Ethereum News Today:

The concept of zero-marginal-cost technology—long considered a hallmark of the digital economy—faces growing scrutiny as artificial intelligence (AI) and blockchain industries reveal hidden operational costs. The traditional view that digital products scale with near-zero additional cost is increasingly challenged by the reality of infrastructure, compute, and transactional expenses in these sectors.

In AI, for example, the marginal cost of serving additional users is no longer negligible. While the initial development of models like large language models (LLMs) can be seen as a one-time investment, the ongoing cost of inference tokens, compute power, and data center usage continues to rise. The shift to third-party model providers—such as OpenAI, Anthropic, and Google—has further complicated cost structures. Developers and application companies now face usage-based pricing models that erode gross margins, particularly when demand spikes or models require more tokens per task. Anthropic, for instance, has priced its Claude Opus model at $75 per million output tokens, significantly higher than the $10 per million charged by OpenAI’s GPT-5 model, which is increasingly adopted by AI wrapper companies like Lovable and Cursor.

The blockchain sector, long celebrated for its scalability and low operational cost per transaction, also shows signs of economic inversion. While blockchains like Ethereum and Solana technically offer near-zero marginal costs for validators, users pay for priority access during peak demand periods. This is evident in mechanisms like Ethereum’s EIP-1559, which formalized congestion pricing with base fees and optional tips, and Solana’s “priority fees,” which reflect the scarcity of blockspace. MEV (Maximal Extractable Value) has emerged as a significant revenue stream for block builders and searchers, with Ethereum generating $32.5 million and Solana $68 million in MEV during August 2025 alone. This suggests that, while the nominal cost of transactions remains low, the value extracted from transaction ordering and timing introduces a new dimension of pricing power.

The implications for business models in both sectors are profound. In AI, companies are increasingly behaving like utility services, where margins are compressed by high fixed costs and variable pricing. AI startups and developers often rely on venture capital subsidies and negotiated discounts to remain competitive, but these strategies come at the expense of long-term financial sustainability. Similarly, in blockchain, the economics of MEV and priority fees challenge the traditional SaaS model, where high gross margins are assumed. Investors like Kyle Samani of Multicoin have argued that the value of blockchains should be measured not only by their utility but by their ability to capture MEV as a form of infrastructure rent.

Capital flows in both industries reflect the urgency of these challenges. In 2025, OpenAI raised $10 billion in June and returned to the market just weeks later with an additional $8.3 billion, signaling intense competition and rapid capital burn. Anthropic, despite generating over $400 million in monthly sales by mid-year, is projected to lose billions for the year. Elon Musk’s xAI is reportedly burning more than $1 billion per month. These figures highlight a sector where scale amplifies losses rather than reduces them, and where venture capital is increasingly used to subsidize infrastructure and user acquisition rather than drive profitability.

The absence of liquidity events compounds the problem. Generative AI has produced few meaningful exits in 2025, with most deals structured to repay investors or capture talent rather than to sustain long-term value creation. Even AMD’s $665 million acquisition of Silo AI in 2024 stands out as a rare integration of a viable business. For venture capital, the lack of exits and public market readiness leaves investors with paper gains and no clear path to realization.

At the macroeconomic level, the AI and blockchain sectors have become significant contributors to U.S. GDP growth. AI-related capital expenditures accounted for 1.2% of U.S. GDP in the first half of 2025, with some estimates attributing half of total GDP growth to data center construction alone. However, the sustainability of this growth remains questionable. These projects are financed with debt, depreciate quickly, and depend on demand that is not guaranteed. A pullback in spending by hyperscalers like Microsoft and Amazon could ripple across the economy, affecting everything from semiconductor sales to employment in related industries.

As these trends converge, the zero-marginal-cost model of the digital economy is proving to be more nuanced than previously assumed. For AI and blockchain, the economics are increasingly shaped by hidden costs, infrastructure dependencies, and competitive pressures that challenge the assumption of scalable, high-margin growth. Investors and developers must now navigate a landscape where capital efficiency and profitability are not guaranteed—even in the most advanced technologies of the 21st century.

Source: [1] Is zero-marginal-cost tech no more? (https://blockworks.co/news/zero-marginal-cost-tech) [2] Why Generative AI Burns Billions Without Returns (https://www.mavensolutions.tech/blog/when-growth-consumes-itself-why-generative-ai-burns-billions-without-returns) [3] What Will Remain After the AI and Crypto Bubbles? (https://www.project-syndicate.org/onpoint/ai-crypto-bubbles-american-innovation-ecosystem-breaking-down-by-william-h-janeway-2025-08)