Satya Nadella just published the most important thing a hyperscaler CEO has said all year, and telco execs, you should read it TWICE. His thesis: in the AI era, the moat moves off the model and onto the learning loop built on top of it—what he calls “token capital.” His ownership test is brutal: can you swap the underlying model without losing your “company veteran” expertise? If not, you don’t own your IP. You’re renting it. I’ve been making this exact case for telco since 2024. A telco’s token capital lives in an operational ontology—the encoded logic of how YOUR company actually works: billing rules, provisioning states, eligibility constraints. A layer YOU own, that gets smarter with every churn save, every order, every network fault, and survives any model swap underneath. A Copilot seat won’t get you there. Neither will GPT-5 piped in through Amdocs. Both are model access. Token capital lives ABOVE the model. The Totogi Ontology is exactly the architecture Satya is describing, built specifically for telco. Read his post. Then ask: where’s your learning loop?
SK Telecom invested $100 million in Anthropic in 2023 at a $5 billion valuation. Anthropic is now approaching a $1 trillion valuation, making SKT’s stake worth north of $5 billion against a total market cap of $11.5 billion. Morningstar just raised SKT’s fair value 40% and recategorized it as an AI stock. Not a telco stock. An AI stock. One early bet, made at the right moment, changed how the entire market sees the company. Kind of makes me sad I didn’t buy some SKT stock in 2023…! 📈
For three glorious days, the world got a taste of what AI can really do. On June 9, Anthropic launched Claude Fable 5, the most capable public model ever shipped: 80.3% on SWE-Bench Pro, better than nearly every human software engineer alive. Then on June 12 the US government issued an export-control order, and Anthropic pulled Fable 5 and Mythos 5 worldwide within hours. Gone. Every other Claude model stayed online, but the best one (by a very wide margin) vanished overnight over a national security dispute. Here’s hoping this gets resolved fast, because Fable was insanely good. We all just saw the future for 72 hours. Bring back Fable!
BT just joined Anthropic’s Project Glasswing, and is now the second telco inside the program alongside Verizon, with access to Claude Mythos Preview (currently suspended; see above), the model that found a 27-year-old OpenBSD vulnerability nobody had ever detected. Smart move, but evaluating which BT code, exactly? Telcos don’t write their own software. They run vendor solutions where they have zero source-code access. BT can scan whatever it has built in-house. But the Ericsson, Nokia, Cisco, and Samsung stacks running the RAN, core, transport, and OSS are closed boxes the operator has never seen inside. If a frontier model can find a 27-year-old zero-day in open code, what’s sitting in the closed code running your network? You can’t audit what you can’t access. How will that problem be solved?
Cloudflare recently revealed that agentic AI traffic on the internet has surpassed human traffic for the first time. Simultaneously, NVIDIA announced chips that run AI agents directly on Windows laptops, putting agents on-prem, in the cloud, and on every device at once. F5’s VP of AI said the danger isn’t just the volume; it’s that ALL traffic is now a potential attack surface. As mentioned above, Anthropic’s Fable 5 was public for three days before the government pulled it, but not before a jailbreak proved the same model capability that finds zero-days in hours could be turned loose, to defenders and attackers alike. WHOA!
Verizon CEO Dan Schulman told Bloomberg that AI will replace “a large percentage” of customer service work. No hedging, no caveats. Good. I love how he’s being honest. But the framing most operators reach for, that “AI will assist our agents,” misses the point entirely. The goal isn’t to make the old job cheaper or faster. It’s to give AI the work it can genuinely do better and free humans for the work only humans can do: judgment calls, escalations, retention conversations that require human empathy. The key is to redesign the job; don’t just automate it.
Meanwhile, T-Mobile US launched Dynamic CX, AI that scans public event data, anticipates crowd surges, and pre-positions spectrum before congestion hits, timed for the FIFA World Cup across 11 US host cities. This works, and it’s worth understanding why. T-Mobile has a largely homogeneous network where RAN telemetry speaks one language across every cell, and the AI has clean inputs to act on. Most operators don’t have that. A mixed Ericsson/Nokia RAN estate has the same fragmentation problem as your BSS: different vendors, different schemas, different definitions of the same metric. T-Mobile’s AI win is real. It’s also a reminder that AI only performs when the data layer underneath it speaks one language. ⚽
TelecomTalk just published a list of where AI is actually delivering in telecom: predictive maintenance, fault detection, network automation, revenue leakage—and Totogi made the list. The takeaway everyone missed: the bottleneck isn’t the data. What’s missing is the company bible: the canonical layer that tells AI what “customer” actually means across 180+ systems, what rules govern every decision, and what actions are valid in which context. Without it, you’re feeding your AI contradictions and wondering why it hallucinates. That’s the problem the Totogi Ontology solves. That’s why Totogi makes the list. Give it a try!
AT&T’s CFO called SpaceX the “elephant in the room” and then dismissed it: satellite is for the 1% rural uncovered; urban and suburban belong to fiber and wireless. The math is right. But analysts are looking at $SPCX as part-telco, because Starlink is the actual business. AT&T is optimizing cost-per-bit. Elon is optimizing for Mars. Your move, Elon. 🛰️
SpaceX priced at $135/share—a $1.77 trillion valuation and a $75 billion raise, the largest IPO in stock market history. It opened at $150 and closed at $160.95. The stock is the headline; the capital is the story. It will go into Starlink infrastructure, direct-to-device buildout, and spectrum—the one profitable business in the SpaceX portfolio that actually competes with telco revenue—with a side of AI revenue in the form of SpaceXAI. Elon just raised the capital to accelerate the business that operators should fear most. Watch out, telco. 🚀
