AI
Gemini 4 includes Argon, Carbon, Neon and Hydrogen, with reported benchmark leadership and product availability across the lineup, but much of the claimed performance remains unverified in the listed coverage. Carbon is described as a cyber-focused model with limited access; Argon’s reported frontier and coding capabilities also lack independent corroboration. Neon has reported strong coding-agent rankings, while Hydrogen is in Search. The latest plugin update exposes an additional reasoning setting, but its output comparison is not an independent model evaluation.
stories per week
last 8 weeks22 stories41 reports10 outletslatest 22 Nov
AIHeating up
Google is expanding agentic AI across consumer, developer and enterprise products. Gemini Tasks includes Duplex calling and Mac computer use, while Google has opened Gemini and Search to more third-party apps and reported enterprise agent and agent-isolation offerings. A new Gemini API preview adds Tool Retrieval for large tool libraries, with claimed efficiency gains not independently quantified. The stories do not establish broad adoption, reliable task completion or measurable security benefits.
stories per week
last 8 weeks8 stories11 reports4 outletslatest 21 Nov
AIHeating up
Reports describe a proposed staged financing for Nvidia GPUs to expand SpaceX’s AI compute capacity, with earlier coverage citing a $40 billion Apollo commitment and investor solicitation. The financing, bond issuance, terms and completed GPU purchases remain unconfirmed. The latest report says a lender-requested restriction on using financed chips in space could complicate funding for orbital data centers; the restriction’s terms and outcome are unclear.
stories per week
last 8 weeks6 stories8 reports2 outletslatest 20 Nov
AI
The Megaspeed-related espionage and chip-diversion allegations remain unverified in the supplied reports; they establish no finding that information was transferred or chips diverted. Reports describe U.S. scrutiny, Nvidia’s reported investigations and leadership turmoil, but provide no findings. Coverage also alleges Nvidia connected Alibaba with Alice Huang before U.S. chip curbs and cites links to China’s gaming investment network and Taiwan’s XPEC case, without establishing operational impact. New leadership-focused commentary raises possible supply-chain and compute-planning uncertainties but does not confirm changes to chip cadence or supplier orders.
stories per week
last 8 weeks66 stories157 reports10 outletslatest 20 Nov
AIHeating up
Reports point to a potentially narrowing Nvidia lead in China’s AI accelerator market and Huawei claims that Ascend is gaining ground, but market-share evidence remains limited and unverified. DeepSeek’s reported plans to train V5 on Ascend 950 and chip co-design remain unconfirmed. Recent coverage focuses on Jensen Huang’s reported departure and scandal; reports raise supply-chain and China-exposure questions but do not establish operational or market changes. Allegations remain unconfirmed, and Microsoft has publicly reaffirmed its Nvidia partnership.
stories per week
last 8 weeks15 stories38 reports3 outletslatest 20 Nov
AI
Anthropic’s IPO is reported as successfully completed at a $1.5 trillion valuation, following earlier reports that it priced shares at $83. The latest account would make it a record-sized financing milestone, but the supplied assessment says the completion and valuation still require confirmation. Anthropic’s compute strategy spans four accelerator families, while the scale and execution of reported compute commitments remain unclear.
stories per week
last 8 weeks22 stories48 reports9 outletslatest 20 Nov
AIHeating up
AI-memory capacity expansion remains underway, but the reports do not establish a quantified broad supply increase. SK Hynix is reported to be advancing HBM4 packaging capacity and cleanroom work; Micron has moved equipment into its Singapore HBM plant, targeting production in 2027. New reports add SK Hynix’s proposed memory-centric AI rack and reported plans to buy HBM bonding tools, while Samsung and SK Hynix are reportedly aligning on a proposed chip-park division. These plans and designs remain subject to execution or validation.
stories per week
last 8 weeks66 stories73 reportsmostly DigiTimeslatest 20 Nov
AI
Reports describe a US$2 billion GlobalFoundries agreement to supply interposers for TSMC’s CoWoS ecosystem, potentially adding US-based packaging capacity. The listed deal coverage does not confirm terms, production timing, scale, or shipments. The latest stories concern US restrictions on Chinese firms’ access to advanced foundries and TSMC’s reported commitment to comply; they do not establish a change in the interposer deal’s status.
stories per week
last 8 weeks26 stories44 reportsmostly DigiTimeslatest 20 Nov
AINewHeating up
US restrictions on Chinese firms’ access to advanced compute are broadening across remote use and chip manufacturing. New reporting says a rule removes a key exemption for Chinese chip designers using third-country intermediaries, and TSMC says it will comply. The reports provide little detail on implementation or the rule’s scope. Separately, commentary says overseas AI compute routes face greater scrutiny, without specifying new rules or effects. Earlier reports describe uncertain impacts on Southeast Asian shipments and investment.
stories per week
last 8 weeks9 stories21 reportsmostly DigiTimeslatest 20 Nov
AI
AI data-center expansion continues to face power and infrastructure constraints, including grid access, permitting, public opposition and project viability. Suppliers are pursuing cooling, modular delivery and power systems, while 800V DC and power-chip demand remain prospective rather than established deployments. The latest report describes a geothermal deal as a potential source of low-carbon baseload for AI data centers, but its capacity, timeline and whether it serves Google’s facilities are not established. The reported near-100% North American power deficit remains unsubstantiated in the supplied assessment.
stories per week
last 8 weeks19 stories20 reports3 outletslatest 20 Nov
AIHeating up
Reports describe Samsung preparing its Texas footprint for advanced-node production, including planned trial runs in December and commercial production in 2027. The latest report says Samsung is transferring equipment to retool its Austin plant for prospective SpaceX business, building on an earlier report about Starlink chips. The scale and significance of the potential supply remain unclear, and the reports do not confirm production details.
stories per week
last 8 weeks6 stories6 reportssingle outlet · DigiTimeslatest 20 Nov
AI
Terafab remains an ambitious but unconfirmed chip-manufacturing effort. TSMC talks and a possible operating role have been reported, but the stories establish no manufacturing agreement, fab investment, construction plan or production schedule. New reporting says Terafab is in talks with Korean firms to buy chipmaking equipment, adding detail to sourcing activity but confirming neither a purchase nor a change in planned capacity. Its reported DRAM ambition remains unverified.
stories per week
last 8 weeks21 stories32 reportssingle outlet · DigiTimeslatest 19 Nov
AI
GlobalFoundries is positioning FD-SOI, including FDX Fusion and a reported Dresden process plan, for physical-AI applications in robots and cars. Reports also cite nearly US$1 billion in custom-chip design wins, but do not identify customers or establish production impact. Its reported US$10 million investment in India’s Cyient Semiconductors supports the company’s development and supply assurance; the scale and practical effects remain unclear. The stories do not verify FD-SOI performance, investment, capacity, or commercial adoption.
stories per week
last 8 weeks12 stories15 reports3 outletslatest 19 Nov
AI
Samsung has reported 2nm customer traction in the form of a Faraday ASIC order and repeat AI-chip orders from US cloud firms, but their scale, confirmation and production or shipment timing remain unclear. Reports also describe 2nm deals taking longer, while yield progress and the planned Taylor fab ramp remain prospective or unverified. The latest story points to incremental movement in foundry-market competition, but does not establish a corresponding change in manufacturing capability.
stories per week
last 8 weeks8 stories8 reportssingle outlet · DigiTimeslatest 18 Nov
AI
Solidigm’s US NAND expansion remains uncommitted, with funding and timing unresolved. SK Hynix has halted the contemplated IPO review, and no alternative financing or Kioxia-linked asset sale is confirmed. The proposed US$45 billion US fab also remains uncertain amid reported US approval issues related to equipment transfers for the Dalian plant. A new report says Solidigm’s 2026 profit is on track to exceed the shelved IPO’s US$10 billion target, but no financial disclosure or new financing decision is reported.
stories per week
last 8 weeks13 stories13 reportssingle outlet · DigiTimeslatest 18 Nov
AI
Reports describe strong AI-related demand for chips and semiconductor equipment, with capacity expansion constrained by supplier bottlenecks and broader issues including cleanrooms, workforce, construction, materials, power and financing. Applied Materials reported record fourth-quarter revenue amid accelerating demand and now targets doubling equipment output by 2028; the report does not quantify its baseline or establish how much of the target will be delivered. TSMC has reported surging AI-chip demand, without a quantified capacity change.
stories per week
last 8 weeks17 stories23 reportsmostly DigiTimeslatest 16 Nov
AI
US policy on AI data centers remains politically contested, with no listed report establishing that Texas or other states have enacted a pause or new oversight rules. Texas Governor-elect Hinojosa’s proposed freeze and review of SpaceX land deals remain plans, while Taiwan firms have committed US$2.9 billion to Texas. In North Dakota, a reported referendum tally indicates substantial local opposition to development, but its effects on projects are not established.
stories per week
last 8 weeks4 stories9 reportssingle outlet · DigiTimeslatest 14 Nov
AI
Meta’s Muse lineup now includes Muse Ember, described in new reports as a lightweight, lower-cost model for agent execution and high-volume work. The reports do not substantiate comparative capability claims with scores. Earlier releases include hosted Muse Wonder and Wonder Full, plus Muse Spark Open, whose weights are reported as available; benchmark claims, licensing, and practical openness remain to be verified.
stories per week
last 8 weeks5 stories18 reports8 outletslatest 9 Nov
AI · quiet
Starlink’s expansion spans satellite capacity, terrestrial mobile ambitions and aviation connectivity. The latest report names Qualcomm as an infrastructure partner for SpaceX’s planned network, but does not specify the agreement’s scope or practical effects. EasyJet plans a fleet-wide rollout beginning in the first half of 2027 and completing by early 2028. The reported 80,000-site push remains unclear, a proposed spectrum transfer awaits FCC review, and satellite deployment at routine scale is not established.
stories per week
last 8 weeks10 stories13 reports3 outletslatest 5 Nov
AI · quiet
LG is advancing AI data-center cooling and robotics alongside broader AI activities. The latest reported development is the partial start of operations at its robotics Data Factory, an incremental step whose scale and effect are not yet clear. Earlier reports describe cooling capacity and supply agreements, including a 5GW North American pipeline, and plans for production lines and a 2030 humanoid target, but provide limited evidence of realized deployment or commercial impact.
stories per week
last 8 weeks10 stories10 reportssingle outlet · DigiTimeslatest 2 Nov
AI · quiet
Mistral announced Large 4 in early October, with its capabilities and planned weights release then still unverified. By October 29, the weights had been released, with Mistral positioning the model for coding and defensive cybersecurity. Performance claims remain company-stated or otherwise unverified in the reports, and licence terms are not clearly stated.
stories per week
last 8 weeks6 stories12 reports8 outletslatest 29 Oct
AI · quiet
Reflection AI introduced Beam as a 501B-parameter, 23B-active open-weight model for coding and agentic workloads, and a later report described its release under Apache-2.0. Reported efficiency, benchmark and deployment claims remain unverified. Nvidia's reported exploratory interest in Reflection AI is unconfirmed, with no settled deal structure reported.
stories per week
last 8 weeks4 stories7 reports6 outletslatest 23 Oct
AIEmerging
Reports assess how the US midterm elections could affect semiconductor ties, AI data-center development and hardware trade. UK chip leaders and ODMs expect political change and increased opposition not to derail those plans, while earlier coverage flags potential risks for Samsung and SK Hynix and possible effects on China's AI exports. The reports provide little evidence of actual disruption so far.
stories per week
last 8 weeks3 stories4 reportssingle outlet · DigiTimeslatest 9 Nov
AI · quietEmerging
Amazon says it has stopped using NDAs with government agencies on data center projects, responding to community concerns about secrecy. Microsoft has made a similar move. The reports say the practical effect on local negotiations, approvals, and opposition remains unproven.
stories per week
last 8 weeks3 stories4 reportssingle outlet · TechCrunch AIlatest 11 Oct
AIEmerging
The administration has announced a Super Intelligence Force and appointed an AI czar, presenting these as steps in AI governance. A reported non-binding safety pact and rebranding may influence the policy conversation, but the reports do not establish changes to oversight or safety practices. The task force’s remit, membership, authority and resources remain unclear; a report is expected within 120 days.
stories per week
last 8 weeks3 stories4 reports3 outletslatest 5 Oct