Verified SEC figures show Nvidia's cost of revenue rose about 3.8 times in two fiscal years. A survey says every respondent plans to use AI agents, but company data and trust lag. That survey's source failed our check on all 11 claims we tested, so we flag it and lean on the data that held.

SEC-filed figures show Micron nearly doubled its capital spending in fiscal 2025 and now holds about 2.5 times the cash it had a year earlier. Analyst enthusiasm for power, data-center and quantum stocks rests on sources that held up only about half the time in our checks.

Companies from Manulife to Box customers are rolling out AI agents, and funding is flowing. But a trade-press study reports that in 'data laggard' firms, AI can reach 30% or less of company data. That source has a poor measured record, so we flag it as a claim, not a fact.

Every company surveyed by MIT Technology Review plans to use AI agents within two years, yet firms with weak data give those agents access to 30% or less of it. Only some of this is firmly established: that survey's source scores 0% on our reliability check, and the Nvidia figures are SEC-verified.

Adobe's weak guidance, Alphabet's negative free cash flow, and a wave of safety-driven departures from Anthropic, Google and OpenAI show investors and insiders both demanding proof, not promises — even as California moves first to regulate AI's reach into children's lives.

A single summit produced roughly $950 billion in AI infrastructure deals and locked Nvidia into South Korea's memory supply chain, while one research firm says Wall Street is still underpricing the memory chip boom behind it. Here's what's verified, what's just one analyst's call, and what's still speculative.

C3.ai's CEO sold $4.8 million in stock the same week the industry's biggest AI-agent-trust survey turned out unreliable on 11 checked claims, even as Manulife, Box and a wave of banking and healthcare startups commit real budgets. Via News separates the verified from the unverified.

Google now lets users strip visible watermarks from Gemini-made images, video and music — a move that cuts against the story of an industry rushing toward stronger AI-content labeling, and lands the same week OpenAI lost another senior executive.

Nvidia and South Korea's chipmakers signed roughly $950 billion in new AI agreements in July, including a deal to help secure Nvidia's HBM memory supply from SK Hynix, even as chip stocks logged a sharp correction and the Nasdaq 100 hit a 2.75-month low. Via News traces the deals, earnings and balance-sheet data behind the divergence.

A late-July selloff cut deep into AI-chip stocks — Micron fell nearly 40% from its peak — even as Nvidia expanded its SK hynix memory deal, GlobalFoundries signed a $300M federal photonics award, and Skyworks-Qorvo named a combined leadership team. The data show a maturing boom repricing risk, not a broken one.

Nearly all enterprises surveyed say they'll deploy AI agents within two years, but agents can typically reach under half of company data — and as little as 30% at 'data laggard' firms. Platform giants are selling governance; a new wave of startups is skipping the data problem by targeting narrow, auditable back-office tasks instead.

Brad Lightcap, OpenAI's special projects lead and the company's former chief operating officer, has announced his departure from the company.

Insurer Manulife, software firm Box and data platform Snowflake are each rolling out enterprise AI agent tools this year, spanning insurance, content management and data operations worldwide. A fintech startup, Casap, is building agents for dispute resolution, while enterprise leaders across regions report unclear governance for the technology.

Micron and SanDisk are among 2026's top-performing chip stocks on AI memory demand, but both have fallen sharply from June record highs. Figures dated and sourced.

Jeff Dean and Sanjay Ghemawat quit Google within the same week to launch startup Discovery Loop, while Demis Hassabis stepped back from running DeepMind. Alphabet responded by investing in the new venture rather than treating the exits as a pure loss, a hedging strategy now spreading across global AI markets.

Oriol Vinyals, co-leader of Google's Gemini models, has left to co-found Discovery Loop, a new frontier AI venture. The move puts a veteran of one of the world's best-funded AI labs into a startup that must build compute, capital and talent from scratch amid intense global competition.

Bloom Energy signed a $1.7 billion fuel-cell agreement with Nebius AI, one of the largest on-site power deals for AI data centers to date. The company also struck a deal with utility giant AEP and expanded distribution in South Korea, showing on-site generation is now a global bottleneck for AI buildout.

Bloom Energy signed three fuel-cell partnerships in quick succession, led by a $1.7 billion deal with Nebius AI, plus deals with US utility AEP and South Korea's SK Ecoplant and SK Eternix. Analysts say the geographic spread signals a global scramble for AI data center power beyond traditional grid capacity.

A cluster of AI-native drug discovery platforms launched worldwide in early-to-mid 2026, all running on NVIDIA GPU infrastructure. Novo Nordisk's decision to outsource cell therapy research to an AI-native partner signals a broader industry shift from in-house labs to external, GPU-powered specialists.

BlackRock priced a $12.5 billion bond tied to the Meta-BlackRock Texas Data Center Complex at a steep 7.53% yield after weak demand from investors worldwide. Credit-default swaps on major AI-linked firms jumped in tandem, signaling that credit markets globally are pricing AI infrastructure risk far more cautiously than equity markets.

KLA raised its 2026 wafer fabrication equipment forecast to the low-$150 billion range, citing AI chip demand. The California-based supplier also lifted quarterly revenue guidance to $4 billion and its advanced packaging target to $1.1 billion, alongside a 10-for-1 stock split.

Amkor Technology posted record second-quarter revenue as AI demand lifted chip packaging volumes, though weaker Q3 guidance points to uneven growth. Nvidia, the U.S. government, and chipmakers on three continents are all moving capital and consolidating to keep pace with AI infrastructure demand.

KLA Corporation lifted its advanced-packaging process control revenue forecast to about $1.1 billion, up 70% year-over-year, citing sustained AI infrastructure demand worldwide. The growth concentrates in one niche tied to chip packaging hubs in Taiwan, South Korea and the US, raising structural risk if global AI capex slows.

KLA CFO Bren Higgins says US export restrictions are ceding Chinese semiconductor-fab market share to non-US rivals unbound by the same rules. The disclosure places KLA alongside Applied Materials and Lam Research in flagging China controls as a recurring earnings drag, part of a wider realignment of the global chip-equipment supply chain.

SunPower's gross profit fell 41% to $27.6M last quarter, far outpacing a 23% revenue decline to $56M. New CFO Tom Kowalczuk inherits a margin gap that mirrors pressure across the global solar sector, from China to Europe.