Skip to content
Christmas & Holiday Lights 2026Explore the community map

Data Center Power, Memory & Voice AI

US Data Centers Use 82 TWh as Altman Notes RAM Costs and Tone Debuts

A Harvard study in Nature Sustainability estimated that 403 U.S. hyperscale data centers consumed 82 terawatt-hours of electricity, OpenAI CEO Sam Altman addressed 5x to 7x DRAM price jumps at Semafor's summit, and Jordan Belfort launched AI voice startup Tone.

By Danville Times · Syndicated from Concord Chronicle
Syndicated reporting from Concord Chronicle · Concord Chronicle
Published · 3 min read

Restrained Economist-style conceptual print illustration on warm cream cotton paper showing an electrical substation transformer and stacked DRAM memory modules balanced on an analytical scale
Hyperscale electrical grid demand, semiconductor memory supply chains, and low-latency voice synthesis models reshape enterprise computing economics. (Illustration: Concord Chronicle) · Concord Chronicle Editorial Illustration
Your listening Queue & catch up
Unheard stories first, timed at 1×. Your selections stay first.

Add a story or build a catch-up queue to start listening.

Energy consumption across hyperscale computing facilities, global semiconductor memory shortages, and commercial voice synthesis startups drew industry scrutiny on Monday as a Harvard study in Nature Sustainability quantified the electricity and carbon footprint of 403 U.S. hyperscale data centers, OpenAI chief executive Sam Altman addressed surging DRAM prices at a Semafor summit, and co-founders Jordan Belfort and Shervin Pishevar launched AI voice startup Tone.

Study Finds 403 U.S. Hyperscale Data Centers Consume 82 Terawatt-Hours Annually

A study published Monday in the journal Nature Sustainability by lead author Gianluca Guidi, a postdoctoral fellow in the Department of Biostatistics at the Harvard T.H. Chan School of Public Health, used satellite imagery and Environmental Protection Agency data between May 2024 and April 2025 to identify 403 hyperscale data centers in the United States consuming roughly 82 terawatt-hours of electricity annually—about 1.8% of total U.S. electricity consumption, according to The Verge. John Sterman, director of the MIT System Dynamics group and MIT Sloan Sustainability Initiative, noted that 82 terawatt-hours exceeds the annual electricity consumption of about 148 nations including Austria, Israel, and Venezuela, while the resulting 25.7 million metric tons of CO2 emissions—comparable to Denmark's annual footprint—exceed the total carbon emissions of about 113 nations because roughly 54% of the facilities' electricity came from natural gas and coal.

The researchers found that coal-reliant Wyoming produced roughly 7 times more carbon emissions per kilowatt-hour than hydropower-rich Oregon, prompting Cornell University energy systems engineering professor Fengqi You to warn of long-term carbon lock-in and former White House climate adviser and EPA administrator Gina McCarthy to call for transparent, independent data on data center power usage.

Sam Altman Addresses 5x to 7x DRAM Price Jumps at Semafor Summit

Speaking on stage Monday at Semafor's Silicon Valley & The World summit, OpenAI CEO Sam Altman clarified earlier remarks made on the Cheeky Pint podcast and X regarding personal computer RAM purchases, stating that he supports home tinkering and wants consumers to be able to buy ample RAM for their own computers, The Verge reported. Altman noted that OpenAI has faced periodic supply chain constraints across memory, rack assembly, and data center capacity, observing that global memory manufacturing did not expand fast enough and that DRAM prices have risen roughly 5x to 7x, raising questions about whether scarce memory is utilized 100% of the time in shared infrastructure versus edge devices.

Jordan Belfort and Shervin Pishevar Launch AI Voice Startup Tone With $10M

In commercial voice software, Jordan Belfort—who pleaded guilty in 1999 to securities fraud tied to Stratton Oakmont, served 22 months, and completed $110.4 million in restitution obligations across 1,513 people in April after earlier $40,000 (1 bitcoin) Miami Beach workshops—has teamed with co-founder Shervin Pishevar and vice president of engineering Ben Sussman to launch an AI voice platform called Tone, TechCrunch reported. Having raised about $10 million in friends-and-family and seed funding and generated roughly $700,000 (approaching $1 million) in revenue, Tone uses compact models operating at roughly 600-millisecond latency trained on 30 years of Belfort's sales audio across 15 strategic tonalities—a system he began teaching in the late 1980s. Competing in a market alongside ElevenLabs and Cartesia while complying with 2024 Federal Communications Commission Telephone Consumer Protection Act opt-in rules, Tone serves law firms, health insurance companies, polling operations, and automotive dealers, and is introducing a self-service narration and live-voiceprint cloning platform for authors.

Sources

Comments (0)

No comments yet. Be the first to share a thought on this story.