The AI market is undergoing unprecedented transformation. According to a recent comprehensive report from the Poe platform, the competitive landscape in the AI field experienced a significant shift in early 2025, with emerging companies rapidly eroding the market share of established giants.
Poe, a platform hosting over 100 AI models, provides a unique perspective on usage patterns of text, image, and video generation technologies, based on interaction data from millions of users over the past year. This usually closely guarded usage data offers invaluable market insights for technology decision-makers.
Market Fragmentation and the Rise of New Players
While established companies like OpenAI and Anthropic still dominate the text generation space, new entrants such as DeepSeek and Black Forest Labs have rapidly gained significant market share, showcasing the dynamism of this ecosystem.
Here are five of the most surprising findings from Poe's analysis:
1. Inconsistent Performance of Google's AI Models
Google's performance varies significantly across different AI models. Its Gemini series of text models experienced growth until October 2024, but subsequently declined. In contrast, Google's Imagen3 series in image generation commands 30% of the market share, while its video generation model, Veo-2, rapidly captured 40% share.
This uneven performance demonstrates that technological prowess doesn't guarantee across-the-board market leadership. Business decision-makers need to evaluate AI capabilities individually, rather than assuming leadership in one area translates to overall excellence.
2. Fierce Competition in Video Generation
Video generation, the newest frontier of generative AI, has witnessed intense competition and rapid shifts in leadership. This category, only emerging in late 2024, has quickly expanded to over eight providers.
Runway, an early pioneer, despite having only one API model, consistently maintains a strong 30% to 50% market share. However, Google's Veo-2, since its launch, has captured nearly 40% of the market within weeks.
Chinese-developed models collectively account for approximately 15% of video generation usage. Models like Kling-Pro-v1.5, Hailuo-AI, HunyuanVideo, and Wan-2.1 continuously push the boundaries in terms of capabilities, inference time, and cost, demonstrating that international competition remains a crucial driver of innovation.
3. Fundamental Shifts in Image Generation
The image generation space is perhaps the most dramatically changed area of the AI market. DALL-E-3 and various Stable Diffusion versions, as early pioneers, saw their relative usage share drop by nearly 80% as the number of official image generation models increased from 3 to approximately 25.
Surprisingly, Black Forest Labs' Flux series of models, having gained popularity since mid-2024, has maintained a clear lead, holding nearly 40% of the market. Google's Imagen3 series, launched in late 2024, has steadily grown, capturing nearly 30% usage share and becoming a strong second-place contender.
4. Rapid Replacement of Older Models
Poe's data reveals a concerning trend: the usage of new flagship models rapidly cannibalizes older versions as more powerful models are released. Users swiftly migrated from GPT-4 to GPT-4o and from Claude-3 to Claude3.5.
This phenomenon suggests that maintaining support for older models may yield diminishing returns. Businesses may need to reconsider product lifecycle strategies, focusing resources on fewer models with more frequent updates.
5. Text AI Duopoly Faces New Challenges
OpenAI and Anthropic maintain dominance in text generation, accounting for approximately 85% of text interactions on the platform. Since the release of Claude3.5 Sonnet in June 2024, both companies have seen nearly equal text usage, indicating intensifying competition.
Even more striking is the emergence of DeepSeek as a legitimate third contender. DeepSeek-R1 and V3 went from zero usage in December 2024 to a peak of 7% market share, significantly higher than any previous open-source model. This sharp rise suggests that the barrier to entry for new text AI providers may be lower than anticipated.
Future Outlook: An Uncertain Landscape for AI Decision-Making
Poe's report clearly illustrates the rapidly evolving nature of the AI market: technological superiority alone does not guarantee sustained market leadership. Business decision-makers face an increasingly complex vendor landscape, where today's leaders could become tomorrow's laggards.
User preferences can dramatically shift with the release of new models, suggesting organizations should build flexible AI stacks to adapt to constantly evolving capabilities, rather than locking into single-vendor solutions. The emergence of different leaders in text, image, and video AI further complicates enterprise strategy.
As Poe concludes: "We hope these findings offer a glimpse into the dynamic shifts in the AI model landscape." The message is clear for businesses navigating this fluid ecosystem: the AI revolution is unfolding at breakneck speed, rewarding those who remain agile while penalizing those overly reliant on yesterday's technology leaders.
