
In a video released at the end of 2025, several partners at a16z, who have long invested in US industry, financial services, and enterprise software, pointed out that in early 2025, OpenAI and Google accelerated their deployment of consumer AI almost simultaneously. From models to interface design, the goal was to make it easy for ordinary users to get started. a16z believes that the consumer AI market has already shown an early structure of "winner-takes-all" or "a few leaders controlling the majority of usage," but as demand diversifies, the LLM market is becoming more segmented, and new entrepreneurs remain the key drivers in application and product experience.
The battle for consumer AI began at the start of the year.
Looking back at the beginning of 2025, OpenAI and Google accelerated their investment in the consumer AI market at almost the same time. Whether it was launching new models, updating features, or trying out brand-new user interfaces, the direction was quite clear: to make AI easy for general users without technical backgrounds.
a16z points out that the reason this competition is so crucial is that the consumer LLM assistant market has already shown early signs of a "winner-takes-all" scenario, or at least the winner taking the majority of usage. Whoever successfully establishes a foothold first may have their advantages amplified over time.
In terms of current usage, ChatGPT is significantly ahead.
Based on actual usage data, most consumers currently only choose one primary AI product. Surveys show that only about 9% of users simultaneously pay for two or more of ChatGPT, Gemini, Claude, or Cursor.
In terms of overall scale, ChatGPT currently boasts approximately 800 to 900 million weekly active users, significantly outpacing its competitors. Gemini accounts for about 35% of ChatGPT's scale on web and about 40% on mobile. Claude, Grok, and Perplexity, on the other hand, mostly fall within the 8% to 10% usage range.
Gemini usage is soaring, and market segmentation is becoming increasingly apparent.
However, a16z also pointed out that the market situation has changed significantly in the past 3 to 6 months. With the rapid popularity of various image and video models, Gemini's desktop user base has grown by 155% year-on-year, while ChatGPT's year-on-year growth rate is about 23%.
At the same time, the market is beginning to show a clear trend of segmentation. For example, Anthropic's Claude is gradually establishing a clear position among highly technical or professional users, no longer competing head-on with ChatGPT and Gemini for the general mass market.
The biggest highlight of 2025 will focus on image and video generation.
From a model perspective, the most virally successful technologies in 2025 will no longer be text models, but rather image and video generation. OpenAI's initiatives include the "Ghibli-style" craze sparked by GPT-4 Images, and Sora and Sora 2. Google, on the other hand, launched VO, V3, V3.1, and the later wildly popular Nano Banana and Nano Banana Pro.
a16z describes Nano Banana's reach as being comparable to, or even surpassing, the influence of Studio Ghibli's iconic moments. Meanwhile, the technological focus of image models has gradually shifted from aesthetics and style to realism and reasoning capabilities, enabling them to simultaneously understand multiple images and text, handle background dynamics and physical plausibility, and even generate complex infographics and market maps.

Looking ahead to 2026, AI is moving towards multimodal integration.
In terms of product strategy, OpenAI and Google have taken different paths. OpenAI integrates most of its functions directly into ChatGPT, releasing Sora only as a standalone video app. Google, on the other hand, distributes its products through Gemini, Google AI Studio, Google Labs, and multiple independent websites, giving each product its own interface. a16z points out that this difference directly affects whether ordinary users know where to start after opening the product.
In terms of social features, OpenAI's group chat and Sora video wall have received a relatively conservative response overall. a16z believes that ChatGPT's core function remains a productivity tool, making it difficult to satisfy the motivation of social platforms to "be seen and recognized."
Looking ahead to 2026, a16z generally believes that AI will continue to move towards multimodal integration, enabling "any form of input and any form of output," while large language model providers will likely continue to focus on optimizing core experiences. Meanwhile, the development of consumer AI applications with strong product characteristics will still be dominated by new startups.
This article, a16z, looking ahead to 2026: Startups will remain a key driver of consumer AI, and the outline of LLM (Local Management Model) is taking shape . It first appeared on ABMedia ABMedia .





