Telcos, AI and the growing dependency on hyperscale service providers
Up until 2021, hyperscale cloud service providers’ prominence at MWC grew at the same gradual pace as telcos’ interest in moving applications into the public cloud. And even when hyperscalers became markedly visible in 2021, it was a Covid-impacted show where big telco vendors such as Ericsson were absent.
Fast-forward to 2024, however, and there is every reason to believe that Microsoft, AWS and Google will be star attractions, rather than the network and IT vendors that have traditionally set the show’s technology, product and service agenda.
GenAI, which exploded into the market in the last year to 15 months, will be a large part of the reason why. In the short term there is no way around working with one or more of the hyperscalers to experiment with GenAI. They own, or have deep relationships with, the dominant large language models. More importantly, they provide services that make it easy to consume GenAI capabilities.
Today most of the telecoms operator groups and large national telecoms operators already engage with Microsoft, Google and AWS to buy GenAI services and solutions.
GenAI is only one form of AI, and it is not about to replace traditional, predictive AI and machine learning, which telcos have been using for years. But if GenAI and predictive AI together become as important as some hope, then it opens up a question for smaller operators, particularly in those countries where public cloud usage is relatively modest and hyperscalers have little or no presence. Namely, can they count on the support of Microsoft, Google or AWS when looking to adopt and integrate GenAI capabilities into their IT and network systems? And if they can’t, will a lack of access to cloud supercomputing and AI capabilities create a digital divide within the mobile communications sector?
AI haves and have-nots?
Mobile communications evolved as a truly global sector in the period 1990 to 2010. Walking the exhibition floors at MWC you are as likely to bump into an executive working for a mobile operator in Iraq, Bolivia, Senegal or Mongolia as one from a European or North American telco.
Amidst all the soul-searching and uncertainty about the future of the telecoms business it is good to be reminded of how successful the mobile business has been at democratizing access to communications and broadband services across the world. Indeed, even the smallest, most economically challenged island nations have built mobile broadband networks. It is testament to the success of cellular technology standardization, global adoption of a single standard and the emergence of a healthy market for mobile networks which has made it viable to deploy networks in small countries with modest consumer spending power.
But how long will it be before operators in these small and emerging economies get access to the same AI tools and capabilities as operators in larger, higher-GDP markets? Will the adoption of AI be as widespread as that of mobile technology itself? Are we starting out on a path to “AI haves” and “AI have nots?”
Over the last two months TM Forum has been speaking to telecoms operators from across the world about their AI journeys. Many large operators are making solid progress through partnerships with systems integrators and hyperscalers and skilling up their own organizations. But other operators are already falling behind. “They (hyperscalers) don't have the scale to support customers to get things meaningfully done”, commented the CIO of one large African telecoms group in relation to the deployment of GenAI co-pilots. This particular telecoms operator even has a strategic relationship with one of the companies the executive was referring to.
Telco AI
In reality, AI is a technology that is being applied across all industries. Hyperscale service providers have set up teams dedicated to serving telecoms operators but their corporations are not focused on the telecoms industry in the same way as Ericsson, Nokia or Huawei. Decisions about whether to set up a data center facility in a new country are based on the potential business that can be generated across all sectors of the economy. Added to that, many countries have strict rules on data security and privacy, limiting how telecoms operators can use public cloud services and AI.
Reducing telecoms operators’ dependency on the LLMs which are gaining most traction is one of the motivations behind the creation of the Global Telco AI Alliance, an initiative to build a telecoms-operator LLM. Chung Suk-geun, Chief AI Global Officer at South Korean telecoms operator SK Telecom, and founder of the Alliance, told delegates at last September’s DTW Ignite conference last September that telecoms operators were having problems getting the attention of the companies building LLMs.
With SK Telecom, and its partners Singapore Telecom, Deutsche Telekom and E&, due to launch its LLM in spring 2024 it’s a fairly safe bet that its stand will be an MCW 2024 focal point. If the Global Telco Alliance is successful, it will give the telecoms industry some ability to prioritize new AI-driven capabilities for the telecoms-operator community. But there is little or no chance of telecoms operators being able to capture the potential of AI without relying to a greater or lesser degree on hyperscale service providers.
Only time will tell whether those operators that have to wait for AI capabilities to become available miss out on efficiency gains or enhanced customer experiences. There is a general assumption with new technologies that early adopters derive business benefits. But, at the same time, there may be lessons to be learned from the deployment of mobile network technologies. Operators which have been late in building next-generation networks have benefitted from significantly lower prices as a result of global economies of scale.