although Google may consider expanding the TPU ecosystem over time, any such efforts must be limited to avoid potential cannibalization of Google Cloud Platform (GCP) revenue. From this perspective, Meta and Apple are prime candidates for internal TPU deployments, as they have large AI projects and massive internal AI clusters to support workloads, and are relatively less dependent on GCP.

OpenAI Faces Multiple Competitive Pressures

The background for OpenAI triggering the red alert this time is the pressure from several competitors. The new version of Google’s Gemini AI model released last month outperformed OpenAI in industry benchmarks, sending Google parent company Alphabet’s stock price soaring. Last week, Alphabet’s shares rose over 14%; since Gemini 3 was released less than two weeks ago, the stock surged more than 10% as of last Friday.

Since the launch 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 under pressure from Anthropic, which has become increasingly popular among enterprise customers. Although OpenAI still has more than 800 million weekly active users and is dominant in overall chatbot use, users are gradually being attracted to Google.

Nvidia Responds to TPU Challenge

Faced with the rise of Google TPUs, Nvidia, in communications with UBS, emphasized its strong relationship with Google Cloud Platform, pointing out that Google uses both TPU and GPU for Gemini inference workloads.

Nvidia believes that cloud service providers are unlikely to run TPUs in their cloud stack, as significant workload optimization is required to achieve total cost of ownership (TCO) advantages with application-specific integrated circuits (ASICs). Nvidia also stated that so far, its performance advantage over peers has not narrowed.

Looking ahead to 2026, Nvidia notes that Anthropic’s 1 GW capacity and HUMAIN’s planned 600,000-unit expansion add upside potential to the $500 billion worth of orders expected for 2025-2026.

Nvidia's CPX chip is aimed at advanced programming applications requiring context windows of over 1 million tokens. Nvidia has not officially announced the market size, but previously hinted that context window applications account for about 20% of the inference market.

Altman said last month that over the next eight years, OpenAI’s data center projects are committed to a total investment of approximately $1.4 trillion. In other words, OpenAI has $1.4 trillion in committed funds to maintain its industry leadership.

All in all, OpenAI does have reason to be nervous, but this turmoil is still confined internally for now. As for whether Nvidia, the world’s most valuable company by market cap, is also facing a similar "red alert," the market is still watching closely.

Risk Reminder and DisclaimerThe market involves risk, and investment needs to be cautious. This article does not constitute personal investment advice, nor does it take into account individual users' particular investment objectives, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in this article are appropriate for their specific situation. Acting on this information is at your own risk. ```