Happy New Year! 🎆 To usher in 2025, telco industry journalists and pundits have been busy making predictions for the year ahead. It’s no surprise that most everyone is forecasting 5G expansion, AI- and generative AI-powered solutions, and my favorite: more migration to cloud-native networks. But there are some interesting outliers, too. ABI Research thinks network APIs will face monetization challenges. Forbes sees the possibility for real-time holographic communications to appear this year. Dean Bubley wants us to tone down the hype. And yours truly says (via Telecoms.com) that this is the year telcos will divide into those who really, really get AI, and those who sit on the sidelines and watch the opportunity pass them by. I can’t wait to check back at the end of the year and see how it all panned out!
NTT DOCOMO is partnering with AWS on its 5G core network and virtual radio access network (vRAN) projects. Using Amazon EKS Anywhere for virtualization, DOCOMO is looking forward to more flexible feature deployment, shorter time to market, better energy efficiency, lower costs, and improved service availability. Chalk up another win for telco networks running on the public cloud!
December was a wild month for AI breakthroughs. OpenAI’s 12 Days of Shipmas peaked with the preview launch of its new o3 model. Google followed with Gemini 2.0, designed for the new era of AI that’s all about agents, with everyone I know raving about its Deep Research functionality. Anthropic launched Claude 3.5 Haiku, its “next-generation fast model,” hot on the heels of Model Context Protocol, a new framework for connecting AI assistants across diverse platforms. Nvidia CEO Jensen Huang made a monster keynote at CES (which was technically in early January, but we’ll count it). We saw a surge in AI tools for business and marketing, and the venture capital is flowing to AI-focused startups. Still think it’s all hype? I don’t think so. AI is happening BIG TIME.
Anthropic just shared tips on how to build effective agents, a topic I predict will be the talk of MWC25. In a nutshell: keep it simple at the start and add complexity only as needed. I think the key to writing good agents is making them ACT, not just CHAT and making sure the action they take doesn’t require human intervention to complete the task. We’re going to see a bunch of "bad agents" at MWC—meaning co-pilots disguised as agents. If you want to see some kick-ass telco agents at MWC, swing by the Totogi stand in Hall 2!
Forget becoming “techcos”: Telcos have set their sights on becoming AI giants. Companies like Deutsche Telekom, SoftBank, Fastweb, Indosat, Singtel, and Swisscom have begun investing in data centers built specifically for AI workloads, and Telenor has formed Skygard to pursue AI startups, while SK Telekom is developing its own large language model (LLM). The key to making this all work? Completely reimagining the telco and rebuilding it, from the ground up, to be AI-first. It’s as much a cultural shift as a technical one.
Microsoft President Brad Smith says the next four years are a golden opportunity for American AI. He envisions a three-part strategy of strengthening AI tech and infrastructure, upskilling workers, and prioritizing AI exports to allies and friends to outpace China’s rapid innovation. The hyperscaler has announced a massive AI infrastructure investment in FY2025—approximately $80 billion to build out AI-enabled data centers globally, with over half earmarked for U.S. facilities. Telcos, are you still planning on your GPUaaS strategy and banking on AI sovereignty to bolster your chances? I wouldn't bank on a sole infrastructure play to make this strategy work…
SoftBank is ready to invest $100 billion in US AI and related infrastructure, and another $960 million in Japanese AI initiatives. Sounds great to me! CEO Masayoshi Son, since you’re handing out money: Have you ever heard of a little telco-focused AI company called Totogi? 🤔
The cost of LLM inference has dropped by a factor of 1,000 in three years. I’ve been blogging about why you’ll regret locking in today’s AI rates—and now venture capital firm Andreessen Horowitz is backing me up with a ton of awesome charts and graphs that prove just how fast costs are plummeting. This will continue to happen in the foreseeable future, so don’t lock in—ride the prices down, down, down instead.
Is Elon building a global telco? Yet another operator, VEON, has partnered with Starlink to launch direct-to-cell satellite connectivity in Ukraine later this year. That’s on top of 12 others, including T-Mobile, Rogers, KDDI, Optus, One NZ, Salt, Entel, Telstra, Kyivstar, Claro, Verizon, and AT&T. For VEON specifically, the service will initially provide SMS and OTT messaging, then expand into voice and data. On one hand, Starlink is a great solution for rural areas (brings down the cost of the network), and definitely helps anywhere infrastructure has been destroyed (like in California or Ukraine). But on the other hand, operators are letting Elon use their valuable spectrum and connect to their networks. Are we letting the dragon into the castle on this one? Hmmm… 🤔
If you find yourself (or your tech team) in need of killer AI engineering skills, allow me to introduce you to Gauntlet AI! It’ll be a repeat of the process from the legendary, old-school Trilogy University program I used to run in the 90s, which built the epic Trilogians you know and love today (including me!). Totogi will be hiring from among the graduates—I can’t wait!
And finally, a quick shout-out to Jim Patterson at RCR Wireless who suggested listening to my podcast with Chris Penrose, NVIDIA’s global VP of business development for telco. It was a great episode covering how NVIDIA’s AI-RAN concept can help telcos use unused network capacity for enterprise AI workloads to boost the bottom line—including the results of a field trial with SoftBank. Thanks, Jim! 💜
