In 2020, the US government made a bet. To approve the merger of T-Mobile and Sprint, regulators needed to pretend the country would still have four wireless carriers, so they picked a successor. Dish Network, a satellite TV company run by Charlie Ergen, a former professional poker player who had spent two decades quietly hoarding spectrum licenses, would inherit Boost Mobile and its 9 million subscribers, buy divested spectrum, and build a brand new nationwide 5G network to challenge Verizon, AT&T, and T-Mobile.Ergen built it. Between 2020 and 2025, Dish Wireless installed more than 144,000 radios on roughly 24,000 towers, covering over 80% of the US population. Court filings put the total investment at about $46 billion, more than $30 billion on spectrum and more than $16 billion on construction. And it was not a copy of anyone else’s network. Dish built the world’s first nationwide cloud native Open RAN network, with radios from Samsung and Fujitsu, software from Mavenir, and a core running in Amazon’s cloud. Engineers around the world studied it as the future of network construction.But customers never showed up. Boost shrank rather than grew, sliding from 9 million subscribers to 7.53 million by early 2026, while quarterly net additions collapsed to a rounding error. A network sized for tens of millions of users carried a small fraction of that, and the bills kept arriving anyway. Tower rent alone ran to $567.8 million in 2025. By last year, Dish had quietly given up, moving Boost’s traffic onto AT&T’s network under a wholesale deal and, in August 2025, beginning the abandonment of the network it had just finished building.

Plans follow the familiar tiered structure: Basic (10M tokens/month), Standard (50M tokens), Business Unlimited. The operator controls pricing by setting ...

The Tokenomics Foundation has been established to unite enterprises, hyperscalers, and frontier model developers in addressing escalating artificial intelligence costs. As organizations struggle to balance AI expenditure with usage, major technology providers are introducing cost management solutions. Oracle is rolling out limited token bundles to enhance spending predictability, while AWS launched its FinOps Agent to identify cost anomalies. The foundation, operating in partnership with the Linux Foundation, aims to establish clear financial controls for AI consumption. This initiative reflects a broader industry shift, with FinOps teams expanding their traditional cloud cost management responsibilities to encompass AI spending oversight and business value alignment. Organizations increasingly recognize the urgent need for comprehensive financial visibility and governance frameworks in their AI investments.

FinOps X 2026 opened with a chart that did most of the arguing for the week: token consumption climbing from single-digit trillions to tens of trillions in ...

Mavenir's CEO highlights significant cost implications of AI adoption, revealing that transitioning to a usage-based model for large language models tripled software development expenses. Despite initial hesitations, Kohli advocates that telecommunications companies invest in owning GPU infrastructure and deploying open-source models, citing seven to eight-year hardware lifecycles and modest performance gaps with frontier models. Financial analysis suggests potential savings of twelve million dollars annually through strategic infrastructure investment. Additionally, Kohli raises critical data security concerns, asserting that public language models may be harvesting proprietary developer information for training purposes, underscoring the importance of maintaining in-house AI capabilities for protecting sensitive corporate knowledge and intellectual property.

Yup, you can be a trillion-dollar company; you can have the cheapest plans on the market; you can hire the best engineers in the world. But if you don’t have enough spectrum, everything is an uphill battle. Spectrum is the land your network is built on. Everything else, such as radios, software, and marketing, is what you build on top of it, and no amount of construction skill compensates for a small lot.The United States is about to prove this twice. Once with fresh performance data that shows exactly how spectrum, not equipment, decides who wins today. And once with what’s coming with the mother of all spectrum auctions, 160 MHz of prime mid-band, the rules for which the FCC votes on July 22, 2026, which could redraw the American wireless map for the next decade.

On July 1, 2026, the Wall Street Journal dropped a story that moved $130 billion of market value in an afternoon: SpaceX had quietly shown investors a prototype handset during its IPO roadshow. Slimmer than an iPhone, with a proprietary operating system, a Qualcomm Snapdragon chip and xAI’s Grok woven into its core. Elon Musk’s response took four words, “utterly false”, and the market’s response took about 7% of SPCX shares, erasing more than $50 billion of Musk’s net worth and, for a day, his title as the world’s first trillionaire.

A Token Policy architecture functions as a hybrid system, moving beyond the traditional network layers. It requires a convergence of functions that bridge the transport layer, where the 5G core operates, with the application layer, where token-based requests are actually generated.Because the network must now “read” the intent of the traffic, we are essentially building a cross-layer gateway that sits between the user equipment and the service destination. The long-term viability of this approach depends on whether the 3GPP decides to standardize these functions.

The question is not if, but how. Mapping the strategic paths for SpaceX’s entry into the terrestrial mobile space, and why, while difficult, it is far from impossible for Musk.

Sam Altman has proposed reimagining artificial intelligence as a metered utility similar to electricity, where users pay monthly bills based on token consumption rather than subscription tiers. Under this model, light users would pay minimal amounts while heavy users incur substantial costs, with the ultimate goal of achieving AI abundance where intelligence becomes "too cheap to meter." However, critics raise concerns about potential gatekeeping and forced minimum charges, while real-world evidence—such as Uber exhausting its 2026 AI budget in just four months—suggests the industry remains far from achieving affordable abundance. Drawing parallels to failed 1950s predictions about nuclear power becoming too cheap to meter, questions persist about whether Altman's vision represents inevitable pricing evolution or merely consolidates intelligence access behind expensive paywalls, particularly without sufficient infrastructure investment.

Agentic AI systems are generating unexpected cost explosions for major enterprises, with Uber exhausting its entire annual AI budget by April 2026 due to token consumption rates five to thirty times higher than standard chatbot queries. As organizations scale agentic workflows in production, per-engineer monthly costs reach $500 to $2,000, fundamentally differing from previous AI pricing models. The economics extend beyond simple model pricing to encompass planning, context retrieval, tool integration, state management, and error handling. Industry leaders acknowledge that cost sustainability has emerged as a critical concern, with customers depleting 2026 budgets prematurely. Understanding the true total cost of task completion and implementing effective cost management strategies are essential before significant financial impacts materialize.

Tokens have emerged as the fundamental economic and technical unit for foundation model services, connecting information processing, computation, and pricing. This paper presents a comprehensive framework for AI tokenomics, examining how tokens are generated, consumed, priced, and allocated across AI systems. The research links token-level technical costs to enterprise workflows and resource allocation while demonstrating that token expenditure and economic value are distinct—value depends on marginal productivity, workflow position, hidden reasoning, risk, and downstream effects. The paper identifies critical research directions including hidden-token measurement, empirical calibration, token productivity optimization, dynamic allocation mechanisms, and emerging token-based markets, offering essential insights for understanding the economics of modern artificial intelligence.

The Token Delivery Network represents the next evolution in AI infrastructure, shifting focus from GPU capacity acquisition to distributed inference delivery. Similar to how Content Delivery Networks revolutionized internet content distribution, TDNs bring AI model endpoints closer to users, applications, and systems based on proximity, performance, capacity, sovereignty, and cost considerations. Rather than serving cached static content, TDNs deliver real-time generated tokens from optimized edge locations, enabling low-latency AI access at scale. This distributed operating model provides simplified governance and improved performance while addressing the latency-sensitive nature of modern AI applications, establishing a new paradigm for global AI service delivery.

T-Mobile's chief technology officer has proposed "kinetic tokens" as a novel mechanism for telecommunications service providers to monetize their role in supporting physical AI and 6G networks. Unlike traditional compute tokens that measure processing resources, kinetic tokens would account for network connectivity and edge infrastructure needs essential for AI-powered robots and devices. However, industry skepticism prevails, as many executives question whether this concept represents a meaningful departure from conventional gigabyte-based network pricing. While telcos possess substantial edge infrastructure advantages through their extensive radio access networks and distributed data centers, transforming this infrastructure into a distinct currency mechanism remains contentious. The proposal highlights broader industry debate regarding whether telecommunications providers can evolve beyond connectivity provision to capture greater value from emerging artificial intelligence deployment models.

The question of whether Elon Musk can execute a Jio-style entry into the US wireless market is easily answered: yes, he can, and it is highly likely to happen. There is a specific reason Musk has been selling a massive vision to his financial backers; during the recent IPO roadshow, SpaceX suggested to investors that Starlink’s addressable market is the entire global communications sector, representing an estimated $1.6 trillion in total value.

When a consumer installs a satellite dish on a roof, they establish a pristine, uncompromised line of sight to the sky using an active phased-array antenna with massive gain (30+ dBi). That architecture scales. Direct-to-Device does not. When you remove the dedicated roof dish and attempt to close that same 500 km link with an unmodified smartphone, you are forced to rely on an isotropic, omnidirectional antenna yielding roughly 0 dBi of gain and operating at a fraction of a watt of transmit power.Despite the massive satellite apertures currently being deployed to brute-force the uplink, the laws of physics remain unforgiving. Orbital D2D cannot provide meaningful concurrent sector capacity, cannot provide low latency, and, due to a microscopic penetration margin, absolutely cannot provide indoor coverage.Orbital D2D is more like an insurance policy. It is a brilliant, necessary solution for the 0.0001% of edge cases like the stranded hiker, the mid-ocean SOS, and extreme remote telemetry. But for Telcos tasked with delivering gigabits of data to dense urban and suburban populations, where 80% of data is consumed indoors, orbital D2D is practically useless as a core capacity layer.

Telecom autonomy has struggled to evolve under traditional machine learning models because these older systems operate as uninterpretable black boxes. Machine learning excels at deterministic tasks and pattern recognition across massive datasets. However, network engineers managing critical infrastructure require clear explanations before executing system changes.This lack of explainability stalled the progression of autonomous networks. Handing over control of complex actions without understanding the underlying math creates an extremely risky trust gap for Telcos. To resolve this, Nokia and Google Cloud designed the Gemini-powered agents using a “glass box” architecture.Instead of executing an unverified command, the action reasoner agent functions as an advisory layer. It processes the network data and presents a confidence-based recommendation to the human operator. Because these specialized agents can explain their conclusions and reasoning, they elicit greater trust from users. Human engineers retain control and final approval over critical control points before logging and executing fixes. By combining autonomous data analysis with human observability, the platform establishes the required safety to deploy machine-speed operations.

The Linux Foundation announced the creation of the Tokenomics Foundation to establish open standards and best practices for managing AI costs, addressing enterprises' growing struggle with unpredictable artificial intelligence spending. Unlike traditional cloud expenses, AI token costs—representing units of text processed by models—exhibit volatile consumption patterns that finance teams find difficult to forecast and control. Major industry players including Google, Microsoft, IBM, and JPMorgan Chase have pledged support for the initiative. Recent data reveals alarming spending trends, with some companies experiencing token cost increases of up to fifty percent quarterly. The foundation, launching formally in June at FinOps X, aims to bring transparency and accountability to the AI economy through established benchmarks and best practices.

As artificial intelligence becomes increasingly embedded in organizational operations, token usage has emerged as a critical cost consideration for executives and boards. While AI technology's potential is widely acknowledged, scaling deployment effectively remains challenging, with organizations struggling to capture full value. Tokenomics—the discipline of understanding and managing AI token consumption—has become essential as companies transition from experimentation to production-scale implementation. Though per-token costs have declined dramatically since 2022, exponential increases in usage offset these savings, particularly as organizations explore transformative agent-led applications. Managing token efficiency effectively is now crucial for controlling AI expenditures and maximizing return on investment in scaled deployments.

Billing is fully integrated: tokens are metered and charged through the operator's existing BSS using the same mediation infrastructure as today's data plans.