At Anthropic’s developer conference last week, I met Guillermo Rauch, CEO of AI startup Vercel. We started chatting, and Anthropic wasn’t the first company he mentioned. It was Google.
Overall, he said, the demand for AI is extraordinary. But Google’s model is especially in demand among Vercel customers. Rauch even said that he recently had to call Google executives to add the Gemini token, a core unit of AI usage.
While all the talk these days is about Anthropic and OpenAI, Google’s Gemini is quietly gaining traction, he said.
This can be seen with Vercel’s AI Gateway, which allows businesses to connect apps to different AI models through a single system. It is primarily used by AI startups, software companies, and enterprise product teams running AI functions such as chatbots, coding assistants, search tools, and co-pilots.
Take a look at this chart. Based on the number of tokens (traffic) processed by Vercel’s AI gateway, the Anthropic model was on top in March. In early April, Google’s Gemini 3 Flash model jumped to the top and has remained there ever since. And this is before Google’s big annual conference I/O, which starts next week. The company is likely to announce a number of more capable AI models, tools, and features. (I’ll be there too, so stay tuned next week.)
Data from Vercel’s AI Gateway Vercel
Gemini Flash is less powerful than the full Gemini 3 model, but it is faster and cheaper to use. This makes it especially popular among Vercel’s corporate customers.
“Enterprise teams tend to choose Gemini Flash and Claude Haiku, which are the smallest, fastest, cheapest models that each lab ships,” Rauch told me. “Flash, in particular, is seeing strong B2C adoption because it has few hallucinations, is an effective tool to use, is fast, and is affordable.”
However, with AI, the answer is never simple. There are also other ways to measure success, such as how much users spend on the model
“We often get asked which lab is ‘winning,’ but what we see in production looks nothing like a benchmark leaderboard,” Rauch says. “AI Gateway reflects different models that enable different use cases.”
When it comes to token usage, Google was a clear winner in April. However, based on spending, Anthropic came out on top with a 61% share. “Some models are cheap and succeed in high-volume traffic, while others are expensive and succeed in quality-critical work. They solve different problems,” Rauch explained.
With the launch of the new GPT-5.4 and 5.5 AI model series, OpenAI’s spending share tripled from March to April (from 4% to 12%). As usage of Gemini Flash grew, Google rose from 8% to 21% in terms of spend.
“A snapshot of one month cannot predict the next,” Rausch warned.
That’s especially true now that Google I/O is just around the corner.
Sign up for BI’s Tech Memo newsletter here. Please contact us by email. abarr@businessinsider.com.