Neoclouds: improving time-to-token for AI companies
In AI, a “token” is a basic unit of data (like a piece of a word) that an AI model processes or generates. AI infrastructure generates revenue only when it is actively generating tokens for users. According to RCRTech publisher Jeff Mucci, who recently attended Yotta in Las Vegas, “time to token” frameworks are increasingly crucial to measuring the the total elapsed time from the purchase of AI hardware to the moment in which the hardware is fully installed, configured, and actively generating tokens (AI data) and revenue.
In his story, “AI Infrastructure Economics: Neoclouds race to cut time to token,” Mucci examines how neoclouds shorten time-to-token by getting computing power up and running faster than would be possible if AI labs had to do it themselves. By buying the chips, renting the data centers, setting up the networks, and handling the risks, neoclouds are helping AI labs rent ready-to-go capacity, also absolving AI labs of the logistical, financial, and technical delays that normally would go into building AI infrastructure.
Check out this story, and other “top stories” below.
Susana Schwartz
Technology Editor
RCRTech
AI Infrastructure Top Stories
Analyst Angle: AT&T’s John Stankey says small terrestrial radios will cost as much as a mobile networks, but Elon Musk says the bandwidth and coverage matter more. How could direct-to-cell systems change thinking around edge compute?
Schneider Electric’s PTC acquisition: The deal will expand Schneider’s industrial software, AI and digital capabilities, as PTC specializes in CAD, product lifecycle management, application lifecycle management and service lifecycle management.
AI Today: What You Need to Know
Data centers on federal land: As the Trump administration considers permits for several data center developments on lands managed by the Bureau of Land Management, the Wilderness Society is calling for a moratorium on federal lands.
Spain’s data center boom: Fast-track permits have turned the sparsely populated region of Aragon into a $70 billion data center magnet, with QTS, AWS, and Microsoft taking advantage of the abundant wind and solar energy.
10 defining data center deals: Global Data Center Hub counts down the highest-ranked transactions from recent months, such as the fact AI labs are becoming key “anchor tenants,” with Anthropic and OpenAI behind some of the biggest deals.
AI rewrites power demand: McKinsey data shows data centers are the fastest-growing load segment in OECD power markets, with a Continued Momentum baseline projecting 24% CAGR in data-center electricity demand through 2030.
DayOne files for IPO: Despite recent net losses on revenue, Singapore-based DayOne Data Centers filed for a U.S. IPO this week, with surging bond yields and elevated interest rates prompting postponement of other major listings.
RCR Events
Industry Resources
Report: NTN in motion — evolving standards, expanding services