The priority of the artificial intelligence industry has changed. For the past few years it seemed like the rule was simple: invest more, buy more capacity and use the largest model possible. That stage is coming to an end.
Now the important question is another: how much concrete value does AI produce for the business? Managers such as Brian Armstrong, CEO of Coinbase, and Guillermo Rauch, CEO of Vercel, are already talking about more efficient strategies: using cheaper models when sufficient and combining several providers depending on the task.
What was talked about in Sun Valley
The Sun Valley conference, in Idaho, United States, brings together some of the most influential executives in the technology sector every year. This year leaders such as Ted Sarandos, Dara Khosrowshahi, Tim Cook and Andy Jassy participated.
The topic that dominated many conversations was efficiency. Sam Altman's statements come just as more companies are wondering how to get the most out of their investments in artificial intelligence.
Coinbase: the right model for every task
Brian Armstrong said Coinbase is experimenting with cheaper AI models as a standard solution for certain development tasks. He also explained a key idea: direct each query to the model that best fits that specific task.
The logic is simple: not every task needs the most powerful model on the market. Using the most expensive model to answer a simple question is like sending a trailer to deliver a letter.
Efficiency doesn't mean using cheap AI for everything. It means not paying more when a task can be solved with a simpler, faster and sufficient tool for the objective.
Vercel: do not depend on a single supplier
Guillermo Rauch, CEO of Vercel, defended a multi-vendor strategy: working with several AI laboratories at the same time and choosing the appropriate model for each part of the system.
In practice, a company can combine models from OpenAI, Anthropic, Gemini or other providers. Each has different strengths, different prices, and best uses depending on the type of task.
Use the largest model for everything
More cost, more dependency and less clarity about which part of the process actually generates value.
Design the process and choose well
Different tools for different tasks, with clear indicators of time, cost, response and monitoring.
What this means for your business in any market
For companies in different markets, this trend is good news. AI is becoming more accessible, but the real difference is not in having a trendy tool. It's about using it well.
Lower barrier to entry
Competition between vendors makes certain solutions more accessible to small and medium-sized businesses.
More strategy
It is not about installing AI for the sake of installing it, but rather identifying processes where time, customers or monitoring are lost.
More control
With small pilots you can measure impact before investing in a larger implementation.
At Ezelero we apply that same principle: diagnosis first, clear process, controlled pilot and tools chosen according to the budget, maturity and real objective of each company.
The era of spending for the sake of spending in AI is ending. The era of using it intelligently is just beginning.
Frequently asked questions
What is the Sun Valley conference?
It is an annual meeting that brings together some of the most influential executives in the technology sector in Sun Valley, Idaho, United States. This year, efficiency in artificial intelligence was one of the central topics.
What strategy is Coinbase following to reduce AI costs?
According to its CEO, Brian Armstrong, Coinbase experiments with cheaper AI models and directs each query to the most appropriate model to avoid unnecessary token consumption.
What does the CEO of Vercel recommend to companies?
Guillermo Rauch recommends working with several providers and choosing the right model for each task, instead of depending on a single platform for everything.
What does this trend mean for companies across different markets?
It means that a company can start smarter: identify a specific process, try a small pilot, measure results and then decide whether to expand.
Do you want to know what combination of AI is best for your business?
We review your processes, your service channels and your repetitive tasks to find a first measurable automation opportunity.