TLDR
- Armstrong says near-infinite AI demand will meet cheaper models, shifting focus toward compute economics rapidly.
- Cheaper specialized models could handle routine enterprise tasks while frontier systems serve harder reasoning workloads.
- Energy supply, cooling, grid access, and chips may shape the next AI infrastructure race worldwide.
- Crypto compute networks are positioning GPUs, miners, and verification tools for cheaper inference demand growth.
- Lower model prices may expand usage, but data centers and power remain harder constraints globally.
Coinbase chief executive Brian Armstrong has said demand for artificial intelligence is “near infinite,” while predicting that 80% of AI workloads could move to models that are 99% cheaper within 12 to 18 months, according to comments circulating among technology and crypto market observers.
He added that energy and compute, rather than user interest, are likely to become the main limits for AI expansion, a view that has drawn attention from investors watching the overlap between cloud infrastructure, digital assets, data centers, and decentralized computing networks.
The statement reflects a broader debate over the economic structure of AI, as companies test whether high-value tasks require frontier reasoning systems or whether many routine workloads can be handled by smaller, specialized, open-source, or purpose-built models running at lower cost.
Cheaper AI Models Move Into Routine Workloads
Armstrong’s forecast places cost compression at the center of the AI market, with cheaper models expected to handle tasks such as classification, drafting, retrieval, customer support, internal workflow automation, and other functions where maximum reasoning performance is not always required.
The shift would not necessarily reduce overall AI usage, because lower prices can increase deployment across companies that previously limited experimentation due to model costs, compliance concerns, or uncertainty about returns from early AI spending.
In sectors such as health care, finance, logistics, commerce, and software development, specialized models may also give organizations more control over latency, privacy, customization, and operating expenses than constant reliance on general-purpose systems.
Energy and Compute Become Central Constraints
The energy argument has become more prominent as AI demand expands beyond research labs and into consumer products, enterprise software, robotics, medical systems, coding tools, search, advertising, and back-office operations that may run continuously throughout the day.
Data centers need electricity, grid connections, cooling capacity, land, transformers, chips, networking equipment, and construction timelines that cannot be expanded as quickly as software demand, which means physical infrastructure may determine how widely cheap AI can be deployed.
That view has led some crypto market participants to compare AI’s growth path with earlier periods in computing and internet bandwidth, when falling costs accelerated adoption while network capacity, hardware availability, and power access shaped competition.
Crypto Infrastructure Enters the AI Debate
Armstrong’s remarks have also renewed attention on decentralized compute, encrypted data access, and on-chain verification systems, including projects that propose allowing AI systems to use selected private data slices while recording proofs of work or usage through blockchain networks.
Supporters of decentralized physical infrastructure networks argue that GPU marketplaces, energy coordination protocols, and mining operators could supply compute during the transition toward cheaper inference, though these systems still face questions over reliability, regulation, data handling, and enterprise adoption.
Bitcoin miners are part of that discussion because they already operate energy-intensive facilities and may have access to power contracts, cooling systems, and industrial sites that could be adapted for AI-related workloads where business conditions allow.
According to analysts the main commercial question is no longer whether AI demand exists, but which providers can deliver secure, affordable and verifiable compute at scale as businesses divide routine tasks from workloads requiring frontier models. Armstrong’s prediction points to a market where model costs fall sharply while energy, chips, data centers, privacy controls and deployment networks become core competitive factors for future growth.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
