Forrester’s assessment is that an AI infrastructure crash is not imminent, but businesses should prepare for a market correction. Around 90 percent of current investment comes from large technology companies with healthy cash flows, limiting the immediate danger. The bigger risks lie with financially strained builders and deals in which AI companies and infrastructure providers depend on one another’s funding. Meanwhile, spending is far ahead of demonstrated returns: the build-out would break even only if AI added 5 percent to US GDP by 2035. Expected productivity gains account for about two percentage points, leaving roughly $1.3 trillion in value to come from other uses. Another estimate puts the industry’s required annual revenue at $6 trillion by 2031.
The eventual correction will depend less on oversupply alone than on whether demand and lasting business value materialize. If they do not, providers may cut investment, merge, raise or lower prices, or withdraw products, leaving customers to replace services and redesign plans. Businesses should measure returns from individual AI projects and monitor suppliers’ finances, including the gap between contracted business and recognized revenue. Dependence on providers relying heavily on outside investment or customer prepayments warrants particular caution. Firms should also rehearse scenarios such as a vendor ending support or a model supplier failing to pay its computing provider, and identify in advance which uses to abandon, what data and workloads to move, and the cost. The advice is to keep investing where value is measurable while limiting exposure and preparing exit plans.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers tested 16 LLMs on all 55 TIS-100 puzzles, giving each one attempt and time to refine its code. The models found 133 new performance tradeoffs across 18 puzzles and set three outright cycle records, though human players improved on most solutions within days. (tis-100.wincon.dev)
Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent harm caused by their models. The government is also weighing an AI safety watchdog with power to impose significant fines, following a hack of a Medicare statistics portal involving OpenAI. (abc.net.au)
AI is increasingly becoming a gateway through which people find information. Audiences still resist having AI produce their news. (niemanlab.org)
Nvidia holds a central position in the AI industry, making disruption to the company potentially catastrophic for the wider ecosystem. The race for advanced AI is also intensifying economic and geopolitical conflict, while concentrated corporate power and infrastructure face growing political resistance. (france24.com)
LLM inference can require reading about 26 GB of model weights to generate each token from a 13-billion-parameter model. The explanation shows how model size drives memory bandwidth demands during text generation. (youtube.com)
Deletion Tuning matches full fine-tuning across three LLM post-training methods by zeroing fewer than 0.05% of model weights. Its compact masks also preserve prior-task accuracy better than full fine-tuning, LoRA, and random sparse tuning. (arxiv.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.