Totogi’s wins are building, per Alex Davies of Rethink Research. In his post-DTW Ignite analysis (paywalled), Davis notes that BSS Magic can replace legacy BSS from vendors like Amdocs, Ericsson, and Netcracker “in a shockingly short amount of time, with very small manpower requirements.” Davis further notes that AI is both creating new opportunities for new players like Totogi, and also spotlighting how behind the incumbents are. Remember, BSS Magic isn’t a BSS replacement; it’s an AI layer that unlocks your entire BSS estate. Wanna give it a try? Give me a call.
Gartner predicts that over 40% of agentic AI projects will be abandoned by 2027 due to high costs, limited business value, and poor risk management. They’ll get stuck in pilot phases and hampered by underestimated complexity. (That’s why you should start small—read my blog about it.) Legacy vendors will “agent wash” older tools like chatbots, calling them agentic AI. Still, even with early setbacks, Gartner sees strong long-term potential, forecasting that by 2028, 15% of daily work decisions and a third of enterprise applications will incorporate agentic AI. Want to accelerate your AI roadmap? Start with business value and a real pain point, then use Totogi’s BSS Magic to connect to your legacy systems. You’d be surprised what you can do!
Apparently, private “AI factories” are all the rage among enterprises. I think it’s a mistake. Hyperscalers are building massive public data centers, while companies like Teradata are pushing private, on-premises environments. Why go private? Organizations want to keep everything on-prem to fine-tune models, maintain data security, and meet sovereignty requirements. But here’s the problem: You're still going to need the public cloud. You need their chips, their scalability, and their AI software and services to do what you actually want to do, and do it in a pay-as-you-go framework. I say ditch the private cloud and go public cloud all the way, baby!
You know who isn’t building a private AI factory? OpenAI. Instead it’s tapping its rival Google for public cloud compute capacity, despite being direct LLM competitors. OpenAI has dealt with increasing demand for computing capacity since ChatGPT launched and can't get enough—so it’s turning to Google's TPUs in addition to the Azure resources it already uses. If the world’s biggest AI company isn’t building its own infrastructure, what makes you think you should? The hyperscalers have the chips you need. Use them!
Steve Saunders went on an anti-public cloud rant in his Fierce Network newsletter from June 20. He recommended two public clouds—Huawei and Oracle—to avoid the three big hyperscalers. Trouble is, the clouds he promotes have no shot as they don’t have the services, scale, or the CapEx investment that telcos need. Remember: don’t use the public cloud as a data center. It’s enterprise software with infinite scaling capacity, priced so that you pay as you grow. Know the difference.
Why do hyperscalers want to understand telco standards? To get those telco network workload on their clouds! As operators increasingly adopt public cloud for daily operations and network automation, staying aligned with industry standards is crucial. Google Cloud’s Angelo Libertucci (coming soon to the podcast!) highlighted his group’s role in helping define network autonomy levels and lending global network expertise, while Dr. Ishwar Parulkar emphasized AWS’s dedicated telecom division and investments to support telcos running network functions in the cloud, like Nokia’s 5G core for Dish and Telefónica Germany. We now have the Big Three hyperscalers all supporting network operations on the public cloud. When are you going to give it a shot and try it for your network? You could save a bunch of money...
Data center capacity is shifting rapidly to the hyperscalers, according to a report from Synergy Research. Hyperscalers now operate 1,189 data centers, which accounts for a whopping 44% of worldwide capacity. Only 34% remains on-prem, down from 56% six years ago. And Synergy projects that hyperscalers will control 61% of global capacity by 2030, with on-prem shrinking to 22%. I know there’s sentiment right now that organizations are repatriating their workloads because of AI sovereignty, but Synergy’s report also says GenAI and graphical processing unit (GPU) demand are only temporarily boosting on-premise growth—not enough to reverse the overall decline. Long live public cloud!
Speaking of expanding the hyperscaler footprint, SK Group, SK Telecom’s parent company, and AWS are teaming up to build a 100 megawatt, AI-focused data center in Ulsan, South Korea, that will form the backbone of an AWS “AI zone” and become a key component of SK Telecom’s AI infrastructure superhighway strategy. Aimed at boosting data sovereignty and AI competitiveness, the facility will support local AI workloads and offer services like Amazon SageMaker, Bedrock, and Amazon Q. Construction starts this September, with operations expected by 2027. I don't know if this is an AI sovereignty thing, but I do know the best way for AWS to boost revenue growth is to open more regions. Expect more announcements like this happening all over the world...
Beware of NVIDIA’s AI-RAN pitch, says Dr. Ishwar Parulkar of AWS, the hyperscaler’s telco CTO. Nvidia’s concept envisions telecom operators buying and deploying Nvidia GPU chips in their radio access network (RAN) sites to both run mobile network functions and lease GPU capacity for AI applications—aiming to evolve beyond traditional ASIC-based RANs and compete with public cloud giants. Recent podcast guest Parulkar argues that simply installing GPUs won’t get you there without a robust software stack, tools, and operational chops, which the hyperscalers have—and I agree. File this idea under: "hard and expensive to execute," and PASS for now.
... and finally, brava to BT Group CEO Allison Kirkby who’s signaling that AI may drive more job cuts than the 40,000-55,000 by 2030 previously forecast. Kirkby is one of the few execs telling the truth, and that’s good leadership. It’s important to be honest with your people and get them ready for reality. Hopefully, BT is also proactively training people to use AI, so they gain valuable skills to take to their next gig. Read more about my perspective on this in my recent blog, AI will cost jobs, lots of them. Stop lying about it.
Happy Fourth of July! 🎆
