Susana SchwartzThe Economics of AI Infrastructure
ICT expert and WiseHarbor founder Keith Mallinson, whose “Nerds and Herds” predictions set realistic timelines for monetization around 5G AI, today highlights the risks in AI infrastructure investment in RCRTech’s “AI may succeed spectacularly while infrastructure investments fail.”
Based on his recent research paper, The Economics of AI Infrastructure, Mallinson questions whether or not investors can recover the capital they are currently committing to AI infrastructure. With massive projected spending of $6.7 trillion by 2030, he draws parallels to the fiber market, warning that high usage demand does not guarantee profitable returns. As a result, he urges caution in forecasting demand, supply, and price, even putting forth a summary of key questions to ask around AI infrastructure.
To see those questions, read the story, “AI may succeed spectacularly while infrastructure investments fail,” and check in each week for the subsequent articles in this four-part RCRTech series.
Susana Schwartz
Technology Editor
RCRTech
AI Infrastructure Top Stories
Samsung foundry price hike: Amid tightening fab capacity, wafer prices for the 4nm SF4 and 5nm SF5 processes are up roughly 10%–15% for new contracts compared with June, and older 8nm technology, rose by almost10%.
Singappore allocates 200 MW to 4 operators: Singapore will give 50 MW each to 4 data center operators, including Digital Realty, Equinix, Keppel Data Centres and ST Telemedia Global. The government’s will reassess in 18 to 24 months.
AI Today: What You Need to Know
Liquid cooling penetration: By 2027, liquid cooling will approach 60% penetration, according to TrendForce, which notes Google is the most aggressive, using it in more than 80% of its AI servers kand extending it to rack-scale system designs.
Cisco – SuperMicro rack-scale AI computing: For neocloud and sovereign cloud customers, Cisco’s high-density server systems extends the use cases for Secure AI Factory architecture, from trillion parameter training models to edge inferencing.
Rebellions – Collosum partnership: South Korea’s Rebellions will co-design a reference architecture with UK-based AI startup Callosum to supply over 100 neural processing unit computing racks for a UK government-backed AI research cluster.
Bull and Bear case for CoreWeave, Super Micro: Shares of CoreWeave and Super Micro Computer recently jumped to their highest levels thanks to booming demand for AI computing capacity, but is it sustainable?
OpenAI executive shift: OpenAI reorganized its infrastructure division following the departure of its head of data centers, Chris Malone, whose exit comes as data center pushback becomes a “sleeper issue” for the midterm election.
RCR Events
Industry Resources
Report: NTN in motion — evolving standards, expanding services
