#dotdot - Frictionless Wireless Connectivity for Incremental Revenue Generation. Extend network reach, enable opportunistic connectivity, and unlock new revenue streams — without deploying new infrastructure.

6–7, matcha latte, two-second attention span, Ramen lovers, five hours a day inside mobile apps, visual-first communication, loyal but always ready to change, perfectly fine consuming and creating AI-generated content.That is the generation telcos are struggling to understand. This is Gen Z. Born roughly between the late 1990s and the late 2000s, they are the first generation to have grown up entirely within algorithmic systems. No memory of life before smartphones. No novelty around connectivity. Always on, always available, always assumed. Their expectations were shaped by feeds that adapt in real time, services that anticipate intent, and tools that integrate creation, consumption, and monetization into a single flow. They experience technology as an environment. And that difference is what makes them so hard for telcos to read, price, and serve using models designed for an earlier era.

Super Bowl LX is not just the biggest sporting event in the world. It is the most extreme connectivity experiment ever deployed in a single venue. More than 35 terabytes of in-venue traffic are expected to be generated by roughly 70,000 spectators, each acting as a live content producer. This load is not driven by messaging or browsing, but by uplink-intensive behavior: high-definition video uploads, multi-angle replays, immersive applications, real-time social sharing, and continuous background streaming.To meet that demand, Levi’s Stadium is running the most ultra-dense stack that combines mmWave, mid-band 5G, Wi-Fi 7, private 5G networks, network slicing, satellite-backed mobile cells, and real-time AI-driven optimization. That is why Super Bowl LX has become the place where Verizon, AT&T, T-Mobile, Ericsson, and Nokia show everything they have. It is a full-stack stress test of the future network. Let us break it down layer by layer.

Deutsche Telekom announced that its sovereign public cloud inside Germany will match core hyperscaler features by 2026, backed by a €1 billion NVIDIA-powered AI data center designed to host industrial and regulated workloads. In parallel, Nokia is integrating GPU-based AI inference directly into the radio access network, pushing compute from centralized data centers into live mobile networks. At the same time, hyperscalers are expanding across multiple regions, zones, metro deployments, and sovereign stacks, bringing cloud execution closer to enterprises and machines rather than keeping everything centralized.Against that backdrop, global AI spending is approaching $2.52 trillion, with roughly 60% directed to infrastructure. Power, data centers, cooling, fiber, and silicon dominate the spend, not models or applications. This resembles previous infrastructure cycles in which capacity was built before demand clarity. Cloud is not reversing direction; it is expanding with the use of AI. As execution becomes more local and physical, local networks and infrastructure operations move back to the center of the AI economy.

On January 30, 2026, SpaceX filed with the Federal Communications Commission for authorization to operate up to one million satellites described as orbital AI data centers. The number dominated coverage, but the underlying driver received less attention.Global data centers consumed about 460 terawatt hours of electricity in 2024. Multiple forecasts from IEA and industry analysts converge between 1,200 and 1,700 terawatt hours by the mid-2030s, driven primarily by AI workloads. In Northern Virginia, the world's largest data center market, grid interconnection queues exceed 40 months. In Ireland and the Netherlands, permission for new campuses has already been frozen or delayed due to power and water constraints. Cooling now represents a larger share of capital expenditure growth than servers.

New research isolates an old foe — smoking — as the principal culprit behind U.S. midlife mortality gaps defined by place and education.

The AI value chain economics at this stage is focused on infrastructure. AI value creation is concentrated upstream, where capital intensity is highest and scale advantages compound most rapidly. The market rewards balance sheets, not clever prompts. Training and inference workloads demand dense compute, high-bandwidth memory, optical interconnects, and ample power. Those inputs scale poorly for small players and extremely well for incumbents.Gartner frames 2026 as the Trough of Disillusionment. Enterprises prioritize predictable ROI over experimentation. AI gets sold by incumbent vendors and bundled into existing contracts rather than procured as standalone moonshots. That behavior reinforces concentration. Buyers choose suppliers already embedded in their infrastructure stack.

The United States sets the global benchmark for telecom intensity. With over 300 million connections, deep smartphone saturation, and fiber builds colliding with fixed wireless, it remains the industry’s highest-stakes market. Pricing, retention, spectrum, and infrastructure decisions made here ripple worldwide. To a greater extent, AT&T, Verizon, and T-Mobile set the pace of telecom strategy worldwide.In 2026, these three players seem to be moving in sharply different directions. AT&T scales fiber to lock in convergence. Verizon executes a stripped-down reboot under Dan Schulman, targeting precision over volume. T-Mobile shifts to cash extraction mode, layering dividends and buybacks on top of its subscriber lead, all while pushing enterprise and network slicing into new territory. One market, three models, and a competitive field that leaves no room for mistakes.

96 $, c’est ce que coûte aujourd’hui la minute de calcul quantique chez un grand cloud provider. Le quantique n’est plus un fantasme de laboratoire : les machines sont disponibles, les tarifs clairs, les capacités démontrées. En d’autres termes, il suffit à un acteur malveillant, étatique ou criminel, d’estimer ce qu’il lui en coûterait de déchiffrer des données en sa possession et le bénéfice qu’il pourrait en tirer.

L’accord noué avec TerraPower, société cofondée par Bill Gates, retient particulièrement l’attention. Il prévoit le financement et le développement de deux réacteurs capables de fournir jusqu’à 690 mégawatts, dès 2032 si tout se déroule comme prévu. Mais ce n’est pas tout : six autres installations pourraient suivre, totalisant potentiellement 2,1 gigawatts supplémentaires d’ici trois ans. Particularité technique : ces sites reposeront sur les réacteurs « Natrium », utilisant le sodium comme fluide caloporteur en lieu et place de l’eau – un choix jugé plus prometteur par certains experts.

En pratique, ces QR codes frauduleux se nichent un peu partout : e-mails, documents numériques ou supports imprimés. Ils poussent insidieusement leurs victimes à délaisser les environnements sécurisés de leur entreprise au profit de terminaux personnels, souvent bien moins protégés. Un simple scan conduit vers de fausses pages imitant habilement des plateformes cloud ou des accès VPN. Le but ? Dérober des identifiants, contourner l’authentification multifacteur (MFA), puis ouvrir la porte à d’autres intrusions.

L’année 2025 aura marqué un tournant préoccupant dans l’intensification de la censure d’internet. Selon les dernières données recueillies par le spécialiste du secteur, Surfshark, plus de 4,6 milliards d’individus ont été affectés par de nouvelles restrictions. Pas moins de 81 mesures additionnelles ont été introduites dans 21 pays, un bond de 29 % par rapport à l’année précédente, faisant de ce millésime le plus restrictif depuis 2021.Au centre du cyclone, l’Asie concentre à elle seule la majorité des limitations. Le cas le plus emblématique reste celui de l’Inde, où les autorités ont instauré 24 nouveaux blocages, principalement lors de périodes troublées : dix en lien avec des manifestations et quatorze résultant d’agitations politiques. Les régions du Kashmir et du Jammu, déjà coutumières du fait, ont même subi un bannissement temporaire des VPN, preuve supplémentaire d’une volonté farouche de contrôler l’accès au réseau. Derrière ce géant démographique, l’Iraq et l’Afghanistan n’ont pas été en reste avec respectivement neuf et sept restrictions.

Loin du mythe de la rupture, l’usage de l’intelligence artificielle dans la cybersécurité s’oriente vers une efficacité accrue et un pragmatisme revendiqué. Comme le souligne Thierry Bedos (KEEPIT), « L’IA restera un outil pratique, centré sur l’automatisation et l’efficacité plutôt que sur la révolution ». Les entreprises adaptent ainsi leur stratégie, privilégiant la modernisation des outils, la montée en compétences des équipes et surtout la qualité des données exploitées.Ce réalisme s’impose d’autant plus que l’explosion des identités non humaines bouleverse déjà les modèles établis : « Les identités machine, déjà 80 fois plus nombreuses que les humaines, deviendront le premier vecteur de compromission dans le cloud », alerte Liat Hayun (TENABLE). Cette évolution appelle une refonte profonde de la gestion des permissions et du contrôle d’accès.

Rappelons-le, depuis 2015, le cadre européen garantit à chacun la liberté d’accéder sans discrimination aux contenus et services numériques de son choix. L’Arcep, chargée de veiller à cette application en France depuis 2016, observe que l’essor fulgurant des IA génératives pose aujourd’hui de nouvelles interrogations. Peut-on encore assurer une égalité d’accès pour tous face à des algorithmes qui filtrent, sélectionnent et synthétisent l’information ? Le rapport s’appuie ici sur une cinquantaine d’experts issus de différents horizons – public, privé ou associatif – ainsi que sur des tests menés avec le Pôle d’expertise de la régulation numérique (PEReN). Tous pointent vers un même constat : les conditions mêmes d’ouverture du web se trouvent reconfigurées par ces outils.

Most financial analysts frame 2026 as flat-to-low-growth revenues with improving margins. ING anchors the European view: sales growth around 2% and median EBITDA growth around 2.5%, with the spread explained by cost rationalisation rather than a demand step change.After a dip in 2025, analyst models show EBITDA growth rebounding above the four-year median in 2026, reflecting the delayed impact of restructuring, automation, and cost rationalisation programs rather than any improvement in underlying demand or capex intensity. Even that 1.5% to 2% revenue growth is a grind. In many competitive markets, operators fight ongoing high price erosion while discounts and value segment pressure keep ARPU fragile. Penetration is above 100% in most countries, which doesn’t leave room for “new customer growth” but rather forces a sum-zero game. That's an expensive endeavour for the consumer market.Analysts call out exactly that dynamic, with upside coming from bundles and selective price increases, and downside coming from competitive discounting and weak ARPU markets.

Davos 2026 marked a clean break for the telecom industry. The industry shifted the discussion away from speed, coverage, and generational labels toward control, autonomy, fragmentation, capital, and geography. Telecom CEOs converged on one reality: Networks are no longer passive infrastructure; they are becoming intelligence systems operating under geopolitical pressure, financial stress, and physical limits. What emerged in the Alps were 5 narratives that will set the roadmap for Telcos in the coming years.

On January 21, the U.S. Supreme Court will hear oral arguments in Trump v. Cook — a case that will, at minimum, decide whether the president has the power to fire Federal Reserve Board of Governors member Lisa Cook. But according to Harvard Law Professor Daniel Tarullo, the Court’s decision could have much broader impacts. “Cook is really the test case of whether there is any limit to untrammeled power for the president,” he says. “Thus, one way or another, I think it will end up being one of the most important decisions of the Roberts Court.”

Telcos attempted multiple monetization strategies without achieving any change in outcome. Quality tiers, zero rating, sponsored data, and later network slicing promised differentiation but failed to command durable price premiums. Enterprises adopted private networks selectively, yet volumes remained insufficient to move group-level revenue. Latency improvements delivered technical gains but limited willingness to pay, since applications abstracted transport and captured the value layer.Cost per bit continued to fall faster than achievable price increases, driven by silicon scaling, higher-order MIMO, densification, and fiber economics. Transport approached marginal cost behavior while remaining capital-intensive.

For most of its history, telecom venture capital was symbolic. That changed after 2018 and accelerated sharply after 2022, as tier-one operators and vendors began treating venture investing as a structural tool rather than innovation theater. Competitive pressure, AI-driven compute shifts, and geopolitical risk forced telcos to engage with emerging technology earlier and more deliberately. Today, the ten most relevant telco and vendor-backed venture platforms collectively deploy or manage an estimated $10B to $15B, with annual deployment of roughly $1.5B to $2B. Capital is highly concentrated, and portfolios are tightly scoped around network software, AI infrastructure, cybersecurity, edge compute, silicon adjacencies, private networks, enterprise platforms, and regulated data services rather than consumer growth plays.

Last week, Bob Sternfels, CEO of McKinsey & Company, stated that the firm has 60,000 employees and that 25,000 of them are AI agents. Not tools or copilots but employees. The wording was deliberate as it places AI agents in the same category as human workers inside an enterprise. That is not a visionary statement; It is a technical bullcrap. Counting AI agents as employees implies equivalence in autonomy, judgment, accountability, and decision ownership. Anyone with a serious background in AI and machine learning knows this equivalence does not exist today, and is not even close to it. What exists are probabilistic systems wrapped in orchestration layers, heavily supervised by humans, brittle under long-horizon execution, and fundamentally incapable of owning outcomes.