US tech stocks see sharp falls over AI cost fears
bbc.co.uk ·
US technology stocks fell sharply for a second consecutive day as investors questioned whether enormous artificial-intelligence spending will produce sufficient returns. The Nasdaq dropped more than 3.5%, while Intel shares fell as much as 30%, putting the chipmaker on track to lose more than $130bn in market value. Reddit fell 26%, Roblox nearly 20%, and Amazon was down about 1.6% in Friday trading after recovering from a steeper overnight decline.
Alphabet had lost more than $500bn in market value on Thursday after saying its AI-generated search summaries were affecting core search-advertising revenue. The company expects to spend up to $205bn this year, almost entirely on AI-related investments. Amazon has committed $220bn this year for areas including AI data centres, network equipment and servers. Meta, Alphabet, Oracle, Amazon and Microsoft have collectively committed $800.6bn of spending in 2026.
Intel forecast lower-than-expected fourth-quarter revenue, said it does not expect cash inflows to exceed spending this year, and cited a PC slowdown expected to continue through at least the first half of next year. Analysts and investors have raised concerns that AI infrastructure and software costs could outstrip anticipated revenue, although major technology stocks remain above year-earlier levels in many cases.
What people are saying
The discussion largely rejected the headline’s simple story that markets are recoiling from AI infrastructure bills.
What readers liked
- Google’s AI summaries appear materially better for routine questions, letting some users get a useful answer without wading through links and…
- AI demand still looks robust, with model-usage rankings and AI-centered workflows suggesting the technology delivers enough value to become hard to…
Concerns raised
- Google’s AI search is accused of serving outdated or incorrect claims with enough confidence that users must audit its citations, undermining its…
- Answer engines may cannibalize the click-driven web economy: fewer search-ad clicks and fewer referrals threaten the sites whose content feeds AI…
Read full discussion analysis
Source: Hacker News · Based on 13 sampled comments · Analyzed
Discussion overview
The discussion largely rejected the headline’s simple story that markets are recoiling from AI infrastructure bills. Several commenters argued the sell-off instead reflects AI disrupting the businesses expected to finance the buildout: answer engines keep users from clicking search ads and publisher links, while referral-dependent sites lose traffic. Others noted that expensive datacenter expansion is still an AI-cost story because it has tightened supplies of memory and compute hardware, raising PC prices and creating a broader capital-spending hangover.
The most concrete AI debate centered on Google’s search summaries. Critics described a product that confidently delivers stale or false information and appears rushed into search to keep pace with the model race; even cited sources allegedly fail to support many answers. Defenders said the system has improved sharply, now resolves ordinary questions accurately enough to avoid opening reference pages, while retaining caution on niche subjects. That split fed a wider anxiety: AI demand is plainly strong and deeply embedded in some developers’ work, but cheap access may not survive if providers cannot turn massive compute spending into durable revenue.
There was also frustration that consumer GPUs and local-model hardware remain scarce while capacity is concentrated in datacenters. Some see on-device inference as the practical escape hatch, while others stress that fabs, advanced lithography, memory, and skilled labor cannot expand on command. A minority welcomed a bubble correction as overdue discipline for inflated AI claims, despite the economic damage it could cause.
Points of disagreement
The sharpest fight is whether AI search is an unusable, unverified product being forced into a core information service, or a rapidly maturing convenience tool. Skeptics say its error rate and weak grounding make replacing conventional search reckless; supporters say recent model upgrades already answer most everyday queries correctly and save substantial browsing effort.
What readers liked
- Google’s AI summaries appear materially better for routine questions, letting some users get a useful answer without wading through links and reference pages.
- AI demand still looks robust, with model-usage rankings and AI-centered workflows suggesting the technology delivers enough value to become hard to give up.
- Local inference hardware is seen as a promising route to faster, more affordable private LLM use instead of renting every task from a datacenter.
- AI lowers the effort required to build software and games, potentially broadening who can create interactive products even if its near-term market effects are unclear.
Concerns raised
- Google’s AI search is accused of serving outdated or incorrect claims with enough confidence that users must audit its citations, undermining its role as a search replacement.
- Answer engines may cannibalize the click-driven web economy: fewer search-ad clicks and fewer referrals threaten the sites whose content feeds AI answers.
- The AI buildout is consuming enormous capital and scarce memory and compute capacity before companies have demonstrated revenue sufficient to justify the spending.
- Developers who now rely on inexpensive AI assistance fear a severe productivity shock if providers raise prices to cover inference and datacenter costs.
- Datacenter-first GPU allocation leaves consumers unable to buy capable hardware for local models, concentrating AI access and experimentation in large platforms.
Keep reading
Sources and details
bbc.co.uk · Published
US tech stocks see sharp falls over AI cost fears
Read report →
Community analysis available for 1 of 1 linked discussions.
Additional discussion searches are still incomplete.
Coverage details and dates
Latest available community analysis:
Summaries: 1 current · 0 awaiting update · 0 not verified
Latest summary update:
Hacker News · bbc.co.uk