Tag
Epoch AI analyzes whether financing will bottleneck AI compute scaling, using Anthropic's $50B infrastructure buildout funded by debt as a case study. It argues institutional investors are willing to lend against long-term payment commitments, especially with supplier backing, making financing unlikely to be the immediate limit on frontier compute growth.
Kalshi CEO Tarek Mansour predicts compute will become a $10 trillion industry by 2030 and is racing to build a futures market for hedging AI compute costs, competing with CME Group and Intercontinental.
B3 Labs has released B3IQ, allowing universities, enterprises, and AI professional users to purchase and host NVIDIA GPU servers in installments with a 30% down payment. When idle, the computing power can be rented out for income, which can be used to offset the purchase price or as direct earnings. It has already been used by Stanford, NYU, and many other universities.
Elon Musk estimates that Terafab's AI compute output will be split roughly 25% for Tesla Optimus and 75% for AI spacecraft, consistent with SpaceX's long-term target of 1TW of compute hardware annually.
Google's aggressive monetization of TPU capacity to outside customers like Anthropic is fueling internal frustration and driving key AI researchers to depart for competitors, intensifying talent and competitive tensions.
SpaceX's first public earnings reveal it is primarily a telecom and AI-compute company, with Starlink and data-center leasing far outpacing its rocket business. The analysis highlights the company's pivot toward renting compute capacity and its heavy AI spending.
British AI neocloud Nscale acquires software startup Anyscale for $1.65 billion to strengthen its AI compute stack by adding workload management and scaling capabilities.
The article analyzes factors that could drive AI compute costs up 10x in coming years, including rising lab revenue, margins, and spot prices, while suggesting that inference spending may signal stalled progress.
A deep-dive list of ten major breakthroughs expected from scaling AI compute to ~150 million H100-equivalents by 2028, including advances in mathematics, drug discovery, materials science, biology, fusion, and climate modeling.
At AMD's Advancing AI event, tinygrad founder George Hotz argued for commoditizing petaflops and increasing competition to lower AI compute costs, opposing monopolies.
A new report from BloombergNEF predicts data centers will quadruple their electricity use by 2035, consuming one-fifth of U.S. electricity, driven by surging AI compute demands and straining already burdened power grids.
A 64GB CMP 170HX NVIDIA GPU with A100 tensor cores has appeared on China's secondhand market for $1,170, and a software-level vulnerability has successfully removed its computing power limit.
The article analyzes recent moves by SpaceX and Meta to sell excess AI compute capacity, questioning whether this signals an end to compute scarcity. It argues the deals are short-term and high-priced, and that underlying demand remains strong, refuting the bear thesis.
An analysis of the cost-effectiveness of building a $20,000 local AI rig, calculating the breakeven point compared to cloud AI services.
NVIDIA announces a new business model to provide AI compute at scale through partnerships with AI clouds, enabling faster access to infrastructure via revenue-sharing and credit support.
Meta is developing plans to sell excess AI compute and models via a new cloud business called Meta Compute, similar to SpaceX/xAI, aiming to monetize its massive AI infrastructure investments.
A discussion of the idea that placing AI compute in space could solve energy problems, leveraging solar power without atmospheric interference, and citing SpaceX's potential to launch massive compute capacity.
The x86 Ecosystem Advisory Group has published the AI Compute Extensions (ACE) specification, defining new x86 instructions and register state for accelerating matrix multiplication and reduced precision data formats in machine learning workloads.
This article examines the history of CUDA alternatives like OpenCL and SYCL, explaining why they failed to become dominant in AI compute due to slow committee-driven development and the challenges of open coopetition.
Google will pay SpaceX $920 million per month from October 2026 through June 2029 for AI compute capacity using approximately 110,000 NVIDIA GPUs, as a bridge to meet demand for Gemini Enterprise while Google expands its own AI infrastructure.