- OpenAI this week unveiled two new open weight GPT models
- Open weight models allow users to customize the models for specific tasks
- That's great news for telcos looking to serve up sovereign and specialized AI services
Telcos have been trying to carve out a niche in the burgeoning AI market for a while now. The likes of Orange, Telus, Telenor, Telefonica Fastweb, Indosat, Singtel and Swisscom, have gone the AI factory route, looking to build local compute muscle to meet workload demand. But it seems a new door has just opened courtesy of OpenAI.
OpenAI just dropped two new open-weight, open-source GPT models designed to deliver top-tier performance at a lower cost. They notably require much less compute power to run, with the gpt-oss20B version able to run on just 16 GB of memory.
That’s key given other open weight options, like Meta’s Llama 4 family, need significantly more muscle to run efficiently (though Mistral’s 7B model comes close).
Why does this matter? The smaller size means OpenAI’s models can run in a wider variety of environments, including at the edge. And the open weights mean those using the models can tweak the training parameters and customize them for specific tasks.
That’s great news for telcos looking to cash in on AI.
“The availability of open weight models gives telcos a choice of sovereignty (data never leaves controlled infrastructure), customization (models fine-tuned for specific tasks, computing footprints, regulatory and business needs), and economics (avoiding hyperscaler or AI provider markups while leveraging existing infrastructure investments),” AvidThink Founder Roy Chua told Fierce by email.
Telco model momentum
Both Meta and Mistal’s open weight models were already popular in telco circles. But Chua noted that “the strong performance of the OpenAI OSS models, along with their convenient sizes, plus the OpenAI brand will likely catapult them to one of the top choices in the months ahead.”
Indeed, French operator Orange wasted no time, announcing plans to deploy the new OpenAI models in its regional cloud data centers as well as small on-prem servers and edge sites to meet demand for sovereign AI solutions.
Steve Jarrett, Orange’s Chief AI Officer, noted the decision to use state-of-the-art open-weight models will allow it to drive “new use cases to address sensitive enterprise needs, help manage our networks, enable innovating customer care solutions including African regional languages, and much more.”
All of this means telcos have an opening to chase what is tipped to become a massive sovereign AI market.
Though a fraction of the overall AI infrastructure market, Bank of America’s financial analysts recently indicated the sovereign AI segment could become a $50 billion a year market.