ChatGPT and Gemini both pass 1 billion users — right as EU AI rules go live

The race, in numbers

Google CEO Sundar Pichai confirmed that the standalone Gemini app surpassed 1 billion monthly active users on August 11, making it the fastest-growing product in Google's history and the company's 14th to reach the ten-figure threshold. The trajectory is steep: 750 million in February, 900 million+ in May, 950 million at the end of July, and 1 billion now (via The Verge). That is roughly 50 million new users in three weeks.

OpenAI got there first. ChatGPT passed 1 billion monthly active users in June 2026 and then 1 billion weekly active users in July, alongside more than 2 million enterprise customers. So "both passed 1 billion" is true — but the two companies are measuring different things. ChatGPT's weekly figure is the stricter metric.

The caveat behind Gemini's 1 billion

One nuance matters: Gemini's count covers only the standalone Gemini app and web interface. It excludes Gemini features embedded in Google Search, Workspace, and Android. Google's wider AI surface is much larger — AI Overviews exceed 2.5 billion monthly users, and AI Mode in Search maintains over 1 billion monthly active users (75 million daily).

The user behaviour behind the growth has also changed. Google reports that 63% of Gemini users interact via voice, one in five voice conversations uses camera video or screen sharing, and Gemini generates over 150 million images per day. More than 100 million monthly active users access Gemini on iOS alone. Voice and vision, not text prompts, are the growth engine — something European product teams should note before betting everything on chat UIs.

The EU angle: Article 50 is no longer theoretical

Here is the part that matters most for European readers: on August 2, 2026 — nine days before Gemini's milestone — the EU AI Office gained binding enforcement powers under Article 50 of the EU AI Act. The grace period for general-purpose AI is over. In practice, this means:

• Interactive chatbots (ChatGPT, Gemini, Claude, and their derivatives) must clearly disclose to users that they are interacting with AI.
• Deepfakes and AI-generated media require machine-readable watermarking, not just a barely visible label.
• GPAI providers must keep compliance documentation and answer binding information requests from the EU AI Office.

The EU AI Office can now conduct model evaluations, demand risk mitigations, recall models, and impose administrative fines. For ordinary European users, the most visible change will be clearer wording in chatbot interfaces. For companies running ChatGPT or Gemini in production, the compliance paperwork just became legally binding. The Digital Omnibus recalibration pushed the broader high-risk obligations to late 2027, so this first wave is specifically about foundation models and consumer-facing transparency.

What the rest of the field looks like from Europe

Anthropic's Claude Opus 5 (July 24, 2026) is bundled into Pro and Team subscriptions at $20–$30 per user per month (roughly €18–€27). On the API side, xAI's Grok 4.5 costs $2.00 per million input tokens and $6.00 per million output tokens (about €1.80 / €5.45). OpenAI's GPT-5.6 Sol is the price aggressor: $0.20 per million input and $1.20 per million output tokens — around €0.18 / €1.08 at current exchange rates.

Nobody else publishes a comparable user count, which is exactly what makes the ChatGPT and Gemini figures stand out. European alternatives are not racing for consumer scale: Mistral's recent releases — the Shieldstral 1.0 safety model (August 3) and Mistral OCR 4 — are open-source tools aimed at enterprises that prioritise sovereignty. Aleph Alpha's Pharia and self-hosted Mistral Large deployments target the same segment. Open-weight options like GLM-5.2 (MIT), DeepSeek-V4-Flash, and Meta's Muse Glimmer (30B, Apache 2.0) give European developers full control over data locality, which is increasingly the deciding factor for regulated industries.

Production reality: price stopped being the bottleneck

From our own experience running AI services on production servers — transcription pipelines, text-to-speech, article workflows — the practical takeaway is that token price is no longer the main constraint; trust and regulatory risk are. At roughly €0.18 per million input tokens on GPT-5.6 Sol, usage costs are almost negligible for many workloads. What keeps European companies tied to specific vendors is the ability to demonstrate GDPR compliance, data residency, and — as of this month — Article 50 documentation.

In AI Arena, our RTX 5060 Ti 16 GB benchmark rig, we keep measuring exactly these production trade-offs: how fast a model actually runs, what it costs per million tokens, and whether it can be self-hosted in Europe. The consumer race, however, is being won elsewhere — by voice latency, camera input, and the sheer polish of the free tier. ChatGPT and Gemini both have free tiers available across the EU, which makes them the default entry point for hundreds of millions of European users. The next question is whether European regulators will shape these platforms as much as their product teams do.