Recently, Stripe agreed to acquire OpenRouter at a reported $8-billion. Just last week Nvidia reportedly agreed to buy Hugging Face, the open-source AI model hub, for $12,9-billion.
Different headlines, but they reflect the same broader shift. The value in AI is increasingly moving beyond individual models to the infrastructure and platforms that give businesses greater choice, flexibility and control.
Neil Dhar, senior vice-president at IBM Consulting, unpacks why it’s important.
Routing is one of the clearest examples. Axios reported last week that businesses are increasingly turning to routing to automatically match AI workloads with the model best suited to the task, based on factors like cost, performance and security.
For the past few years, everyone has been trying to figure out which AI model is going to win. Companies have evaluated and experimented with leading LLMs, picked providers and made bets on which technologies to build around. But as models become increasingly commoditised and open alternatives improve, business leaders have realized one model will not rule them all.
Research from the IBM Institute for Business Value shows that by 2030, most organisations will primarily be using smaller, customized models as CEOs shift to a hybrid AI strategy encompassing foundation models and specialised models based on specific business requirements.
The question then isn’t which model will win. It’s if you built an organisation that can change when the answer does.
Model routing offers a useful blueprint. Its value isn’t simply that it selects the right model for a given task. It enables companies to direct tasks to different models as performance, costs and business needs change. That same flexibility needs to extend across the enterprise.
The reality is that companies aren’t navigating a multi-model world alone. They’re increasingly operating across multiple models, agents, and data environments. If the model can change but everything around it is rigid, the enterprise hasn’t gained much flexibility at all.
The goal should be an AI environment where those pieces work together, but where changing one doesn’t mean rebuilding everything else.
That ability to change course is as much of a financial advantage as it is an architectural one. If AI capabilities, costs and use cases can change in months, the organisation needs to be able to evolve just as quickly.
The flexibility to choose between models is one thing. But can you move capital when a better opportunity emerges? Can you redesign a process when a new capability makes the old one obsolete? Can your people build new skills quickly enough to take advantage of it?
This isn’t an argument for business leaders with commitment issues. CEOs and CFOs still need to make big bets, but every investment should be judged on two things: the value it creates today and what it leaves the business free to do six months from now, when the model, the cost curve or the opportunity has changed.
That puts a premium on investments that remain valuable even as the technology changes. Workload portability, open technologies and the ability to operate across hybrid and partner environments avoid locking in today’s assumptions.
Ultimately, the flexibility to change models doesn’t mean much if the rest of the business can’t keep up. Models will get better, costs will shift and new capabilities will emerge. Companies need to be able to redirect investment, redesign processes and build new skills in response.
Otherwise, the technology keeps moving while the organisation falls further behind.
Every shift creates another decision about where to invest, what to build and how to operate. The decision you make today might not matter as much as your ability to make a different one tomorrow.