“Personne ne l’a encore jamais fait”. Emeric Lhomme mesure l’ampleur du défi qu’il s’est lancé avec Alatyr. À la tête de cette start-up française de 11 salariés, cet ancien d’ArianeGroup ne veut pas seulement déployer en orbite des data centers dédiés à l’intelligence artificielle générative, à l’image des projets portés par Google ou encore SpaceX. Il ambitionne d’aller plus loin: assembler dans l’espace, à l’aide d’un robot, plus de 15.000 mètres carrés de panneaux solaires pour alimenter en énergie les GPU chargés d’entraîner ou d’exécuter les modèles d’IA. “Un défi d’ingénierie, mais aussi de financement”, reconnaît-il.

Jamais le marché des smartphones n’avait connu une telle dégringolade. Plombées par l’envolée spectaculaire des prix des puces mémoire, les ventes ont chuté de 22% au troisième trimestre, selon les estimations d’IDC. Un plongeon bien plus marqué que ne l’anticipait le cabinet avant l’été. Et le pire reste à venir, prévient-il: le recul devrait s’accentuer au quatrième trimestre, qui comprend la période cruciale des fêtes de fin d’année. Sur l’ensemble de 2026, IDC table désormais sur une contraction historique de 16,7%, qui ramènerait le marché à son plus bas niveau depuis 2013. Le rebond, lui, n’est pas attendu avant 2028.

On October 22, Bloomberg reported that SpaceX is negotiating access to at least 80,000 cellular sites across the United States, including discussions with Crown Castle. This is roughly half the country’s 158,500 purpose-built towers, although not every cellular site is a traditional tower. The ambition is becoming obvious. Musk is not trying to build a better satellite service. He is preparing to compete with AT&T, Verizon and T-Mobile as a fourth nationwide mobile operator.

Twenty years after the PayPal Mafia helped shape the internet economy, a small group of telecom companies is making a similar bet on artificial intelligence. NVIDIA is bringing them together, and some are already making serious money. But who will actually get rich?

"We will also have direct-to-cell, high-bandwidth direct-to-cell capability, and we will be building out our terrestrial network.” Musk was not describing Starlink as a satellite-only mobile service. He was describing a future network that combines space and ground infrastructure, with broad connectivity anywhere SpaceX is allowed to operate. That is important because one of the most common arguments I hear against Starlink becoming a serious mobile competitor is indoor coverage. Satellites struggle to reach phones deep inside buildings, and much smartphone usage happens indoors. Therefore, the argument goes, existing mobile operators have a natural protection that Musk cannot easily overcome. The physics are correct; however, I think the business conclusion is wrong. Most heavy indoor smartphone traffic already runs over WiFi. The fact that people use their phones indoors doesn't mean the cellular network carries all those bytes. Add WiFi to the picture, then add the terrestrial network Musk says SpaceX intends to build, and indoor coverage starts looking much less like a defensive wall around the mobile industry.

SpaceX reported $14.6 billion in revenue in the third quarter of 2026, up 161% from the prior year. More revealing is the explosion of the AI business. Its superintelligence division produced most of the growth, while connectivity generated roughly 85% of operating profit. SpaceX is building a business in which rockets lower the cost of connectivity, connectivity brings customers and profits, and compute opens another market. For the Telecom industry, the question is where that combination shifts control. SpaceX can supply a Telco with coverage while developing the network, applications, and direct customer relationships that could eventually weaken that Telco´s position.

In mobile communications, without spectrum you are nothing. Yes, nothing. You can have the most advanced satellites, billions of dollars, revolutionary antennas and incredible software, but without the legal right to transmit on usable frequencies, you cannot operate a commercial mobile network. You need spectrum that works with the devices you want to connect, in the countries where you want to operate. Otherwise, you need somebody willing to let you use theirs. Elon Musk understands this very well. Over the past year, SpaceX has moved aggressively to secure spectrum rights, including its approximately $20 billion agreement with EchoStar and, more recently, an agreement to acquire up to 14 MHz of nationwide 800 MHz spectrum in the United States. These are not random investments. They are steps toward a business that could compete with traditional mobile operators without starting with all the infrastructure and legacy obligations of a traditional telco. Now European operators appear to be learning from what is happening across the Atlantic. On October 19, Deutsche Telekom, Orange, Telefónica and Vodafone announced a joint initiative to explore acquiring satellite spectrum rights across Europe. My interpretation is that they are preparing for the possibility that Musk, or another satellite operator, eventually decides that selling wholesale satellite coverage to telcos is less attractive than taking a larger share of the mobile business. And this time, the telcos want to be in a position to negotiate.

Every industry has its urban legends. In telecom, we put ours into PowerPoint presentations, attach a five-year forecast, and ask the board for a few billion dollars. I have spent so many years watching vendors promise that the next technology will bring new revenues, lower costs, and happier shareholders. Operators buy the equipment, deploy the networks, and eventually discover that customers never agreed to the business case. With 6G, AI, satellites, and fiber, some familiar stories are coming back. Here are ten of my favorites, and why I don’t buy them.

Telecom operators have been cutting jobs for decades. As technology evolved, we automated network operations, outsourced field maintenance, consolidated customer care, closed retail stores, and eliminated entire departments. With revenues struggling to grow, reducing the workforce became one of the industry’s favorite ways to protect profitability. Now we have artificial intelligence, and suddenly the same old strategy has become an AI transformation story. Apparently, reducing employees is no longer just cost-cutting. It is evidence that your company is becoming AI native.

On October 12, 2026, hackers disrupted refrigeration systems across France. Around 700 supermarkets were affected; some refrigerators reportedly reached 50 degrees Celsius, and Intermarché reported losses running into tens of millions of euros. Software controlling ordinary equipment had become a way to destroy physical inventory. This was a refrigeration controller attack, not an AI sensing attack. But imagine a 200-pound humanoid carrying equipment beside workers, or an autonomous car approaching a crossing. An attacker might not even need to control the machine. Manipulating what it sees or believes about its surroundings could be enough to produce a dangerous decision. For telcos, this opens an interesting question. Could networks help protect Physical AI by independently verifying what machines see? And could that become a new business beyond simply connecting them?

Next week, investors will be asked to value Jio Platforms at about $108 billion. The company has 533 million mobile customers, generates almost all its revenue in India, and collects an average of just $2.27 per mobile customer per month. Those aren't the numbers you'd normally associate with one of the world’s most valuable telecom companies. But Mukesh Ambani isn't simply selling a mobile operator story. Instead, he is selling the possibility of a national digital platform built on a mobile network: AI, cloud, entertainment, enterprise software, connected homes, and distribution for some of the world's largest technology companies. If that works, the network becomes the entry ticket, not the final product. And Jio might become the example other telcos have spent twenty years talking about without delivering. The question is whether investors are already paying for that platform or just for an exceptionally good telecom business.

Emerging healthcare applications including real-time diagnostics, remote monitoring, and computer-assisted surgery demand strict Quality of Service guarantees while managing limited edge resources. Current Internet of Medical Things (IoMT) orchestration approaches prioritize communication metrics like latency and bandwidth but neglect edge resource constraints and dynamic adaptation. This paper introduces a novel resource-slicing-based adaptive orchestration framework integrating 5G network slicing, resource-aware orchestration, and mobility-aware task offloading. The system assigns IoMT services to distinct 5G network slices while jointly considering CPU capacity, memory, and storage availability, enabling more efficient service placement. It additionally supports user mobility and dynamic task offloading to maintain service continuity under changing conditions. SimPy-based simulation validation demonstrates the method enhances deployment efficiency, reduces request blocking, and improves user acceptance across diverse IoMT scenarios. This work addresses critical challenges in distributed edge computing environments where traditional cloud-centric approaches suffer from high latency and connectivity dependency, while edge resources remain limited and heterogeneous despite their proximity advantage.

50G PON represents the third generation of passive optical network technology, standardized by the ITU-T G.9804 series and also known as G.HSP or Higher Speed PON. It delivers a significant capacity upgrade over the widely deployed GPON and XGS-PON networks, offering 50 Gbps downstream and 25-50 Gbps upstream capabilities. This technology enables advanced services including 8K ultra-high-definition video, Industrial Internet applications, and AI computing. Following GPON's introduction around 2007 and XGS-PON's deployment circa 2016, 50G PON maintains backward compatibility with earlier generations through strategic wavelength planning, allowing operators to gradually migrate without requiring complete infrastructure replacement. Each PON generation has pursued increased bandwidth per subscriber while utilizing existing fiber infrastructure. While XGS-PON is expected to remain dominant through 2026, 50G PON achieves its fivefold downstream speed increase through advanced digital signal processing and enhanced forward error correction, maintaining performance over the same 20-kilometer reaches and 1:64 split ratios as predecessor technologies, making it crucial for operators planning next-generation access network upgrades.

Equatys, a new direct-to-device mobile satellite venture, has launched with a $1 billion equity commitment from Space42 and Viasat to address dropped calls for mobile network operators. Announced at the World Space Business Week in Paris, the company is developing a shared satellite infrastructure model leveraging 3GPP non-terrestrial network standards to enable seamless handover between cellular and satellite coverage. The service utilizes S-band and L-band spectrum to integrate satellite and mobile networks without requiring users to manually switch services or pay premium rates. Space42 CEO Ali Al Hashmi emphasized that the approach mimics a TowerCo model in space, providing pooled spectrum resources that benefit MNOs and their customers while maintaining operator sovereignty and avoiding competitive threats. Equatys plans to build a low-Earth orbit constellation and ground infrastructure for this service, which currently does not exist in the market. The collaborative approach enables economically feasible satellite connectivity directly from smartphones, positioning the venture to fill coverage gaps while maintaining seamless user experience aligned with evolving 6G standards.

Sateliot, a European satellite telecommunications operator, is launching five new satellites on SpaceX's Transporter-18 mission this Thursday to expand its 5G NTN IoT network. The Draconis mission, lifting off from Vandenberg Space Force Base aboard a Falcon 9 rocket, will bring Sateliot's total constellation to eleven satellites. The new units, manufactured by Alén Space and measuring 20 x 10 x 35 centimetres at 10 kilograms each, feature significant technological enhancements including optimized payloads, upgraded onboard computers, improved attitude control systems, and radiation-resistant memory. Orbiting at 510 kilometers altitude with a five-year operational lifespan, these satellites will double daily connection passes while halving revisit time for IoT devices in areas lacking terrestrial coverage. The improved platform design reduced manufacturing time by fifty percent compared to the 2024 mission. Exolaush handles mission integration, while KSAT, Leaf Space, and Comfone provide ground segment support. This deployment advances Sateliot's commercial strategy of offering wholesale connectivity services to mobile operators in remote regions.

Post-quantum cryptography (PQC) and quantum key distribution (QKD) represent complementary defenses against quantum computing threats to current encryption standards. This comprehensive survey organizes hybrid PQC–QKD approaches that combine both techniques to ensure security even if one component fails. The authors establish a taxonomy across eight dimensions including integration layer, combiner construction, authentication scope, and QKD topology, then evaluate published protocols such as Muckle, MACsec and TLS integrations, and IKEv2 mechanisms standardized in RFCs 8784, 9370, and 9867. The analysis examines key-combiner constructions against NIST, ETSI, and ITU-T guidance while mapping security arguments onto common evaluation frameworks. Deployment evidence from metropolitan and national QKD networks distinguishes operational testbeds from pure-QKD implementations. The survey addresses standards and policy positions, including national security agencies' recommendations against relying on QKD alone. Key open challenges include authenticating QKD classical channels, shared-infrastructure threat models, device certification versus formal proofs, and verification of hybrid handshakes. This synthesis provides essential guidance as organizations prepare for quantum-resistant cryptographic transitions.

Over the past four months, I’ve advised several Physical AI companies on different topics, including communications. It’s given me an interesting perspective on an industry moving surprisingly fast. Delivery robots are navigating sidewalks, drones are inspecting infrastructure and delivering packages, and robotaxis are expanding into more cities. And these are only the first applications. Over the next decade, we could have millions of autonomous machines roaming around our cities, performing jobs that until recently required humans. But there is an interesting question behind all this excitement. How the hell do you train all these little guys to operate in the real world?

There was a time when computing was so expensive and complex that it had to run on massive centralized machines called mainframes. Then advances in hardware, software and applications made it possible to move computing into personal computers. That shift created an entirely new industry and changed the economics of technology forever. Does the story sound familiar?

Telcos have spent enormous effort improving customer service, yet something as simple as broken roaming, a billing error, or an eSIM problem can still turn into a nightmare. Customers no longer compare their telco with another telco, but they compare their experience with Amazon, Apple, Uber, their bank, and every other digital service they use every day. Against that benchmark, telecom has a problem.

Wall Street wiped tens of billions of dollars off the combined market value of Verizon, AT&T, and T-Mobile after SpaceX announced its latest spectrum acquisition. Investors are suddenly worried about Elon Musk building a fourth mobile operator in the United States. But I think another problem is developing, and it could affect the industry much sooner than most people expect. The United States is preparing one of its largest spectrum release cycles in years, with at least 800 MHz covered by auction mandates extending through 2034. Verizon, AT&T, and T-Mobile will need to decide how much additional spectrum they require for 5G and eventually 6G, while SpaceX is assembling the assets to become a credible competitor in parts of the mobile market.