Technology procurement has been upended in little more than a year by a mix of geopolitical tension, economic pressure and the pace of AI development. What once looked like a straightforward buying decision now carries a strategic risk: if an organisation builds too much on a single platform, it can find itself exposed to pricing shocks, service changes or even political intervention.
That concern has become more visible as companies have watched the on-off-on episode around An...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
thropic’s Fable 5. For enterprise leaders, the lesson is increasingly hard to ignore: dependence on one AI supplier is no longer just a technical preference, but a business vulnerability.
Former French prime minister Édouard Philippe put the issue starkly when he said: “An infrastructure whose models and computing power we don’t control is an infrastructure that others can unplug.” For CIOs, that warning is pushing AI strategy towards resilience rather than loyalty.
The answer, for many, is diversification. Unlike core systems such as CRM and ERP, where running multiple suppliers side by side can be messy and expensive, AI workloads lend themselves to a more flexible approach. Enterprises can route different tasks to different models depending on cost, performance, reliability or policy requirements. That is where model gateways and orchestration layers are beginning to matter.
Tools such as OpenRouter, LiteLLM and Portkey are becoming part of the plumbing that sits between applications and model providers. They allow firms to switch traffic between vendors, set fallbacks and manage access without rewriting every workflow each time a model changes. According to OpenRouter, the volume of tokens it coordinated rose from 5 trillion a week to 20 trillion a week in the year to April 2026, suggesting that this kind of abstraction is moving rapidly from experiment to infrastructure.
A June 2026 survey cited in the coverage found that two-thirds of 145 enterprises had already adopted diversified AI deployment strategies before the Fable 5 shutdown, reinforcing the sense that vendor spread is no longer niche thinking. In practice, organisations are trying to avoid being trapped by one provider’s pricing, limits or outage risk.
The market for gateways is itself becoming more sophisticated. Comparative guides published this year describe LiteLLM as appealing to teams that want more direct control, OpenRouter as useful for broad model access and low-friction experimentation, and Portkey as geared towards enterprise guardrails and oversight. Other round-ups add Vercel AI Gateway, Cloudflare AI Gateway, Kong AI Gateway, TrueFoundry and Requesty to the list of tools vying to solve routing, caching, fallback and cost control.
Governance is emerging as a central theme. OpenRouter argues that visibility into how models are used should come before any claim of governance at all, and recommends managed gateways where LLM access is fragmented and hard to monitor. That reflects a wider shift in enterprise AI, where observability, authorisation and redaction are increasingly treated as design requirements rather than afterthoughts.
For CIOs, the message is becoming clear: AI sovereignty is not about shutting out external providers. It is about making sure no single supplier can dictate how critical systems run. In a market where models change quickly and access can be unstable, diversification is turning into a basic condition of control.
Source: Noah Wire Services
