Chip designer
Designs the processors most AI runs on. Now describes its own products in dollars per gigawatt, which tells you what the scarce thing is.
$89.0B data centre revenue in one quarter, up 117%
· 5 min
The accord is one page of text, voluntary, and asks each company for four layers of checks, one of them an outside auditor. It came on the same day Trump ordered federal agencies to call AI 'Super Intelligence'.
· 5 min
MLCommons published MLPerf Inference v6.1 with a record 30 submitting organisations and, for the first time, a test of AI agents doing multi-step work such as coding.
· 5 min
Eight companies, organised by Nvidia, say they will build enough AI computing capacity in Australia to draw as much power as a large power station many times over.
· 4 min
A market research firm's count of DRAM sales shows the price pressure already visible in Nvidia's margins and Amazon's capex is larger than those numbers alone suggested.
· L2 Silicon · $4,999 · Confirmed
Nvidia is launching a new, lower-memory version of its DGX Spark desktop computer, a small machine meant for running AI models at a desk rather than in a data center. The 64GB version starts at $4,999 and goes on sale October 23 through PC makers including Dell, HP, and Asus. Two of these machines can be linked together to match the memory of the original, more expensive model. This gives developers and small businesses a cheaper way to experiment with AI models at home instead of paying for cloud computing time.
· Band Money · +$150bn to $235bn · Confirmed
Nvidia's board approved spending another $150 billion buying back its own shares, on top of what was left of an earlier plan, bringing the total it is now authorised to spend to $235 billion. Buying back shares doesn't change what Nvidia makes or sells; it signals the company has more cash than it currently has other uses for, and it has no direct effect on the price of Nvidia's chips.
· Band Rules and society · 6 companies; 4 layers of checks, one an outside auditor · Confirmed
President Trump posted the signed text of the White House Accord on Super Intelligence on 29 September. It is signed by him and by Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Greg Brockman of OpenAI, Elon Musk of xAI and Jensen Huang of Nvidia. It says every company training and deploying frontier models, the most advanced AI systems, should run four layers of checks: internal controls that watch for risks such as cyber, biological and chemical threats and make sure models do not hack or access systems in unintended ways; an internal team to make sure those controls work; an independent outside auditor or evaluator; and an independent board committee that receives the reports and makes sure problems are fixed. The companies say they will meet regularly to set standards. The text says it may make sense to write these steps into laws or regulations over time. It does not say who chooses the auditors.
Read the story: Six AI leaders sign a White House pledge to audit their own models. Here is what it says, and what it leaves out · OpenAI · Anthropic · Nvidia
· L3 Iron and cloud · $240.9m; 2,000+ B300 GPUs; two clusters; take or pay · Confirmed
QumulusAI, which rents out computing for running AI models, said on 29 September that it signed agreements with a combined contracted value of $240.9 million covering more than 2,000 Nvidia Blackwell B300 processors across two dedicated clusters. Both are three-year take-or-pay agreements, meaning the customer pays for the capacity whether or not it uses it. It says the first customer's commitments with it now total more than $283 million and the second's more than $61 million. For the wider market, it is another multi-year contract that pays for new AI computers before they are switched on.
· L5 Tooling and infrastructure · OpenShell software; Sentry on BlueField-4 chips · Confirmed
Nvidia announced its Open Agent Safety Platform on 28 September, open software and a reference hardware design for keeping AI agents inside the limits their operators set. One part, Sentry, runs on Nvidia's BlueField-4 data processors and watches agents from outside the software they run in; Nvidia says it can quarantine an agent that tries to move outside its boundary within milliseconds. Nvidia names Anthropic, Cisco, CrowdStrike, Microsoft, Hugging Face and SpaceXAI among those working with it, and says SpaceXAI uses it for Cursor coding agents and Grok models. For a company letting AI agents act on its systems, this is a tool meant to put a hard stop on an agent that goes past what it was allowed to do.
· L3 Iron and cloud · $3.36bn: $2.36bn at closing, $1bn from Nvidia in mid-November 2026 · Confirmed
Nscale, a British company that builds and rents out AI computing, says it has raised $3.36 billion through convertible loan notes, a loan that turns into shares later, led by Third Point. $2.36 billion came in at closing, and Nvidia has committed a further $1 billion, expected in mid-November 2026. The notes turn into shares automatically when Nscale completes its initial public offering, with Nvidia getting shares that carry no vote. Other investors include funds managed by Apollo, Citadel, Hudson Bay Capital, the Abu Dhabi Investment Council and 8090 Industries. Nscale's listing filing, covered here on 18 September, first described these notes.
· L3 Iron and cloud · 22,000 GPUs in early 2027; 48,000+ in total; 100+ MW · Confirmed
Aolani, a Singapore-founded AI cloud company, said on 21 September it had signed a compute collaboration with Nvidia. Under it, 22,000 Nvidia Blackwell Ultra GPUs are expected to be deployed across two AI data centres, which it calls AI factories, in Malaysia and the Philippines in early 2027, taking Aolani past 48,000 GPUs and more than 100 megawatts of capacity. It says a revenue-sharing and credit-support model will tie the build-out to customer demand, and it gives no dollar figure. It expects the Philippine site to be among the first large-scale AI computing platforms in the country.
· L3 Iron and cloud · 30 orgs (record); DeepSeek R1 best per chip 5.7x in a year · Confirmed
MLCommons published results for MLPerf Inference v6.1 on 16 September, its industry-standard, peer-reviewed benchmark of how fast AI systems run, with a record 30 organisations submitting, including Nvidia, Microsoft Azure, Oracle, CoreWeave, Nebius and Lambda. For the first time it includes a test of AI agents: the Edge Agentic Inference test, which measures multi-turn work such as agentic coding, where each step depends on the ones before. MLCommons says the best per-chip result on its visual-language test improved 2.99 times in six months, and on the DeepSeek R1 reasoning test 5.7 times in a year. For people using AI tools, MLCommons says gains like these could let the same systems serve more users.
Read the story: MLPerf, the AI industry's standard speed test, adds its first test of AI agents as a record 30 organisations take part · Nvidia · Lambda · Oracle · Microsoft Azure · CoreWeave · Nebius
· L1 Power and land · three founding members; no partner count or capacity target given · Confirmed
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.
· L5 Tooling and infrastructure · 22+ petaflops per GPU · Reported
Together AI ported its open source ThunderKittens programming tools to Nvidia's newest chip, called Vera Rubin NVL72, and says it reached over 22 petaflops of real-world performance per GPU, a measure of raw calculation speed, competitive with Nvidia's own closed-source software for the same job. A petaflop is a thousand trillion calculations per second; Nvidia's previous generation chip, the B200, tops out at 9 petaflops per GPU on this kind of workload on paper, so Vera Rubin represents a large jump in raw hardware capability. This is Together AI's own benchmark of its own open source code, not independently verified, and it is a technical achievement inside the AI infrastructure world rather than a product change that affects prices today. For anyone using an AI product, tools like this are part of what eventually lets AI companies use new chips at close to their full speed once they are available.
· L3 Iron and cloud · up to $3.1bn, 1 GW target · Confirmed
Zankore, an Indonesian company, signed a senior term loan facility of up to 3.1 billion US dollars to build data centres using Nvidia hardware. It is building an initial 100 megawatts, says it expects approximately 200 megawatts of initial capacity in the first half of 2027, and is targeting 1 gigawatt eventually. The release does not make clear whether the 200 megawatts includes the first 100 or sits on top of it. This is a large financing in Southeast Asia, a region that gets far less AI infrastructure coverage than the United States or China. Nothing changes for ordinary people yet; this is financing and construction, not a product reaching anyone.
· L3 Iron and cloud · up to 2 GW by 2027 · Confirmed
Nvidia, which designs the chips most AI computing runs on, announced that eight data centre and cloud companies in Australia plan to build up to 2 gigawatts of AI computing capacity between them by 2027. A gigawatt is roughly the output of a large power station; 2 gigawatts would be a very large buildout for one country. In the announcement, the cloud company Sharon AI said it is deploying up to 68,000 Nvidia GPUs, the data centre operator CDC said it already runs more than 550 megawatts of capacity in Australia and New Zealand with 800 more under construction, and IREN pointed to its 800 megawatt Bundey campus in South Australia. This is Australia's first appearance in Strata's graph of tracked AI stack events. For an ordinary person in Australia, nothing changes today, but a buildout at this scale is the kind of thing that shows up later as electricity demand and grid investment in your area.
Read the story: Australia joins the AI buildout race with a coordinated two-gigawatt pledge · Nvidia
· L2 Silicon · $154.73bn industry revenue, +59.5% QoQ · Confirmed
TrendForce, a technology market research firm, says the DRAM memory chip industry took in nearly $154.73 billion in the second quarter of 2026, up 59.5 percent from the first quarter, as contract prices for memory chips rose sharply while shipment volumes grew only modestly. Samsung led with $60.98 billion of that revenue, up 63.4 percent, followed by SK Hynix at $38.59 billion, up 37.9 percent, and Micron at $36.0 billion, up 65.5 percent. DRAM is the memory that computers and AI servers use to hold data they are actively working on. For anyone buying a computer, a phone or anything else with memory chips inside it, these are the price rises TrendForce describes.
Read the story: The world's three big memory chip makers post their strongest quarter in years as prices surge · Micron · Nvidia
· Band Money · approx $11.9bn · Confirmed
Nvidia's filing with the US securities regulator puts the price for Hugging Face at approximately 11.9 billion dollars payable to its stockholders. Up to approximately 1 billion dollars more is set aside in shares for Hugging Face employees who join Nvidia.
· L5 Tooling and infrastructure · $12.93B · Confirmed
Nvidia makes the chips nearly all AI runs on. Hugging Face is the site where researchers publish and share their AI models freely. The company that makes the engines has bought the garage everyone parks in.
· L2 Silicon · 71 to 72% · Confirmed
Nvidia told investors its profit margin will bottom at 71 to 72 percent early next year. The reason it gave is memory. The chips that feed its processors have become expensive enough that the most profitable company in the industry is absorbing the cost. The figure was said on the earnings call, not written in the filing.
· Band Rules and society · Guidance, not a new rule · Confirmed
America told exporters that its existing licence rules for advanced AI chips already apply to companies headquartered in China and a short list of other countries, including their offices elsewhere in the world. No new law was written. It is a statement about how the rule already on the books will be enforced.
Strata, the whole AI stack, explained simply. Every number carries a source and a confidence label.
Today · The Stack · Stories · Archive · Glossary · Break the Chain · Focus · About · RSS feed
Privacy · Terms · Corrections · Report an error
Also from Dheeco: Dheeco · AP Fact Check
Published by dheeco.com · © 2026 Dheeco