DTW Ignite 2026 was an excellent event and not just because Copenhagen in early summer is a near-perfect place to visit with colleagues and take stock of where telecom stands. I was most intrigued by the operator and analyst talks about the state of Open RAN and AI-RAN.
I had the opportunity to sit down with Manish Singh, Dell's CTO for Telecom Systems, on the sidelines of DTW to discuss the technical progress that has been made in the RAN recently and why commercial progress has been challenged.
That conversation starts with Massive MIMO. It’s the clearest example of a technology that did what it promised, then hit a wall that had nothing to do with engineering skill.
Since its debut, Massive MIMO has delivered two to four times the capacity gain in the areas where it’s deployed, with penetration now around 80% in the right conditions. That’s a genuinely good outcome for spectral efficiency, which is still one of the industry’s most expensive constraints.
Massive MIMO’s economics are most notable in mid-band, where antenna form factor, wavelength and other factors line up sensibly. Push it into low-band and the antenna requirements balloon, energy consumption climbs, and uplink starts limiting what you can deliver. It’s also a dense-area technology at heart. The value thins out fast in suburban or rural deployments. Massive MIMO has found its ceiling, and the next round of gains has to come from somewhere else. Enter AI, where distinct paths are forming.
Operators, vendors and boards keep lumping “AI in the RAN” and “AI on the RAN” together but they are not the same thing:
The industry has been a bit of a victim of the metrics it chose for itself. Take churn rate. A carrier can lose a thousand high-ARPU customers and gain a thousand lower-value ones, and churn rate looks perfectly healthy. ARPU is agnostic to margin, so you can win the highest-ARPU customer in the market and still lose money on them once acquisition cost and price competition are factored in.
That’s why we’re paying more attention now to lifetime value against cost of acquisition. That ratio actually tells you whether the business is heading somewhere good. Rakuten Mobile has been pushing on this. The Rakuten Symphony 2026 Industry Growth Report digs into why the old KPIs weren’t giving operators the full picture.
The logic needs to extend to how we account for AI itself. For example, token consumption isn’t a meaningful signal on its own. Plenty of that spend is just automated email drafting. What actually matters is something closer to an “AI EBITDA,” a real measure of whether AI investment shows up in the bottom line.
Governance is also an integral part. Any AI investment comes down to two questions. Where is your spend concentrated, and how ready is your data?
We’ve been stuck in the G-trap: 2G to 3G to 4G to 5G, and now 5G isn’t fully deployed and we are already talking about the next G. The industry’s leadership has historically come up through engineering, which is necessary, but we need more commercially led decision-making. We need to build technical roadmaps to serve business outcomes.
The clearest shift will be in how operators approach verticals. Health care, oil and gas, retail, logistics. These industries don’t care about Open RAN, but they care about downtime and outcomes specific to their business. Selling data bytes and minutes means nothing to a hospital network or a logistics operator. Selling less downtime does.
Enterprise isn’t a simple playbook like B2C. It demands understanding the vertical, the specific use case, where the data actually lives, and what value you’re delivering against a customer problem. Operators who build that now will capture the next wave of investment as 5G matures and 6G approaches. Listen to the conversation for more thoughts on this topic.
