Trouble in paradise? Microsoft just launched two homegrown large language models (LLMs), MAI-Voice-1 and MAI-1-preview, signaling a desperate attempt to reduce their dependency on OpenAI. After investing $13 billion-plus in the partnership, Redmond is essentially throwing good money after bad to avoid being completely at OpenAI’s mercy. A few years ago, both Microsoft and Google had a big, one-horse play: Microsoft with OpenAI, Google with Gemini. Conversely, Amazon Web Services (AWS) went with a lot of different LLMs. Lots of people said AWS had the wrong strategy; turns out, it was the right one, as LLMs have quickly become a commodity. Now, Microsoft is scrambling to catch up.
OpenAI is expanding its options, too, as it’s reportedly ready to start mass-producing its own AI chips next year in partnership with Broadcom, reducing its reliance on NVIDIA. This is the AI supply chain completely reshuffling—everyone’s trying to escape everyone else’s chokehold. NVIDIA has been printing money, charging whatever they want for H100s, so of course OpenAI wants options. But here’s the thing: OpenAI is now fighting wars on two fronts: the LLM war against Microsoft, Google, and Anthropic, AND a chip war against NVIDIA. Building custom silicon is brutally expensive and takes years to get right. Ask Google about its TPU journey, or Amazon about its Graviton chips. It’s not just about designing the chip; it’s about the entire software stack, drivers, and ecosystem. I’ll keep my eye on this one…
Mukesh Ambani is making power moves in India’s enterprise AI market. Reliance is teaming up with Meta for a $100 million AI joint venture that will leverage Meta’s open-source Llama models to create customized AI platforms for Indian businesses. The 70-30 split favoring Reliance makes sense—it brings local market expertise and massive infrastructure, while Meta provides the AI models. But here's what’s really happening: this looks like the same playbook NVIDIA is running with its 18 telco AI factory partnerships. Meta gets to compete in India’s enterprise AI market without building data centers or navigating regulatory complexity, while Reliance does the heavy lifting with infrastructure investment and local execution. The difference? Ambani actually gets majority control and becomes the gatekeeper for enterprise AI across 1.4 billion people. Unlike most telcos getting played by tech giants, Reliance might actually be playing the game right.
Is there a new fourth network in the USA? SpaceX just acquired EchoStar’s AWS-4 and H-block spectrum licenses for $17 billion, giving Starlink direct-to-cell capabilities without depending on T-Mobile or other carriers. But here’s the buried headline: SpaceX just revealed they’re already the “largest 4G coverage provider on planet Earth” with over 6 million Direct to Cell users across five continents. The deal also gives Boost Mobile subscribers access to next-gen satellites that will deliver 100x more capacity using custom SpaceX silicon and optimized 5G protocols via Starship launches. The Wolf of MWC didn’t just become the fourth major US mobile operator overnight—he’s apparently been the biggest global 4G provider all along. For telcos already partnering with Starlink: congratulations, you’re now helping fund your biggest competitive threat. Wowzers.
Thanks to Scott Bicheno and Iain Morris at the Telecoms.com Podcast for giving me a shout-out in their latest episode (at 1:07) which focuses on EchoStar’s $23 billion sale of spectrum to AT&T. The gist: the guys were chatting about how the DISH network was a technical failure, and the egg was on AWS’ face, but I disagree. My take: the EchoStar issues were commercial, not technical. The network DISH built in the public cloud 100% worked, and it was a lot cheaper to do it that way, as predicted. I spoke with CTO Mark Rouanne about it, and he confirmed: the network works great. The business side just never built out the subscriber growth it needed to survive. Proof that to build a successful telco, it’s not just about the network. You need great marketing and sales, too.
Bell Canada is saying the quiet part out loud about AI adoption: it’s not about the technology anymore; it’s about the people. John Watson, Bell’s Group President, nailed it when he said everyone—not just the IT team—needs to be “AI-literate and AI-centric” in their roles. Bell is democratizing AI access across 12,000 field technicians, proving that successful AI deployment requires cultural transformation (check out my podcast with leadership guru Jim Abolt), not just technical integration. This is the reality check telcos need. You can’t just bolt AI (or any other technology for that matter) onto existing processes and expect magic to happen. You need everyone at your telco, top to bottom, living and breathing AI.
It’s a renaissance for mobile virtual network operators (MVNOs) right now, with everyone from fintech firms like Klarna and Revolut to celebrities launching mobile services. UK MVNOs jumped from 14.9% to 19.7% of mobile connections in just three years, growing 9.8% in 2024 while the overall market crawled at 0.7%. The game-changer? New cloud-native platforms and eSIM technology that can launch services in days instead of traditional 12-month implementations—both of which dramatically lower the initial investment to launch an MVNO. Totogi’s multi-tenant charging system is part of this movement—free for the first 250,000 subscribers, and pay-as-you-grow after that. As we told TM Forum, this lets MVNOs “control their destiny and move at their own speed” while cutting out the traditional middlemen, mobile virtual network enablers (MVNEs).
Omdia’s latest report reveals governments worldwide are scrambling to regulate AI technologies, including generative AI, agentic AI, and sovereign AI systems. The EU set the benchmark with the world’s first comprehensive AI regulatory framework, while the UK attracted £14 billion in AI investment and the EU committed €20 billion for AI gigafactories. My take: good luck. By the time you define a regulation, the AI you’re trying to control will be three leaps ahead. If you do manage to regulate it, you’ll just hamper your own economy. Remember the last time we did this, with the public cloud? I’m talking about Gaia-X, the “For Europe, by Europe” public cloud. How’d that work out for you, Europe? ☠️
