Tim Cook turned Steve Jobs’s vision into a $4 trillion utility. Now, John Ternus must do the unthinkable: dismantle the most successful product in history to find Apple's soul in the age of AI.

If you thought Generative AI was wild, you are completely misjudging the scale of what comes next. GenAI is software that replaces clicks and synthesizes text. Physical AI is another league entirely. It is software entering the physical world to directly enhance, or replace, human labor.
We are talking about a direct disruption of a $50 trillion global labor income market, and it is moving at a frenetic, violent pace. In just a few years, telecommunications will face an entirely new customer segment, undoubtedly the biggest and most important one in history. And no, you cannot sell them gigabytes per month!
But right now, the industry is blind to what this new segment is like, what they want, how they buy, and where they live. We need to profile this new persona.

Last year’s US-imposed tariffs sped up significant trade shifts—toward Mexico and away from China—that began years earlier and have diversified American imports among top partners.
While the Trump administration’s major tariff announcement on April 2, 2025, was billed as “Liberation Day,” research by Harvard Business School’s Laura Alfaro suggests that companies were already positioned to adjust to the levies. The recalibration of supply chains has been so profound that US imports from China have returned to near-2001 levels, when the country entered the World Trade Organization.

The network edge is suddenly a very valuable real estate in the digital economy. And right on cue, the telecom industry has fractured into a bitter, trillion-dollar civil war over what the “edge” actually means, where it lives, and who will pay for it.
On one side of the battlefield, we have the “Believers.” Fueled by the relentless ambition of silicon giant Nvidia and championed by operators like T-Mobile and SoftBank, this camp views the base of every cell tower as a potential goldmine. They are pushing “AI-RAN”, a controversial architecture that seeks to drop heavy-duty GPUs at the absolute far edge of the network to handle wireless processing and enterprise AI simultaneously.
On the other side stand the “Doubters.” Led companies such as AT&T and Verizon, and was quietly supported by vendors like Ericsson. This camp views the far edge (understood as Radio) as an economically ruinous CapEx trap. Grounded in the laws of thermodynamics, silicon efficiency, and optical physics, they argue that throwing expensive hardware at a cell site and praying for enterprise customers to materialize is commercial suicide.

When Huawei released its 2025 financials, the eyeballs were not on the massive $128 billion top line but on the 2.2% growth rate. For a company that historically ran laps around its rivals, normalizing to the standard telecom industry sluggishness reality is a concerning shift. However, the real signal came in the segment’s performance: while the global cloud market compounds at over 20%, Huawei’s cloud revenue actually shrank by 3.5%.
Unless you have been living under a rock, you know the Huawei challenge. Years of Western sanctions, hardware bans, and systematic exclusion from every cool alliance or partnership. Huawei’s response has been brute-force vertical integration. They are dumping $28 billion into R&D, roughly 10x what Ericsson or Nokia spends, to build a parallel stack across silicon, cloud, and AI. It kept them alive. But with the 2025 data in hand, we need to ask a different question. Is this a company stabilizing, or are we finally seeing the financial ceiling of operating in a closed ecosystem?

There is a brutal, silent war happening at the atomic level of the Telco´s network, and it is not about 5G speeds or 6G hype cycles.
It is a battle for the “math layer” or the specific silicon architecture that converts radio waves into money. While the global Radio Access Network is a $35 billion market, it acts as the high-stakes “butterfly effect” for a $1.3 trillion industry. A single decision made today about a DU chipset could dictate the power consumption of entire nations and decide whether the future of telecom is a specialized machine, a cloud workload, or a distributed NVIDIA-powered AI factory.
For decades, this war was latent, tucked away in the proprietary “black boxes” of a few Scandinavian and Chinese giants. But the arrival of NVIDIA reframing the base station as a micro-AI factory, Intel’s gravity-well grip on virtualized RAN, and the ideological chaos of Open RAN have finally rocked the boat.

Artificial intelligence is not a telecom strategy. It is the fastest and most unforgiving way to discover whether one exists. The assumption behind much of the current discourse is that intelligence can be added to an organization in the same way capacity or software is added. Deploy a model, integrate it into workflows, automate decisions, and the organization becomes smarter. That logic is appealing because it frames AI as an upgrade. It avoids confronting the structure of the business itself. But organizations are not empty systems waiting to be enhanced.

Derrière ces décisions commerciales se cache une réalité industrielle complexe. Le marché mondial subit de plein fouet une véritable pénurie de puces mémoire, alimentée par la forte demande émanant notamment du secteur des serveurs pour l’intelligence artificielle. Parallèlement, les capacités de production peinent à suivre. Les fournisseurs ne peuvent augmenter leurs volumes aussi vite qu’ils le souhaiteraient. De surcroît, la flambée du prix des matières premières, métaux ou énergie, accentuée par les incertitudes géopolitiques actuelles, alourdit davantage la facture.

It might sound like déjà vu, but it isn’t. According to recent Reuters reports, Amazon is quietly preparing to reenter the smartphone market under the highly classified “Project Transformer.”
Twelve years after the catastrophic failure of the Fire Phone, the company is attempting a surgical hardware reboot. This time, the strategy is radically different. Rather than engaging in a head-to-head battle with the Apple and Google app store duopoly, Amazon is pivoting to an entirely new category: an AI-assisted “anti-smartphone.” This minimalist, agentic dumbphone is not designed to compete on hardware specs or app ecosystems; it is engineered strictly as a frictionless tether to accelerate Amazon’s proprietary flywheel of e-commerce, advertising, and Prime media.

NVIDIA’s ambition is not to conquer a relatively small, low-growth niche business. As is typical of CEO Jensen Huang’s playbook, the strategy is far more visionary. NVIDIA is putting immense effort into telecom because it is the necessary physical gateway to a much larger prize: a $500 billion top-line opportunity driven by the automation of the physical world. They aren’t trying to build better radios; they are building the global nervous system for Physical AI.

In the early 2000s, Linux altered the balance of power in computing. It dismantled the tight vertical integration model that vendors like Sun and IBM relied on and replaced it with a shared software foundation that ran on almost any hardware. The operating system stopped being a proprietary control point and became a common infrastructure. Value moved up the stack. Ecosystems expanded. The industry shifted from closed systems to programmable platforms.
Two decades later, a similar attempt is unfolding in telecom. A Western coalition led by the Linux Foundation, seeded by US FutureG funding and joined by NVIDIA, DeepSig, Ericsson, and Nokia, is pushing to create a “Linux of RAN” through the OCUDU initiative. The target is the CU and DU layer, the programmable heart of the radio network. If this effort gains real deployment, it could reshape a 35 to 40 billion dollar RAN market, open the baseband layer to new entrants, and accelerate a structural split in 6G between Western and Chinese ecosystems.

Create a living AI version of you, or whoever you want to be.
Your AI Self talks, posts, remembers, and grows, so you can live your best life—without all those silly human limits.

Between 2010 and 2020, IoT was positioned as telecom’s next growth engine. Forecasts projected tens of billions of devices, recurring revenues, and deep enterprise integration. The volume materialized, but the value did not.
By the end of 2025, global IoT connections reached roughly 21 billion, according to IoT Analytics. Cellular IoT alone accounts for approximately 4.5 billion connections and is projected to reach 8 billion by 2031, according to the Ericsson Mobility Report. On paper, scale exists. IoT connections are projected to double by 2031, but most growth comes from short-range technologies rather than cellular.
However, the economics of this business tell a different story. Cellular IoT ARPU in many markets ranges from $1 to $5 per device per month for NB-IoT and LTE-M deployments. Even broadband IoT modules, such as LTE Cat 1, operate at narrow connectivity margins. Compare that to satellite IoT, where legacy plans often range from 40 to 70 dollars per device per month. Volume versus premium.

Companies and researchers often fail to disclose negative trial results, resulting in significant gaps in the public record and a publication bias that obscures the true landscape of drug development outcomes—overrepresenting successes and underrepresenting failures. This gap can also create a distorted perception of the safety and efficacy of medical products.

The genius of Bezos gamble and the controversy of it lies in Globalstar’s S-band spectrum (2.4 GHz). In the telecom world, spectrum is the “land” upon which digital cities are built. While Starlink relies on high-frequency Ku and Ka bands (perfect for high-speed home internet but terrible for penetrating walls or connecting to small antennas), the S-band is the “Goldilocks” frequency. It is globally harmonized, meaning a device using it can work in London just as easily as in Tokyo without changing hardware. More importantly, it is the precise frequency needed for Direct-to-Device (D2D) connectivity.

Telecom voice is basically unusable these days. Globally, over 50% of all phone calls are now deepfakes, scams, or extortion attempts. Here in Mexico, that number easily clears 60%. The result is a massive behavioral shift; people simply do not answer the phone anymore. We’ve retreated to the safety of known WhatsApp contacts, leaving the legacy voice network to operate as a $41 billion fraud machine. It’s pretty sad when the ultimate fate of a century-old technology is a society that just lets it ring to voicemail.

Des centaines de navires apparaissent au-dessus d'aéroports ou de centrales nucléaires sur les écrans de navigation. L'Iran brouillerait massivement les signaux GPS dans le détroit d'Hormuz, et le trafic maritime commence sérieusement à en pâtir.

I was reading an interview with Yigal Elbaz, and he said something that by now is an open secret: “We need to completely decouple the ‘Gs’ or the cycle of the ‘Gs’ from our ability to innovate.”
He’s calling for continuous innovation and progress instead of these arbitrary ten-year blocks. No hard feelings with Gs, but just the reality of running a modern network. If you’re actually in the trenches of telecom operations today, you already know the old model is broken. Not long ago, I was working with a Tier One vendor on a RAN software upgrade plan. Ten years back, a nationwide upgrade was a massive, high-stakes event you’d do once a year if you were feeling brave, and everyone was nervous. Today, you can push software to 20,000 radio sites in a matter of days. We’re talking staged rollouts, automated health checks, and rollback logic if something breaks, all running like a standard cloud pipeline. It’s an IT stack now. The network and all the surrounding technology are evolving every week, but the industry is still trying to talk in decade-long cycles. That gap is getting impossible to justify.

Telcos still frame Starlink as a niche solution for rural broadband, aviation, or maritime. That idea breaks the moment you look at the market potential.
In fact, the demand has always been there, but the challenge has been the Telco's cost structure. Extending fiber into low-density areas costs $3,000 to $10,000 per home passed, and often exceeds $20,000 in remote terrain, while monthly home broadband revenue in those same areas is $40 to $80 at best. The return profile does not close, so the investment never scaled beyond minimum coverage.

Depuis quelques années, le spectre des fraudes téléphoniques alimentées par l’IA s’étend bien au-delà des frontières américaines. D’après le rapport « State of the Call 2026 » publié par Hiya, une entreprise internationale spécialisée dans la protection des communications, les appels indésirables n’épargnent désormais aucun marché majeur. En moyenne, un consommateur reçoit 7,4 appels non sollicités par semaine – un chiffre qui grimpe jusqu’à près de 10 appels hebdomadaires aux États-Unis et culmine en France, où les volumes restent les plus élevés d’Europe. Plus inquiétant encore, ce flot d’appels connaît une croissance annuelle de 16 %, tendance observée sur tous les territoires interrogés.

Contrairement à la majorité des solutions du marché qui privilégient le cloud, ce « Personal Computer » mise tout sur une expérience entièrement locale. L’agent IA préinstallé fonctionne en continu sur la machine : il analyse, automatise et assiste l’utilisateur dans sa gestion quotidienne des documents, courriels ou applications, tout en évitant les interruptions entre les différentes sessions. L’accès direct aux fichiers sans passer par un serveur centralisé offre un avantage décisif : la latence serait cinq fois inférieure aux alternatives cloud, selon les annonces de Perplexity. Un argument de taille pour ceux qui exigent réactivité et respect de leur vie privée.

I’m not exaggerating: “Robotic AI Radio” is the exact phrase Jensen used during GTC 2026. What we are witnessing is a fundamental rewiring of how the world computes, and this groundbreaking development was showcased alongside 120 different robots at GTC 2026 today. This deployment is the first fruit of the partnership between Nokia and NVIDIA for building their Aerial RAN product. It is a massive architectural shift that incorporates GPUs directly into the RAN (Radio Access Network) for two critical purposes: improving RF signal processing and, crucially, enabling physical AI inference right at the edge. Although we are really early in the game, it is abundantly clear that AI has left the central datacenters and training labs. It is now all about the physical world and inference, where telcos can play a big role.

MWC 2026 is over, and the hype hangover begins. Back to reality, there is one metric the telecom industry should look at right now: how much money we are actually making from AI. Not how much AI we deploy internally, not how many partnerships we announce, but real revenue generated by operators. This has been the industry’s problem for decades as Telcos build the infrastructure, enable the ecosystem, carry the traffic, and then watch someone else capture the value.
In 2025, the global AI economy generated roughly $1.5 trillion. At best, telecom operators captured about $300 million in direct AI revenue. Even if the definition is stretched to include connectivity and other indirect services, the most optimistic estimates reach around $4 billion. The sense of déjà vu is hard to ignore.
Telcos have already invested well over $300 million in AI and are deploying it across networks and operations, yet we capture only about 0.02% of the AI economy. The opportunity is massive. But if telecom does not fix how value is captured, AI will repeat the same story we already lived through with cloud, apps, and streaming.

A few days ago, John Saw introduced a term that reframes how we think about AI in the physical world. He called it the kinetic token. His argument is simple but structural: If large language models are built on informational tokens that represent fragments of text, then machines that move, lift, weld, or drive must be built on atomic units of time-indexed physical state and not just symbols. Measurements are tied to position, velocity, force, energy, and identity.
That shift marks the boundary between Language AI and Physical AI. In one world, tokens optimize probability distributions over words. In the other, tokens sit inside control loops where milliseconds translate into motion, and error margins translate into safety risk. When intelligence leaves the cloud and enters factories, roads, ports, and robots, the unit of computation changes. And with it, the infrastructure and monetization logic must change as well.

In the last three months, OpenAI has moved from speculation to reality. Reporting based on sworn court filings confirms a working prototype and indicates that the first hardware product will not ship before late February 2027, alongside a branding pivot away from the “io” name. Executive statements at Davos anchor a reveal window in the second half of 2026.
More importantly, first-party hiring signals describe an end-to-end consumer device stack: custom silicon, embedded systems, operating system kernel and frameworks, camera firmware, secure boot, A/B OTA infrastructure, and hardened software supply chain controls. The pattern matches with OpenAI building a vertically integrated hardware platform designed to sit beside the smartphone and control the ambient AI interface layer around daily digital life.
OpenAI is leaving chat interfaces behind and moving into the physical world.