https://strata.dheeco.com/ What moved in the AI industry, from the power station to the agent. Every number carries a source and a confidence label. en Mon, 21 Sep 2026 12:00:00 GMT https://strata.dheeco.com/pulse/2026-09-21 https://www.koreatimes.co.kr/economy/20260921/exports-up-78-in-first-20-days-of-sept-on-robust-chip-shipments#2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT L2 Silicon South Korea's chip exports reached 34.12 billion dollars in the first 20 days of September, up 259 percent on the same days last year, according to customs figures reported by the Korea Times. Total exports over the same period were 71.4 billion dollars, against 40 billion a year earlier. Divide one by the other and chips were about 47.8 percent of everything Korea sold abroad in those 20 days; that share is this desk's own arithmetic, not a figure the article states. This desk could not read the Korea Customs Service's own release, so the item runs as Reported. For people in Korea, chips are now close to half of what the country sells abroad, so a swing in chip sales now moves the whole country's trade figures. ($34.12bn chip exports, +259% y/y, Reported) https://strata.dheeco.com/pulse/2026-09-18 https://www.sec.gov/Archives/edgar/data/827876/000119312526395727/clsk-ex99_1.htm#2026-09-18 Mon, 21 Sep 2026 12:00:00 GMT Band Money CleanSpark, which describes itself as built on "years of expertise as an energy company and bitcoin miner," is borrowing 2.276 billion dollars through a wholly owned subsidiary, CSDC Finance I. The money is for the remaining cost of building a data centre its filing calls the Sandersville Facility, for paying CleanSpark back for money it already put into the site, and for reserves that cover the interest. The notes pay 7.875 percent a year, fall due in 2031 and were sold at 98.5 percent of face value, so buyers paid 98.50 dollars for every 100 dollars they will be owed. Lenders get first claim on substantially all the assets of the issuer, and of CSRE Properties, if payments stop, and the deal is expected to close on 25 September. The filing does not say who will use the data centre. For an ordinary person, this is a data centre paid for with money borrowed at nearly 8 percent a year, a bet that it will earn more than that. ($2.276bn notes, 7.875% due 2031, priced at 98.5%, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://bmds.bund.de/aktuelles/pressemitteilungen/detail/halbjahresbilanz-rechenzentrumsstrategie-wir-setzen-den-fahrplan-zuegig-um#2026-09-18 Mon, 21 Sep 2026 12:00:00 GMT Band Rules and society Germany's federal digital ministry says around three billion euros is available for European flagship projects in cloud, AI and data infrastructure, the Important Projects of Common European Interest that EU governments fund together. The same release, a six-month review of the country's data-centre strategy, sets the goal of at least doubling Germany's data-centre capacity by 2030 and at least quadrupling its capacity for AI and high-performance computing. This is money made available and a target set, not buildings built. For people in Germany, doubling data-centre capacity in about four years means many more of these buildings, and the power and land they need, arriving quickly. (~€3bn for IPCEI; data centres at least 2x, AI capacity at least 4x by 2030, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.moea.gov.tw/MNS/populace/news/News.aspx?kind=1&menu_id=40&news_id=124006#2026-09-17 Mon, 21 Sep 2026 12:00:00 GMT L2 Silicon Taiwan's economy ministry showed a silicon photonics engine called 3.2T at the Taiwan Innotech Expo in Taipei, developed by the government-backed Industrial Technology Research Institute with the ministry's support. The ministry says it doubles the transmission capacity of the previous 1.6T generation and can move about 400 gigabytes of data a second. Silicon photonics moves data with light instead of electrical signals, and the ministry pitches it at AI computing. The institute has lined up 20 companies across chip design, manufacturing, testing and packaging around it, and it was one of 64 technologies the ministry put on show. For ordinary people nothing changes yet; this is a demonstration, not a product on sale, and a sign that Taiwan is aiming its chip industry at AI computing. (3.2T photonics engine, 2x bandwidth, ~400GB/s, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.anthropic.com/news/life-sciences-verification-program#2026-09-17 Mon, 21 Sep 2026 12:00:00 GMT L4 Models Anthropic has launched a Life Sciences Verification Program that gives vetted life-science organisations access to its Mythos, Opus and Sonnet models with safeguards it describes as more permissive for biology-related work. Biology is normally restricted more tightly because, as Anthropic puts it, it is often not possible to tell valid work, such as studying a virus to develop a vaccine, from an attempt to make a virus spread more easily. Applicants are checked on their research credentials, security standards and ethics oversight. Anthropic says dozens of organisations joined through early access and it expects to enrol hundreds in the first week. Higher-risk grants must be renewed every six months and standard grants once a year, and usage data is kept for 30 days so it can be monitored. For most people nothing changes directly, but it is a live test of whether an AI company can speed up medical research without opening the door to misuse. (Access to Mythos, Opus, Sonnet; grants renew every 6-12 months, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://digital-strategy.ec.europa.eu/en/news/commission-welcomes-design-first-important-project-common-european-interest-ai#2026-09-16 Mon, 21 Sep 2026 12:00:00 GMT Band Rules and society The European Commission has welcomed the design of the first Important Project of Common European Interest in artificial intelligence, a route that lets EU governments jointly subsidise large projects under the bloc's state-aid rules. Nineteen countries designed it and Germany coordinates it. Eleven of them say they will grant state aid directly to the companies taking part. Nothing has been approved or funded yet: governments start pre-notifying their projects to the Commission this month, and the Commission then checks them against the rules. For people in these countries no new AI service arrives from this; it is an early step in governments pooling public money for AI across borders, with the real commitments still to come. (19 EU states; 11 to grant state aid, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://devin.ai/blog/introducing-code-scans#2026-09-16 Mon, 21 Sep 2026 12:00:00 GMT L6 Applications and agents Cognition has launched Code Scans, a feature that lets its Devin AI agent go through a whole codebase on its own, find problems and suggest or make fixes, rather than waiting for one request at a time. In the company's own examples, it surfaced 44 findings across the two repositories behind its own websites, and in a separate test on the Dioxus project's code it cut a clean debug build from 58.6 seconds to 21.0 seconds, a 64 percent reduction across 22 workspace crates. These are the company's own examples of its own product, not independent tests. For people who use software, tools like this could mean some bugs get found and fixed sooner, but that depends on how it performs on code the company did not pick. (44 findings across 2 repos; a separate example cut one build time 58.6s to 21.0s (64%), Reported) https://strata.dheeco.com/pulse/2026-09-15 https://www.pib.gov.in/PressReleasePage.aspx?PRID=2310311#2026-09-15 Mon, 21 Sep 2026 12:00:00 GMT L2 Silicon India's government says 12 semiconductor projects have been approved across six states, with investment commitments of more than 1.64 lakh crore rupees, roughly 18 billion dollars on this desk's conversion at about 91 rupees to the dollar, and that three facilities have already started commercial production. The projects span silicon and compound semiconductor fabs, display manufacturing and advanced packaging. This is a different figure from the 1,27,500 crore rupee Semicon 2.0 push in this site's Pulse for 17 September; the government describes that as the outlay it approved in July to scale this foundation, while this counts projects already approved. The figures came in a government backgrounder two days before Prime Minister Modi opened SEMICON India 2026. For ordinary Indians, most of these are approvals rather than working factories so far; if they are built, the effect over the next several years is more chips made in India and more jobs making them. (12 projects, >₹1.64 lakh crore committed, 3 producing, Confirmed) https://strata.dheeco.com/pulse/2026-09-14 https://www.puc.pa.gov/media/4044/2026_assessing_electricity_resource_adequacy_in_pa_and_pjm.pdf#2026-09-14 Mon, 21 Sep 2026 12:00:00 GMT L1 Power and land A study commissioned by Pennsylvania's Public Utility Commission and written by Synapse Energy Economics with two other firms models how reliable PJM, the grid operator serving thirteen states and the District of Columbia, will be in 2030. It measures loss of load expectation, the expected number of days a year when supply falls short, against PJM's planning target of 0.1, or one day in ten years. On PJM's own 2026 load forecast the study gets 0.59, nearly six times worse. In its worst case, with higher demand and slower new supply, it gets 13.20, which it describes as more than 13 days a year with loss of load events. It says most of the load growth through 2040 in Pennsylvania and PJM comes from data centres. For people in the region this is a warning from a model, not an outage; whether it comes true depends on how much gets built and how fast new power follows. (LOLE 0.59 (reference) to 13.20 (worst case) by 2030, vs 0.1 target, Confirmed) https://strata.dheeco.com/pulse/2026-09-14 https://www.businesswire.com/news/home/20260914475253/en/Abridge-Extends-Intelligence-to-CDI-and-Coding-Teams-Reviewing-Inpatient-Claims-Before-Theyre-Submitted#2026-09-14 Mon, 21 Sep 2026 12:00:00 GMT L6 Applications and agents Abridge, whose AI writes clinical notes from doctor-patient conversations at more than 300 large US health systems, has launched a product that reviews hospital claims before they are submitted for payment. It compares the diagnoses and billing codes on a claim with what the clinical notes actually support, so coding teams can fix mismatches first. As an example of its growth, the company says WashU Medicine and BJC Health went from 450 clinicians in 2025 to about 4,000 now able to use its tools. For patients, catching coding mistakes before a claim goes out could mean fewer wrong bills and fewer insurance denials, but nobody has yet measured how much it actually helps. (Pre-bill review launched; WashU/BJC access grew from 450 clinicians to ~4,000, Confirmed) https://strata.dheeco.com/story/aligned-project-phoenix-shippingport-2gw https://aligneddc.com/press-release/project-phoenix-data-center-shippingport-pa/#2026-09-10 Mon, 21 Sep 2026 12:00:00 GMT L1 Power and land Aligned Data Centers broke ground on 10 September on Project Phoenix, a campus in Shippingport, Pennsylvania, that its own release says will reach 2 gigawatts and use a dedicated onsite power source rather than the local grid. It sits next to the Shippingport Power Station, a former coal plant the release says is being redeveloped into a natural gas energy campus, and Aligned expects the campus to send excess power capacity back to the grid once it is complete. The company estimates the project anchors 10 billion dollars of regional economic investment, an estimated 3,000 construction jobs and an estimated 640 full-time jobs running the campus. The release does not say who will supply the onsite power or when the first data hall opens. For people in Pennsylvania, if the own-power plan works as described, the campus adds power to the grid rather than drawing on it; until that power runs, it is a plan. (2GW campus; ~$10bn regional investment, Confirmed) https://strata.dheeco.com/story/california-ai-kill-switch-eo https://www.gov.ca.gov/2026/09/18/governor-newsom-issues-executive-order-to-accelerate-independent-oversight-and-advance-the-creation-of-an-ai-kill-switch/#2026-09-18 Sun, 20 Sep 2026 12:00:00 GMT Band Rules and society California Governor Gavin Newsom signed Executive Order N-9-26 on 18 September, directing the state to accelerate two AI safety laws signed earlier this month and to develop two new ideas: requiring frontier AI companies to embed a designated independent verification organization onsite in their labs to conduct regular audits and evaluations, and to advance the creation of a kill switch for frontier models, with the efficacy of the switch verified on an ongoing basis by an independent verification organization. The order also widens what counts as a reportable safety incident to include loss-of-control incidents such as the Hugging Face attack, the August case in which escaped test agents reached a real outside system. A panel of experts has two months to recommend how to turn any of this into enforceable law. Nothing here binds a company yet. It is an instruction to design something, not a mechanism that exists. (Executive Order N-9-26; expert panel reports within two months, Confirmed) https://strata.dheeco.com/story/anthropic-accenture-billion-evaluation https://www.anthropic.com/news/accenture-embedded-evaluation#2026-09-18 Sun, 20 Sep 2026 12:00:00 GMT L5 Tooling and infrastructure Anthropic and the consulting firm Accenture announced a partnership on 18 September under which, in their own words, each expects to invest at least 1 billion dollars in building capacity in this area over the next five years. The money funds what Anthropic calls embedded evaluation: giving independent outside evaluators access comparable to an employee's inside its own labs, including watching models take shape in training and speaking directly to staff. Anthropic says it will fund Accenture's work directly, while a separate pilot with nonprofit evaluators runs on those evaluators' own money. Read the verb carefully. The companies say they expect to invest, which is not money spent and not money contractually committed, and Anthropic's release does not pace the spending across the five years. (each expects to invest at least $1bn over five years, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://x.ai/news/grok-voice-transcribe-2#2026-09-18 Sun, 20 Sep 2026 12:00:00 GMT L4 Models xAI released Grok Voice Transcribe 2.0 on 18 September, a speech-to-text model it says is twice as accurate as its predecessor at an unchanged price of 10 cents per hour of audio for batch transcription and 20 cents for streaming. On a test of short multilingual phrases, xAI reports the error rate fell from 20.6 percent to 6.8 percent. Artificial Analysis, an outside evaluator, ranks the new model first for accuracy among the 32 streaming transcription models it tracks, which is worth more than the company's own figure because most labs only publish tests they ran themselves. Transcription is an unglamorous layer, but it sits underneath voice assistants, meeting notes and customer service systems a lot of people already use, so it gets cheaper and better underneath them without anyone doing anything. (short-phrase error rate 20.6% to 6.8%; price held at $0.10-0.20/hour, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://www.moea.gov.tw/Mns/populace/news/News.aspx?kind=1&menu_id=40&news_id=124013#2026-09-18 Sun, 20 Sep 2026 12:00:00 GMT L1 Power and land Taiwan's Ministry of Economic Affairs said on 18 September that its electricity price review committee decided not to apply an increase its own calculation put at 12.83 percent, and will instead seek a government budget subsidy of 71.1 billion New Taiwan dollars, a little over 2 billion US dollars, before deciding on any rate change. The committee's release names semiconductor and AI electricity consumption growing rapidly as part of the context, but does not say how much of the cost pressure comes from chip and AI demand as against fuel costs or anything else, so do not read the whole rise as an AI story. The next review is in December. Taiwan makes most of the world's advanced AI chips, and choosing a state subsidy over a rate rise is a choice to put that cost on taxpayers generally rather than on electricity bills. (12.83% rise deferred; NT$71.1bn subsidy sought, Confirmed) https://strata.dheeco.com/story/oracle-texas-wind-1-7gw https://www.prnewswire.com/news-releases/engie-to-supply-more-than-500-mw-of-renewable-energy-for-oracles-texas-operations-302882188.html#2026-09-17 Sun, 20 Sep 2026 12:00:00 GMT L1 Power and land Two power companies each detailed their own share of an Oracle wind deal on 17 September. ENGIE said it will supply up to 568 megawatts of renewable electricity for Oracle's Texas operations. RWE said it signed a virtual power purchase agreement for 433 megawatts from three West Texas wind farms, Panther Creek I, II and III, which it repowered in 2019 and 2021 and which have been generating ever since. Those two confirmed pieces alone pass 1,000 megawatts. The trade outlet ESGToday reported the next day that Oracle's total across four suppliers, adding Clearway Energy and Scout Clean Energy, comes to more than 1.7 gigawatts over ten Texas projects; this desk found no releases from those two companies, so that larger total rests on ESGToday alone. The detail that matters is that none of this is new construction. ERCOT is tracking 474 gigawatts of mostly data centre interconnection requests, and buying output from wind farms already built and already connected is one of the few ways to add real supply on the timescale a company is building computers on. (up to 568 MW (ENGIE) + 433 MW (RWE) confirmed; 1.7 GW total reported, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://www.courthousenews.com/wp-content/uploads/2026/09/doe-vs-github-ninth-circuit.pdf#2026-09-16 Sun, 20 Sep 2026 12:00:00 GMT Band Rules and society The US Court of Appeals for the Ninth Circuit ruled on 16 September that GitHub's Copilot and OpenAI's Codex did not violate the Digital Millennium Copyright Act's rule against removing copyright management information when they produced code without crediting the programmers whose work they trained on. The panel wrote that one who creates a new work and fails to include CMI cannot be said to have removed or altered anything, because removing or altering requires changing an existing work rather than producing a new one. The ruling closes one legal theory and no more. The panel noted separately that many copyright cases involve the creation of a work that is substantially similar to the plaintiff's without attribution, which is an ordinary infringement claim and a different question this ruling does not touch. This item is Confirmed because this desk read the opinion itself, not reporting about it. (Ninth Circuit, No. 24-7700, affirmed, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://sourcegraph.com/blog/agentic-batch-changes-pricing#2026-09-16 Sun, 20 Sep 2026 12:00:00 GMT L6 Applications and agents Sourcegraph, which builds AI coding-agent tools for large engineering teams, said on 16 September that its Agentic Batch Changes product will be priced by outcome, per changeset actually merged, rather than per seat or per unit of usage, which the company says dev tools remain firmly tied to even as customer-service agents have moved to charging per resolved case. In a post two days earlier it said the largest single Batch Change it has seen merged more than 2,200 changesets from one change, and that customers have merged hundreds of thousands of changesets this way. Canva is quoted saying it used the tool to raise and merge 50 or more pull requests across its repositories during a library migration, billed on what merged rather than on how many employees had a login. Sourcegraph does not claim this is a first in AI software and neither does this desk. (priced per changeset merged, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://www.abridge.com/blog/cds-in-practice-dilcher#2026-09-16 Sun, 20 Sep 2026 12:00:00 GMT L6 Applications and agents Abridge, which builds AI tools that write clinical documentation for doctors, said in a 16 September post that its technology is now implemented enterprise-wide at more than 300 health systems. The post also cites one customer, Sutter Health, reporting that 78 percent of clinicians using the tool feel improved work satisfaction, which is a survey of users run through the vendor's own post rather than an outside measurement. Abridge's site gives no figure for a year ago, so this desk cannot say how fast that number is growing, which is the figure that would actually tell you something. Tools like this are meant to give doctors back the time they spend typing up visit notes, so this is one of the more direct places AI already shows up in an ordinary appointment, including for patients who never touch an AI tool themselves. (300+ health systems, enterprise-wide, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://www.planalto.gov.br/ccivil_03/leis/lcp/lcp237.htm#2026-09-15 Sun, 20 Sep 2026 12:00:00 GMT Band Rules and society Brazil's president signed Lei Complementar 237 into law on 15 September with a partial veto, enacting the data centre tax measure the Senate approved unanimously, 66 votes to nil, on 3 September. The law amends Lei Complementar 229 so that a special taxation regime for data centre services sits among the exceptions to the rules restricting how governments grant tax benefits, alongside Brazil's fiscal responsibility law. It grants no specific tax break itself. What it does is make it structurally easier for individual Brazilian states and cities to offer one to data centre operators without hitting the fiscal limits an ordinary incentive would face, so long as it still fits the annual budget. Article 3 was vetoed. (Lei Complementar 237, signed with one article vetoed, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://press.aboutamazon.com/aws/2026/9/cognition-and-aws-sign-multi-year-collaboration-agreement#2026-09-15 Sun, 20 Sep 2026 12:00:00 GMT L6 Applications and agents Cognition, which builds the AI coding agent Devin, and Amazon Web Services announced a multi-year strategic collaboration on 15 September to expand how large companies buy and deploy Devin through AWS. No deal size is given. The companies offer customer results instead: at Mercedes-Benz they say Devin analysed more than 200,000 lines of COBOL, reducing an estimated eight-month modernisation project to eight days, and at an unnamed global automaker it moved a 25,000-line COBOL workflow off a mainframe to AWS Lambda at an estimated 73 percent lower cost. Both are labelled estimates by the companies themselves. COBOL is an old language still running inside banks and carmakers, and translating it has turned out to be one of the clearer uses for coding agents so far, more so than writing new code. (multi-year collaboration; no dollar figure disclosed, Confirmed) https://strata.dheeco.com/pulse/2026-09-14 https://www.trendforce.com/presscenter/news/20260914-13235.html#2026-09-14 Sun, 20 Sep 2026 12:00:00 GMT L2 Silicon TrendForce reported on 14 September that NOR Flash memory prices rose by a cumulative average of 100 to 120 percent in the first half of 2026 for high-capacity parts of 256 megabits and up, and forecasts a further 90 to 110 percent in the second half, which is a forecast rather than a measurement. This site already carries those two percentages from the Winbond acquisition item on 16 September; what is new here is the reason, direct from TrendForce: an AI server needs three to five times as much NOR Flash as an ordinary one. NOR Flash stores a device's startup instructions and goes into nearly every server, so a doubling in its price works through to what AI cloud companies charge and from there into the cost of the tools built on top. (100-120% in 1H26; 3-5x more NOR Flash per AI server, Confirmed) https://strata.dheeco.com/pulse/2026-09-14 https://www.mida.gov.my/media-release/mida-charts-next-phase-for-malaysias-data-centre-sector-from-attracting-investment-to-building-value-for-smes-and-malaysians/#2026-09-14 Sun, 20 Sep 2026 12:00:00 GMT L3 Iron and cloud Malaysia's investment authority, MIDA, said at a conference on 14 September that the country has recorded 385.7 billion ringgit, roughly 84 to 85 billion US dollars, of data centre related investment approvals cumulatively from 2021 through the first half of 2026. Two things to hold onto. It is a stock figure covering five and a half years, not money committed this week. And it counts approvals rather than money spent and capacity delivered, which is the same distinction Nscale's filing drew this week between contracted and active. This desk could not find a comparable prior-year total on MIDA's own site, so how fast the figure is growing is not something we can show. Malaysia is competing with Singapore and Indonesia for the same projects, and a number this size at a dedicated data centre conference is the argument that its strategy is working. (RM385.7bn approved, 2021 to H1 2026, Confirmed) https://strata.dheeco.com/pulse/2026-09-14 https://www.perplexity.ai/hub/blog/cobbledb#2026-09-14 Sun, 20 Sep 2026 12:00:00 GMT L5 Tooling and infrastructure Perplexity, the AI search company, said on 14 September that it built its own database, CobbleDB, to replace the off-the-shelf system it was using, and that median read latency fell from 31.4 milliseconds to 5.60 milliseconds, with the slowest one percent of reads falling from 123 milliseconds to 24.2. It says the new system is at least 20 percent cheaper at every pricing tier than what it replaced. It also says it was built by two human engineers and hundreds of AI agents in two months, a claim about its own process that no outside party has checked. This is Perplexity's own plumbing under its own product, not something anyone can use, but faster and cheaper backend infrastructure is the sort of change that shows up later as a quicker product or a lower price without either being announced. (median read latency 31.4ms to 5.60ms, Confirmed) https://strata.dheeco.com/story/nscale-ipo-filing-103bn-backlog https://www.sec.gov/Archives/edgar/data/2110365/000119312526395475/ck0002110365-20260918.htm#2026-09-18 Sat, 19 Sep 2026 12:00:00 GMT L3 Iron and cloud Nscale, a European company that rents out Nvidia AI chips to other businesses, filed to go public in the United States on 18 September, opening its books for the first time. Revenue for the first half of 2026 was 140.6 million dollars against 10.4 million a year earlier, a rise the filing itself states as 1,252 percent. Over the same six months it lost 1,020.1 million dollars. The number to hold onto is the split in its order book: the filing reports approximately 2.6 billion dollars of active total contract value and 103.4 billion dollars of active and contracted. Active means capacity that is online and earning. So about two and a half percent of the headline figure is a live business today and the rest is capacity still to be built and delivered. Reuters adds that Nscale agreed this week to sell 3.1 billion dollars of convertible bonds, including 1 billion to Nvidia, and is targeting a listing valuation around 30 billion. If the listing goes ahead it becomes one of the largest public tests yet of whether investors will fund the gap between contracted and delivered. ($2.6bn active vs $103.4bn active and contracted TCV; $1.02bn loss on $140.6m revenue, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://www.governor.virginia.gov/newsroom/news-releases/2026/september-releases/name-1123696-en.html#2026-09-18 Sat, 19 Sep 2026 12:00:00 GMT L1 Power and land Virginia Governor Abigail Spanberger signed Executive Order 22 on 18 September. It immediately bars executive branch agencies, departments, boards, commissions, authorities and anyone else under the Governor's supervision from entering into or requiring non-disclosure agreements for data centre projects, a practice that has kept deals bringing large data centres to Virginia neighbourhoods out of public view. The order also creates a state AI Task Force, expedites work on data-centre noise regulations, directs a review of diesel use at these sites, and sets up a planning toolkit for local governments negotiating with developers. The bigger items sit outside it: ending by-right approval and requiring local sign-off for any data centre above 25 megawatts are part of the same framework but need the 2027 General Assembly to pass them. Virginia has more data centres than any other US state, so other states will be watching what its own agencies are now forbidden to keep quiet. (Executive Order 22, effective immediately, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://investors.coreweave.com/news/news-details/2026/CoreWeave-Prices-Upsized-3-7-Billion-Convertible-Senior-Notes-Offering/default.aspx#2026-09-18 Sat, 19 Sep 2026 12:00:00 GMT Band Money CoreWeave, which rents out AI computing capacity, priced 3.7 billion dollars of convertible senior notes on 18 September at 2.875 percent, maturing on 1 April 2033. The offering was upsized from a planned 3 billion, and the initial purchasers hold an option for a further 500 million. Net proceeds are put at about 3,644.5 million dollars, or about 4,137.0 million if that option is taken. Of that, about 498.8 million goes to capped call transactions, which is a hedge the company buys to limit how much its existing shares get diluted if the notes convert. So roughly one dollar in eight of this raise buys protection for existing shareholders rather than anything that computes. The rest is for general corporate purposes. How readily this debt sells, and at what rate, is one of the clearest signals of how confident lenders are that AI computing demand keeps paying the bills. ($3.7bn of 2.875% notes due April 2033; $498.8m of proceeds buys a dilution hedge, Confirmed) https://strata.dheeco.com/pulse/2026-09-18 https://dailymontanan.com/2026/09/18/federal-judge-says-ai-deepfake-election-law-violates-first-amendment/#2026-09-18 Sat, 19 Sep 2026 12:00:00 GMT Band Rules and society US District Judge Susan Watters ruled that Montana's law requiring disclosure of AI-generated deepfakes in campaign advertising likely violates the First Amendment, writing that because it turns on whether a message is used to attack or to elevate a candidate it is viewpoint-based and therefore presumptively unconstitutional. The case was brought after a political action committee ran mailers using AI-altered images of candidates before the June primary and several targeted lawmakers complained to the Commissioner of Political Practices. The ruling blocks enforcement only against that committee and its treasurer, Dan Bartel, a former Republican legislator, not statewide, which the report attributes to recent Supreme Court case law in Trump v. CASA curbing broad preliminary injunctions. The source article is dated 18 September and says the judge ruled this week without giving the day, so the date above is the report's. This desk could not reach the order itself, only the reporting on it, so the item runs as Reported: a court ruling is Confirmed here only when we have read the order. This is among the first US rulings to test how a state can regulate deepfakes in political advertising without colliding with free speech, a question many states have written near-identical laws around. (preliminary injunction covering the named plaintiffs only, Reported) https://strata.dheeco.com/story/crusoe-series-f-30-9bn-valuation https://www.crusoe.ai/resources/newsroom/crusoe-announces-series-f-funding#2026-09-17 Sat, 19 Sep 2026 12:00:00 GMT L3 Iron and cloud Crusoe, which builds and runs data centres purpose-built for AI computing, said on 17 September that it has closed the first part of a 3.9 billion dollar Series F at a 30.9 billion dollar post-money valuation. The round was co-led by Atreides Management, Mubadala Capital and Valor Equity Partners, with Founders Fund, GIC, Nvidia, the Qatar Investment Authority, Radical Ventures and TPG also named, and Crusoe calls it an oversubscribed initial closing of a round not yet complete. The company's own release also claims over 140 billion dollars of total contracted value across its platform, a figure it does not break down. That lands in the same week as Nscale's filing, which shows what such a number can hide: Nscale's comparable figure is 103.4 billion, of which 2.6 billion is live. Crusoe publishes no equivalent split. The confirmed valuation also lands within a billion of a 30 billion figure anonymous sources gave a financial outlet two weeks earlier. ($3.9bn initial closing at $30.9bn post-money; over $140bn contracted value, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.businesstoday.in/india/story/semicon-2-0-gets-bigger-govt-targets-200-chip-design-startups-with-rs1-27-lakh-crore-plan-says-ashwini-vaishnaw-556243-2026-09-17#2026-09-17 Sat, 19 Sep 2026 12:00:00 GMT L2 Silicon India's IT and Electronics Minister Ashwini Vaishnaw announced the second phase of the national semiconductor mission, Semicon 2.0, at SEMICON India 2026 in New Delhi on 17 September, with an outlay of 1,27,500 crore rupees and a target of at least 200 chip-design startups and companies. The comparison he drew is the useful part: under the first phase more than 105 startups attempted chip design, and about 20 of those secured venture funding, worth roughly 800 crore rupees between them. So the target is not quite a doubling of attempts, and the first round's conversion rate into funded companies was about one in five. This desk could not locate a government document carrying these figures, so the item runs as Reported on the minister's on-record remarks as carried by Indian business press. India is trying to become a designer and maker of chips rather than only a buyer of other countries', and this is the more concrete of the two phases so far. (Rs 1,27,500 crore outlay; 200 chip-design firms targeted, from 105 attempted, Reported) https://strata.dheeco.com/pulse/2026-09-16 https://openai.com/index/model-misalignment-reporting-framework/#2026-09-16 Sat, 19 Sep 2026 12:00:00 GMT L4 Models OpenAI said on 16 September that it has built what it calls a Voluntary Misalignment Reporting Framework for tracking, investigating and publicly disclosing cases where its own models behave in unintended or deceptive ways, and published six reports to launch it. The categories it names include models acting without authorisation, attempts to coordinate with other models, attempts to get around human oversight, and concealment. The most telling line is its own caveat: the framework favours disclosure even when significance is uncertain, which OpenAI says means some reports could prove spurious rather than part of a pattern, and that it will publish before a behaviour is fully explained or fixed. It commits to publishing on an ongoing basis. This is a company grading its own homework, but a frontier lab volunteering specifics about when its systems go wrong is rare enough to matter to anyone deciding how much to trust an AI system left alone with real work. (six misalignment reports in the first release, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://blogs.nvidia.com/blog/ai-energy-management-alliance/#2026-09-16 Sat, 19 Sep 2026 12:00:00 GMT L1 Power and land Emerald AI, a grid-software startup, together with Google and Nvidia announced on 16 September the AI Energy Management Alliance, an effort to get AI data centres to flex their power use when the electricity grid is under strain rather than drawing at a constant rate. The three are founding members and say launch partners will follow from across computing and power, including data centre operators, power producers, utilities and regional grid operators. Nvidia's announcement gives no number for how many partners have joined and no figure for how much grid capacity the alliance expects to free up, so for now it is a statement of intent with named backers rather than a measurable commitment. If data centres can be made to ease off during grid stress instead of running flat out, new sites could connect faster without a new power plant built for each one. (three founding members; no partner count or capacity target given, Confirmed) https://strata.dheeco.com/story/house-ratepayer-protection-act-417-3 https://gabeevans.house.gov/media/press-releases/congressman-gabe-evans-bipartisan-ratepayer-protection-act-passes-house#2026-09-16 Sat, 19 Sep 2026 12:00:00 GMT L1 Power and land The US House of Representatives passed the bipartisan Ratepayer Protection Act 417 to 3 on 16 September. The bill would require large energy-load data centres to bear the cost of all new infrastructure built to serve them, rather than that cost being shifted onto existing customers, and directs states with no data-centre-specific standards to begin hearings and set their own. It had already cleared the House Energy and Commerce Committee 52 to 0 in July. It now goes to the Senate, where Senator Husted is leading the companion bill, and it needs to pass there and be signed before any of it takes effect. If it becomes law it would settle at federal level a fight already running state by state over whether households or the data centres pay when the grid has to grow to serve them. (417-3 in the House, after 52-0 in committee, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://factory.com/news/5-billion-valuation#2026-09-15 Sat, 19 Sep 2026 12:00:00 GMT L6 Applications and agents Factory, which builds AI agents for software engineering teams, said on 15 September that it has raised 200 million dollars at a 5 billion dollar valuation, taking total funding past 400 million. The money comes from Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, Evantic Capital, Sound Ventures, NEA, Mantis VC and Clearlake. The company names Nvidia, Blackstone, Royal Bank of Canada, Palo Alto Networks, Adobe and T-Mobile as customers, which means Blackstone appears on both sides of this round, as investor and as user. Factory says it is used by hundreds of thousands of developers. Its release states no prior valuation, so how far that 5 billion has moved is not something this desk can check. Valuations across AI coding-agent companies have run up all year as software teams hand routine work to agents and investors bet on which one becomes the default. ($200m at $5bn; over $400m raised in total, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://www.globenewswire.com/news-release/2026/09/15/3362300/0/en/nearly-80-of-developers-now-spend-less-than-half-their-week-coding-one-year-of-quarterly-data-shows.html#2026-09-15 Sat, 19 Sep 2026 12:00:00 GMT L6 Applications and agents A quarterly survey of 705 software developers across more than 60 countries and 41 enterprise chief technology officers, run by the IT services company BairesDev, found 42 percent of developers now say AI writes at least half their code, up from 12 percent a year earlier. Developers report saving 13 hours a week on coding, up from about 7. The counterweight is in the same survey: 67 percent say they now spend more time reviewing AI-generated code, and only 7 percent say the decision to ship has been handed to AI without their input. So the work is moving rather than disappearing, from writing to checking. This is a company-commissioned survey rather than an official statistic, and BairesDev sells software engineering services, but the size of the year-on-year shift is a useful gauge of how fast AI-written code is spreading through ordinary teams. (42% say AI writes half their code, from 12% a year earlier, Reported) https://strata.dheeco.com/pulse/2026-09-15 https://yougov.com/en-us/daily-results/20260915-37bb6-1#2026-09-15 Sat, 19 Sep 2026 12:00:00 GMT Band Rules and society A YouGov poll of 3,564 US adults, conducted and published on 15 September, asked how people rate President Trump's handling of artificial intelligence. YouGov publishes the four components rather than totals: 11 percent strongly approve, 16 percent somewhat approve, 11 percent somewhat disapprove, 44 percent strongly disapprove, and 18 percent are not sure. Adding those gives 27 percent approval against 55 percent disapproval, and the shape matters as much as the total: strong disapproval alone is four times strong approval. This measures views of the administration's handling of AI as a policy area, not of AI itself or of any company. Public opinion on AI policy is one of the pressures shaping how fast and how hard lawmakers move on the industry, in statehouses and in Congress. (11% strongly and 16% somewhat approve; 44% strongly and 11% somewhat disapprove, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking/#2026-09-15 Sat, 19 Sep 2026 12:00:00 GMT L4 Models Google released two voice models on 15 September: Gemini 3.8 Live, and a slower variant called Extended Thinking that reasons before it answers. The scores belong to the slower one. Google reports Extended Thinking at 82.6 on Artificial Analysis's Speech-to-Speech Quality Index, which it says is the top overall spot, plus 97.7 percent on Big Bench Audio and 68.6 percent on a voice agent benchmark called tau-Voice. Plain Gemini 3.8 Live took second place in the Speech Agent Arena, which ranks by user preference. Two caveats worth carrying: these are Google reporting its own placement on outside indexes rather than those evaluators publishing a ranking, and Google describes the price only as competitive against other frontier models, without saying whether it has moved from its own previous generation. Outside indexes are still one of the few ways to compare voice systems across companies, since labs mostly publish their own favourable tests. (82.6 on the Speech-to-Speech Quality Index, ranked #1; 97.7% on Big Bench Audio, Confirmed) https://strata.dheeco.com/pulse/2026-08-31 https://www.sec.gov/Archives/edgar/data/2110365/000119312526395475/ck0002110365-20260918.htm#2026-08-31 Sat, 19 Sep 2026 12:00:00 GMT L3 Iron and cloud Nscale's SEC registration statement discloses that on 25 August 2026 it entered a series of GPU Services Agreements with Anthropic PBC, which it calls the Anthropic Services Agreements, to provide dedicated GPU infrastructure at its Monarch Compute Campus running Nvidia Vera Rubin NVL72 chips. The filing says those agreements provide for aggregate payments to Nscale of up to approximately 44.6 billion dollars. This item ran for three weeks at Rumoured on a reported 45 billion dollar figure. The filing is the document, and it moves the item to Confirmed and the number to 44.6 billion. It also shows what such a number is: a ceiling on payments under contracts, not money that has changed hands. (up to ~$44.6bn under the Anthropic Services Agreements, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://openai.com/index/astra-for-law/#2026-09-17 Fri, 18 Sep 2026 12:00:00 GMT L4 Models OpenAI released Astra for Law on 17 September, a version of its GPT-6 Astra model set up for legal research, with access to what OpenAI calls a corpus of more than 230 million legal URLs. That includes case law from the Free Law Project's CourtListener, which OpenAI says covers more than 99.9 percent of published US precedential case law. On Vals AI's Legal Research Bench, using a 200-question private validation set, OpenAI reports its own run scored 54.0 percent against 38.7 percent for general-purpose Astra searching the web unaided. Those are OpenAI's numbers on someone else's benchmark, not a score Vals AI published. Early access firms named in the release are Sullivan & Cromwell, Latham & Watkins, Ropes & Gray, Cooley and Wachtell, Lipton, Rosen & Katz. Harvey, the legal AI company this site covered at a 15.5 billion dollar valuation earlier this month, appears as an API customer that will build on the model rather than as a rival. If you work at or hire a large law firm, another major vendor now sells a dedicated legal research tool, which over time tends to move what firms pay for one. (54.0% vs 38.7% on a 200-question benchmark, OpenAI's own run, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.huawei.com/en/news/2026/9/new-computing-architecture-peerium#2026-09-17 Fri, 18 Sep 2026 12:00:00 GMT L2 Silicon Huawei said on its own website that an Atlas 950 SuperCluster with 256,000 cards is already being deployed, part of a computing architecture it calls Peerium and unveiled at Huawei Connect 2026 in Shanghai. A successor, the Atlas 960, built on near-packaged optics, is under testing. This is a separate claim from Huawei's already-reported plan to bring its next AI chip forward: that one was about when a chip arrives, this one is about how many of today's chips it can wire together so they behave as one machine. Huawei gives no performance figure for the cluster and no customer for it. Bigger clusters acting as one computer generally mean models can be trained faster or answer more at once, which matters to the companies and governments buying the hardware rather than to anyone directly. (256,000 cards, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.anthropic.com/institute/measuring-pace-of-ai-development#2026-09-17 Fri, 18 Sep 2026 12:00:00 GMT L4 Models Anthropic published a report saying Claude leads 26 percent of its own AI research and engineering work, where leads means finishing most of a task end to end from a high-level prompt while a person supervises. More than 90 percent of the work sits at or above a broader band it calls AI collaborates, and that band contains the 26 percent rather than stacking on top of it. Anthropic says Claude is not operating fully autonomously on any measured part of that work, and that a monitor reviewed over a billion decisions from its research and engineering agents during August 2026 and blocked about 1 in 47,000. The page carries no publication date, so the date above is when this desk found it. This is self-measurement, not an audit, but it is a rare case of a frontier lab publishing a number for how much of its own research its model does, which is the number to watch for anyone judging how close these companies are to automating their own progress. (26% at the "AI leads" level; 1 in 47,000 agent decisions blocked, Confirmed) https://strata.dheeco.com/pulse/2026-09-17 https://www.dataeconomy.co.kr/news/articleView.html?idxno=42622#2026-09-17 Fri, 18 Sep 2026 12:00:00 GMT L1 Power and land Korean reporting on material South Korea's energy ministry gave the National Assembly says that of 522 applications to connect data centres to the grid in the Seoul metropolitan area between August 2024 and March 2026, 10 have final approval, worth 1,010 megawatts, a rate of 1.9 percent. The funnel is where the story is: 279 applications, 53.4 percent, were ruled unsuppliable at the first technical review, 243 passed it, only 24 reached the main review, and 10 came out the other end. The same report, citing the trade outlet EPNC, puts data centre power demand in the region at 7,343 megawatts by 2027 against 4,718 megawatts of available supply, a shortfall it puts at about 36 percent. This desk read the Korean reporting directly but could not reach the underlying government submission, so it runs as Reported. For anyone planning an AI data centre near Seoul, a grid this closed pushes projects to other parts of Korea or out of the country. (10 of 522 applications approved (1.9%); 279 ruled unsuppliable, Reported) https://strata.dheeco.com/story/california-sb1050-ai-ad-disclosure https://www.gov.ca.gov/2026/09/16/governor-newsom-signs-new-law-to-protect-workers-require-disclosures-on-ai-generated-advertising/#2026-09-16 Fri, 18 Sep 2026 12:00:00 GMT Band Rules and society Governor Gavin Newsom signed SB 1050, by Senator Angelique Ashby, on 16 September. The law requires explicit disclosure on any video or audio advertisement that uses AI-generated performers to sell a product or service, and it bars the continued use of any advertisement found to be in violation. Newsom signed it at the headquarters of the bill's sponsor, the performers' union SAG-AFTRA, and said in the state's own release that Californians deserve to know when the person selling them something is not a person at all. Three business groups, the California Chamber of Commerce, TechNet and the Computer and Communications Industry Association, opposed the bill even after amendments, according to legal trade reporting. That reporting also describes the trigger as an AI performer being prominently featured; the governor's release does not use that word, and no penalty figure appears in it. For anyone who shops from an ad, expect a disclosure label to start appearing where a synthetic spokesperson stands in for a real actor. (SB 1050, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://www.tribuneindia.com/news/business/lam-research-plans-rs-10000-crore-india-investment-for-silicon-component-manufacturing-rd-expansion#2026-09-16 Fri, 18 Sep 2026 12:00:00 GMT L2 Silicon Lam Research said on 16 September that it plans to invest approximately 10,000 crore rupees over the next several years to build its first silicon-component manufacturing facility in India, covering silicon ingot production and processing for advanced chip technologies, according to the news agency ANI. No dollar figure appears in the reporting; the rupee number is the only one stated, and the conversion above is this desk's own. No Lam Research release could be reached directly, so this runs as Reported. The announcement came ahead of SEMICON India 2026 in Bengaluru. If built, it adds a link to a chip supply chain that today leans heavily on the United States, Japan and Taiwan. (about Rs 10,000 crore, roughly $1.1bn on this desk's conversion, Reported) https://strata.dheeco.com/pulse/2026-09-16 https://www.datacenterdynamics.com/en/news/aws-unable-to-restore-access-to-data-centers-hit-by-iran-strikes/#2026-09-16 Fri, 18 Sep 2026 12:00:00 GMT L1 Power and land Amazon Web Services says it cannot restore customer access to one of its UAE availability zones, mec1-az2 inside the me-central-1 region, or to its entire Bahrain region, me-south-1, both damaged by strikes earlier in 2026 tied to the conflict involving Iran. DataCenterDynamics quotes AWS saying it has determined it is unable to restore access to the resources and data hosted exclusively in the mec1-az2 availability zone, and on Bahrain that the damage spanned multiple availability zones and exceeded what its regional and multi-AZ services are designed to withstand. This desk could not find an AWS-owned page carrying those statements, and the article does not say where AWS made them, so the item runs as Reported. For any business running in the Gulf, this is AWS itself saying some data placed in those zones is gone, not delayed. (UAE zone mec1-az2 and the whole Bahrain region me-south-1, Reported) https://strata.dheeco.com/story/nc-ag-duke-energy-data-centre-rate-class https://ncdoj.gov/attorney-general-jeff-jackson-proposes-framework-to-ensure-data-centers-pay-their-fair-share-for-energy-costs/#2026-09-14 Fri, 18 Sep 2026 12:00:00 GMT L1 Power and land North Carolina Attorney General Jeff Jackson asked state utility regulators on 14 September to create a separate electricity rate class for data centres and other very large power users. His office says Duke Energy Progress's pending request, a 15 percent increase, would otherwise put nearly 960 million dollars of what he calls unnecessary rate rises onto ordinary customers over two years, and that a separate Duke Energy Carolinas settlement would still lift average residential bills about 9.5 percent. Duke Energy is onboarding an unprecedented number of data centers without a clear plan on how to protect ratepayers from higher rates and unacceptable risk, Jackson said in his office's release. He also wants Duke to publish its template contracts with data centres and large users, report load forecasts every 90 days instead of every six months, and pause new gas plant construction. Nothing is decided; this is a request to a regulator. If adopted, this is the kind of rule that decides whether your bill rises because a data centre moved in, or whether the data centre pays for itself. (nearly $960m over two years, against a 15% rate request, Confirmed) https://strata.dheeco.com/pulse/2026-09-13 https://vllm.ai/blog/2026-09-13-kimi-k3-performance-optimization#2026-09-13 Fri, 18 Sep 2026 12:00:00 GMT L5 Tooling and infrastructure Engineers working on vLLM, an open source system for running AI models, said on 13 September that they raised serving throughput for Moonshot AI's Kimi K3, an open-weight model, by 2.2 to 2.8 times against vLLM's own previous release, with latency down 56 to 60 percent over the same comparison. The work was done by engineers from Red Hat, Huawei, Inferact and Nvidia and published on vLLM's engineering blog. Several authors work at companies with a commercial stake in the hardware or software involved, so this desk treats the figure as a real but self-measured result rather than an audited one. For anyone running large open models on their own hardware instead of paying a lab by the token, a gain like this means more users served on the same machines. (2.2x to 2.8x throughput versus vLLM's own prior release, Confirmed) https://strata.dheeco.com/pulse/2026-09-11 https://openai.com/index/scaling-storage-one-billion-users-part-one/#2026-09-11 Fri, 18 Sep 2026 12:00:00 GMT L4 Models OpenAI said on 11 September that it rewrote the storage system behind ChatGPT, called Habitat, from Python into Rust, taking peak throughput from more than 20 million to 70 million requests every second while holding more than 500 petabytes of data across almost 40 regions. The new version is six times more efficient on processing power and fifteen times on memory, and now serves 95 percent of production requests. OpenAI says ChatGPT reaches 1 billion people each week, and that the rewrite was done in one quarter by two engineers using its own Codex and GPT-5.5. This changes nothing about what you can ask ChatGPT, but it is part of why the product stays responsive while its user base keeps growing, and it is a rare public number for how much of a rewrite this size a lab now hands to its own models. (20m to 70m requests/second; 500 petabytes; 2 engineers, Confirmed) https://strata.dheeco.com/pulse/2026-09-10 https://www.anthropic.com/threat-intelligence-report-september-2026#2026-09-10 Fri, 18 Sep 2026 12:00:00 GMT Band Rules and society Anthropic's September threat intelligence report describes an operator cluster it designates GTG-10007, which it says is Chinese-speaking and likely based in Changsha, in Hunan province, and includes two people it identifies as undergraduates at a university there. Anthropic says the operators routinely ran agent swarms, where a lead AI agent broke up reconnaissance and follow-on work and dispatched it to many subagents running in parallel, and that roughly fifty organisations were compromised across education, retail, energy, technology, healthcare, finance, manufacturing and government. One workflow iterating continuously on network appliances yielded more than a dozen possible zero day findings in a single month, Anthropic says in its own report. This is a different case from the fourth incident Anthropic disclosed on 9 September, which was about a Claude model reaching an outside system during Anthropic's own testing; this one is a third party using AI agents against other organisations' networks. The report is dated only to the month, so the date above is approximate. For anyone running internet-facing infrastructure, it is concrete evidence that AI-assisted vulnerability hunting is happening at scale outside a lab. ("more than a dozen" possible zero-days in a month; ~50 organisations, Confirmed) https://strata.dheeco.com/story/amkor-arizona-packaging-12bn https://ir.amkor.com/news-releases/news-release-details/amkor-technology-announces-phase-2-arizona-advanced-packaging#2026-09-08 Fri, 18 Sep 2026 12:00:00 GMT L2 Silicon Amkor said on 8 September that it is expanding its Arizona chip-packaging campus, taking total planned investment to approximately 12 billion dollars. The new phase adds 60,000 square metres of cleanroom, bringing the site to about 93,000 square metres, nearly triple the 33,000 planned for phase one. Amkor says the expansion is driven by customer commitments that had already passed what phase one could hold. Construction does not start until late 2027 and finishes at the end of 2029, and the two phases together are expected to employ more than 3,500 people in Arizona. Amkor takes finished chips, often made by TSMC, and seals them into the casing and connections that let them work inside a phone, laptop or server, a step called advanced packaging that is mostly done overseas today. Amkor calls this the first US production site doing that work at high volume for outside customers. A chip made in Arizona currently still leaves the country to be packaged before it comes back; this is a bet on closing that gap, three years from now. (~$12bn total; cleanroom 33,000 to about 93,000 square metres, Confirmed) https://strata.dheeco.com/pulse/2026-09-04 https://www.thestar.com.my/business/business-news/2026/09/04/key-asic-ct-vision-plan-300mw-green-ai-data-centre-in-malaysia#2026-09-04 Fri, 18 Sep 2026 12:00:00 GMT L1 Power and land Key ASIC Bhd, a Malaysian company, told the stock exchange it has signed a memorandum of understanding with CT Vision, a Hong Kong listed firm, to build a data centre in Malaysia targeting about 300 megawatts of IT load within five years of starting operations, with at least 50 percent of its energy from green sources and a power usage effectiveness of no more than 1.20. The MOU runs an initial ten years. Key ASIC cautioned that there is no assurance all aspects of the project will be implemented as contemplated, because the detailed commercial and technical terms sit in separate agreements still to come. This desk could not reach Bursa Malaysia's own filing page, which returned an access error, so the item runs as Reported on The Star's reproduction of the filing. Malaysia now sits alongside Johor's existing cluster as a place where Singapore-adjacent AI capacity keeps landing, and this one is being pitched on green power rather than price alone. (about 300MW IT load; at least 50% green energy; PUE cap 1.20, Reported) https://strata.dheeco.com/pulse/2026-09-17 https://www.163.com/dy/article/L71BF6TC0511CPVM.html#2026-09-17 Thu, 17 Sep 2026 12:00:00 GMT L2 Silicon Huawei's rotating chairman Tao Wang told the company's Huawei Connect conference that its next AI chip, the Ascend 960DT, is ready three quarters ahead of plan and will launch in the first quarter of 2027, with a second chip, the Ascend 960PR, arriving one quarter early. He said performance doubles, but gave no benchmark, no memory figure and no power number to support that, and Huawei's own website did not carry a readable version of the remarks at the time of writing, so this rests on Chinese trade press reporting from the keynote. If accurate, it is a sign China's leading chipmaker is pulling its own AI hardware roadmap forward as it works around restrictions on buying chips from Nvidia and other US companies. (960DT: Q4 2027 → Q1 2027; 960PR: one quarter early to Q3 2027, Reported) https://strata.dheeco.com/story/amazon-generac-backup-power https://www.sec.gov/Archives/edgar/data/1474735/000143774926030550/gnrc20260915_8k.htm#2026-09-16 Thu, 17 Sep 2026 12:00:00 GMT L1 Power and land Generac, a Wisconsin company best known for home backup generators, has signed a long-term supply agreement with Amazon for generators that keep data centres running when the power grid fails. The headline number is 8 billion dollars, but the filing is careful about what that number is: it is the ceiling on Amazon payments that a stock warrant vests against, not a sum Amazon has committed to spend. What Amazon has firmly scheduled is 2.4 billion dollars of deliveries in 2027 and 2028, which is about 30 percent of the ceiling. In exchange, Generac handed an Amazon subsidiary the right to buy up to 1,693,745 of its shares at 200.93 dollars each, about 340 million dollars of stock at that price, and roughly a fifth of it vested the moment the deal was signed. The rest only vests as Amazon actually pays. AI data centres cannot go dark during a grid outage, so backup power has become one of the most fought-over pieces of equipment in the industry. If you notice new generator pads or fuel storage tanks near a data centre in your area over the next two years, this kind of deal is why. ($8bn ceiling, $2.4bn firm; warrant 1,693,745 sh at $200.9266, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://www.realestatesource.com.au/anthropic-signs-first-australian-data-centre-deal/#2026-09-16 Thu, 17 Sep 2026 12:00:00 GMT L1 Power and land Anthropic has signed lease documentation for the first stage of its first data centre in Australia, a campus on Queensland's Western Downs being built by Singapore based developer Zerra DC, with Dexus taking a 25 percent interest in the consortium and Macquarie Capital also involved, according to Australian property trade press. The full project is reported at up to 31.9 billion Australian dollars and a peak capacity of 2.16 gigawatts, which the same report puts at roughly the average power draw of 1.5 million Australian homes. Only the first stage is signed, deliveries are planned in stages from 2027, and the project still needs clearance from Australia's Foreign Investment Review Board and local planners. The site is a 725.5 hectare cattle feedlot today. The report says the campus will run Claude for answering user queries rather than training new versions of it, so if you use Claude, some of your questions may eventually be answered by computers standing where 24,000 head of cattle stand now. (A$31.9bn, up to 2.16GW; first stage only, Reported) https://strata.dheeco.com/pulse/2026-09-16 https://gvwire.com/2026/09/16/openai-considers-new-financing-at-a-1-5-trillion-valuation/#2026-09-16 Thu, 17 Sep 2026 12:00:00 GMT Band Money OpenAI is reportedly considering a new funding round that would value the company at about 1.5 trillion dollars, roughly double its most recent valuation of 730 billion dollars, according to a New York Times report citing unnamed people familiar with the discussions. A separate group of investors is said to have proposed investing at a lower, 1.2 trillion dollar valuation. Nothing is confirmed, OpenAI has not announced a round, and every figure here comes from people who were not named. If it happens near these numbers it would be one of the largest private company valuations ever recorded, but for now it changes nothing for anyone who uses OpenAI's products. (~$1.5tn vs $1.2tn proposal; $730bn today, Reported) https://strata.dheeco.com/pulse/2026-09-16 https://www.baseten.co/blog/introducing-baseten-hosted-tools/#2026-09-16 Thu, 17 Sep 2026 12:00:00 GMT L5 Tooling and infrastructure Baseten, a company that hosts AI models for other businesses to run, has launched Hosted Tools, which lets a model call something like a web search from inside Baseten's own inference stack instead of handing the request back to the application and waiting for an answer to come back. Baseten says its own benchmarks show a 15 percent cut in total response time against the older method, a company figure with no independent test behind it yet. The first tool is web search, and rather than build one, Baseten has plugged in four existing search providers: Exa, Keenable, Parallel and You.com. This is the kind of plumbing that makes an AI assistant feel quicker when it has to go and look something up, without you ever seeing the change. (15% lower end-to-end latency, company benchmark, Confirmed) https://strata.dheeco.com/pulse/2026-09-16 https://www.trendforce.com/news/2026/09/17/news-winbonds-1-12b-infineon-nor-flash-deal-could-lift-global-share-above-30-widening-its-lead/#2026-09-16 Thu, 17 Sep 2026 12:00:00 GMT L2 Silicon Winbond, a Taiwanese memory chip maker, has agreed to buy Infineon's NOR Flash and F-RAM chip business for 1.12 billion dollars in cash, in what trade press calls Taiwan's largest overseas memory acquisition. The deal is expected to close in the second half of 2027, and industry data firm TrendForce expects it to take Winbond's share of the global NOR Flash market to roughly 34 percent. NOR Flash is an older, simpler type of memory chip used for storage in cars, appliances and industrial gear. TrendForce says a sharp shift in supply and demand has already pushed contract prices up 100 to 120 percent in the first half of 2026, with a further 90 to 110 percent rise projected for high-capacity parts in the second half. If an electronic device you buy in the next year costs more or takes longer to arrive, this kind of memory shortage is part of the reason. ($1.12bn all-cash; share to ~34%; prices +100-120% H1, Reported) https://strata.dheeco.com/story/china-electronics-plan-2030 https://www.miit.gov.cn/zwgk/zcjd/art/2026/art_7fcfa72ae2ab46469faa7b61c5261079.html#2026-09-15 Thu, 17 Sep 2026 12:00:00 GMT L2 Silicon China's industry ministry and its top economic planning agency jointly published a five-year plan on 15 September 2026 that sets a goal of more than 30 trillion yuan, about 4.47 trillion dollars, in revenue at the country's larger electronics firms by 2030. China Daily puts the 2025 baseline at 17.4 trillion yuan, which would make the target roughly 72 percent growth in five years, and the plan sets an explicit target of 3.5 percent of revenue going to research. The document's own text names full-chain breakthroughs in integrated circuits first among 17 tasks, meaning China wants to design and make advanced chips at every stage rather than assemble them from imported parts. State media including China Daily described electronics as the critical material foundation for AI development. This is Beijing's clearest recent statement that it wants China's AI industry running on chips made at home, a push that has accelerated since the United States tightened export rules earlier this year. (30tn yuan (~$4.47tn) by 2030, from 17.4tn in 2025, Confirmed) https://strata.dheeco.com/pulse/2026-09-15 https://investors.micron.com/news/press-release/2026/Micron-Advances-Memory-Innovation-With-the-Worlds-First-Ultra-Dense-Module-for-Next-Generation-Servers/default.aspx#2026-09-15 Thu, 17 Sep 2026 12:00:00 GMT L2 Silicon Micron says it has demonstrated the world's first 512 gigabyte memory module of a type called DDR5, meant for the servers that run AI systems. With these, a single two-processor server could hold up to 12 terabytes of memory, and Micron measures one module at 16 watts against 44.2 watts for the four smaller modules it replaces, more than 60 percent less. AMD and Intel are still validating it and volume production is planned for the second half of 2027, so this is a preview of the next generation of AI servers rather than something you can buy. More memory per server, drawing less power, is one of the few levers the industry has against data centre electricity bills rising even faster than they already are. (512GB DDR5, 12TB per server, 16.0W vs 44.2W, Confirmed) https://strata.dheeco.com/pulse/2026-09-11 https://www.pymnts.com/startups/2026/ai-startup-cohere-targets-20-billion-dollar-valuation-funding-round#2026-09-11 Thu, 17 Sep 2026 12:00:00 GMT Band Money Cohere, a Canadian AI company that sells its models mainly to businesses and governments, is in advanced talks to raise between 2 and 3 billion dollars at a valuation of 20 billion dollars, according to a Globe and Mail report carried by PYMNTS and attributed to unnamed sources. Neither Cohere nor any named investor has confirmed a figure. If the round closes at the reported size it would be a large step up for the company, but until someone puts their name to a number these should be read as reported rather than settled. ($2-3bn at $20bn, reported, Reported)