- Fierce Network Research's newest report describes how telcos can escape the dumb pipe trap by building AI factories
Telcos yearn to diversify from their core, low-margin business of delivering commodity data and voice services. Operators want to be more than just "dumb pipes" that watch over-the-top players like Amazon Web Services and Google get all the profits.
Our latest report from Fierce Network Research describes how telcos can escape the dumb pipe trap by building artificial intelligence (AI) factories for AI sovereignty. Download the free report, sponsored by Supermicro and NVIDIA: AI sovereignty: Seizing the opportunity for transformative telco infrastructure investment
What is AI sovereignty?
Let's explain, starting with AI sovereignty.
AI sovereignty is similar to data sovereignty, which you're probably already familiar with. Data sovereignty regulations and laws require data and applications related to people and organizations within a nation or region (like Europe) to be physically located within that nation or region.
The European General Data Protection Regulation (GDPR) is a great example of data sovereignty regulation. Individual nations within Europe, such as Germany, have their own data sovereignty requirements, as do countries worldwide, including the U.K., the United States, Singapore, Russia and Malaysia.
AI sovereignty applies the data sovereignty principle to artificial intelligence. Nations see AI as essential for unlocking economic leadership and leadership in science, culture and more. AI is a strategic resource, like transportation and food production, and nations see the need to control that resource themselves.
It isn't just a question of control. Generic large language models (LLMs) are trained on the entire internet. Nations and regions need AI trained in local languages, laws and business practices.
“Every country needs to have a policy around AI,” Joao Kluck Gomes, NVIDIA director of business development for AI factories and applications, told Fierce Network Research, speaking to us for our report. “They need to enable their national ecosystem to participate in this new economy, not only as consumers of this technology from the American hyperscalers, but as creators.”
What are AI factories?
AI factories are key to AI sovereignty. AI factories are data centers, but they are a special kind of data center. They are data centers that have been optimized to run AI workloads. As the report explains: A conventional data center comprises banks of individual computers operating independently of each other, running different workloads. But an AI factory is different.
“An AI factory produces one product — either training or inferencing,” Michael Clegg, Supermicro VP & GM 5G / Edge told Fierce Network Research. “The entire factory is a set of production lines that cranks out the same output. The whole thing acts in unison.”
AI factories require massive amounts of GPUs, and each GPU is connected by high-speed networking. Additionally, AI factories require resources to manage systems and large storage components.
What does all this have to do with telcos?
As our report explains: Telcos are uniquely positioned to build and operate AI data centers. They already have the relationships, technology, skills, physical real estate and other resources.
This is already happening. More than a dozen telcos worldwide are investing in AI factories and other AI infrastructure and services in nations and regions including Indonesia, India, Singapore, Southeast Asia, the Middle East and North Africa, Japan, Canada and South America.
But the road to success is rocky. Telcos have tried to expand beyond dumb pipes multiple times, and it usually hasn't gone well. Our report discusses that history and what telcos need to succeed in new and innovative business ventures.
Download the Fierce Network Research report now: Seizing the opportunity for transformative telco infrastructure investment.