Corporate Concentration
A small number of labs and chip companies holding the systems everyone else depends on.
October 21, 2026 · Technology & Capability
Nvidia ships AI data centres in shipping containers
Building an AI data centre is getting more like assembling furniture than pouring concrete. At GTC Berlin on October 21, 2026, Nvidia introduced DSX POD, a reference design for complete Vera Rubin AI factories that partners including Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv build and test inside modular containers before shipping, cutting setup at a prepared site from about six months to as little as 14 days. Nvidia also said its DSX platform can produce the same AI output using up to 30% less power through its MaxLPS software, and that closed loop liquid cooling with dry coolers can cut facility cooling water use to near zero, which it estimates at 54 million litres a day saved at a 2 gigawatt deployment.
October 20, 2026 · Technology & Capability
Chinese tech giants order $35 billion of Nvidia cards
Chinese firms locked out of Nvidia's best data centre chips are buying workstation cards by the million instead, and Nvidia says it cannot make nearly enough of them. Reuters reported on October 20, 2026 that ByteDance, Alibaba and Tencent have ordered more than 3 million RTX PRO 5500 cards worth over $35 billion, and that Nvidia has told customers it can fill only a fraction because demand far exceeds planned output. From 2027 Nvidia is expected to shift much of its TSMC capacity to Rubin, its next data centre platform, with RTX PRO 5500 samples due in December and deliveries starting in the first quarter of 2027.
October 19, 2026 · Technology & Capability
Microsoft walks away from UK data centre over power
A flagship British AI data centre lost its anchor tenant because the electricity grid cannot reach it until the 2030s. Microsoft terminated its agreement for Nscale's site in Loughton, Essex, which was meant to start running early next year with 23,040 Nvidia GB300 GPUs and 50 megawatts of capacity for Azure in Britain. A Microsoft spokesperson told Reuters the decision relates to a single site, and Nscale said it is looking at alternative power sources and other UK locations.
October 14, 2026 · Power & Politics
Amazon eyes $7 billion deal to loosen Nvidia's grip
The reason almost everyone building AI buys Nvidia chips is that the software only really works on them, and Amazon is reportedly ready to spend about $7 billion to change that. According to The Wall Street Journal, Amazon is in talks to buy Decart, an Israeli startup whose software lets AI models run efficiently on different brands of chip, which would make it easier for customers to switch to Amazon's own Trainium processors. CEO Andy Jassy says the chip unit would be worth roughly $50 billion a year on its own, and Amazon has already booked over $225 billion in Trainium commitments.
October 13, 2026 · Technology & Capability
Banks get AI that keeps customer data in house
Your bank can now run AI over your account records without shipping them off to an outside tech company. French banking software firm SBS launched the SBS AI Foundation, built with NVIDIA chips and software, which lets a bank use generative AI inside its own systems so sensitive data never leaves. Scottish Building Society is the first named user, and SBS chief executive Eric Bierry says banks want AI they can explain and run wherever their data has to live.
October 12, 2026 · Technology & Capability
Microsoft says its own chip will run most AI work
The company that buys more Nvidia hardware than almost anyone says it now plans to run most of its own AI on silicon it designed itself. Microsoft announced the Azure Maia 300 on October 12, 2026, calling it the primary inference system for Copilot and its other AI products, with deployment at scale starting in 2027 and potentially more than a million units. Microsoft says the chip does more AI work per unit of electricity than the GPU systems that handle most inference today, and that the earlier Maia 200 is already running in multiple regions.
October 5, 2026 · Technology & Capability
Electricity, not chips, is now the AI bottleneck
The thing holding back AI buildouts has shifted from silicon to power sockets. Morgan Stanley told clients on October 5, 2026 that US data centre developers face a 34% net power shortfall through 2028, equal to about 32 gigawatts, even after accounting for behind-the-meter generation and fuel cells. The bank said Nvidia and Broadcom should ride it out because of their visibility into where chips get placed, while memory, optical and other secondary component makers are more exposed if deployments slip.
September 28, 2026 · Cultural & Ethical
Pope Leo says AI fears are not a hoax
Speaking to reporters on the flight back to Rome from France, Pope Leo XIV said concerns that AI could destroy the world are not "fake news," an apparent response to President Donald Trump telling the UN General Assembly the week before that AI danger fears were a "hoax." Leo also questioned tech executives who oppose guardrails, citing Nvidia CEO Jensen Huang, who had told CBS News he agreed with Trump that AI safety concerns were overblown. The remarks came a day before a White House AI meeting with Trump, Speaker Mike Johnson, Huang, Mark Zuckerberg, Dario Amodei and Greg Brockman.
September 28, 2026 · Technology & Capability
Kill switches for AI arrive from two directions: Nvidia, and NYC
The idea that an AI system should have an off switch stopped being a talking point and started showing up in both law and hardware within the same week. New York City Council Speaker Julie Menin introduced a package of bills led by Introduction 2602, which would make it unlawful to sell or deploy an AI system in the city without outside validation of its data, bias, privacy and security, and would require every covered system to carry a human override that can shut it down. Penalties reach $25,000 per instance against the business and the validator, a companion bill would pay whistleblowers a cut of recovered fines, and Council Member Virginia Maloney's Introduction 2600 would let people sue over foreseeable harm. Three days later Nvidia released its Open Agent Safety Platform, combining open source software called OpenShell, which fences in what an AI agent can do and logs its actions, with Sentry, a watchdog running on Nvidia's BlueField-4 networking chips that can cut off a misbehaving agent in milliseconds. Jensen Huang called it full-stack safety engineering, and more than 100 organizations signed on including Anthropic, Microsoft, IBM, Palantir, Salesforce, JPMorganChase and Hugging Face, while the New York bills remain proposals.
September 22, 2026 · Technology & Capability
Alibaba unveils its own AI chip and 20 gigawatt plan
Cut off from Nvidia's best hardware, China's largest cloud company says it has built a replacement good enough to train frontier models. Alibaba CEO Eddie Wu introduced the Zhenwu V900 at the Apsara conference in Hangzhou on September 22, 2026, a chip from its T-Head design unit with 216GB of memory and 1,200 GB/s of inter-chip bandwidth that Alibaba says triples the performance of May's M890 and can scale into clusters of 500,000 units. Alibaba also said it aims to operate more than 20 gigawatts of global data centre capacity by 2032, with mass production of the V900 targeted for the first quarter of 2027.
September 11, 2026 · Technology & Capability
Three AI datacenter buildouts land in the same week
Within three days, three of the biggest names in AI committed to power draws that rewrite national energy plans. Google put at least 13 billion euros (15.1 billion dollars) into new data centres in the Finnish towns of Kajaani, Muhos and Vaala plus an expansion at Hamina, its largest single European investment, with grid, clean energy and battery work attached and a promise of 37,000 supported jobs. The same day, Nvidia lined up eight Australian operators including AirTrunk, NEXTDC, CDC and IREN to add up to 2 gigawatts of capacity for its DSX AI factory platform by 2027, more than doubling a national footprint that currently sits near 1.6 gigawatts. Two days later came word that Microsoft is targeting more than 38 gigawatts of data centre capacity by 2032, up from roughly 12 today, enough electricity to exceed New York State at peak, driven by having to turn customers away for lack of machines. The local benefits are real, from Finnish construction jobs to Australian researchers finally getting compute at home, and the plans are not fixed, since Microsoft dropped about 2 gigawatts of projects earlier this year.
September 8, 2026 · Power & Politics
Mistral raises 3 billion euros, becoming Europe's biggest AI bet
Europe now has one AI company big enough to argue it does not need American models, and investors just put three billion euros behind that claim. Paris-based Mistral AI, founded in 2023 by Arthur Mensch, Guillaume Lample and Timothée Lacroix, closed the round at a valuation above 21 billion euros, roughly double where it stood a year earlier. Samsung Electronics led, joined by EQT's Scaleup Europe Fund, PSG Equity, Advent, BlackRock-managed funds, Andreessen Horowitz, ASML, Nvidia, Salesforce Ventures and the Grand Duchy of Luxembourg. The money is going into computing capacity on European soil, 200 megawatts by the end of 2027 and a target of one gigawatt by 2030, which is what governments and large firms asking for AI that runs under their own rules actually need. Emmanuel Macron described it as France and South Korea building a third way in AI, though Mistral's backers include Nvidia and Samsung and it deepened its Microsoft partnership in July.
September 3, 2026 · Technology & Capability
Nvidia buys Hugging Face for 12.9 Billion, amongst other acquisitions
The place where most freely downloadable AI models live is now owned by the company that makes the chips they run on. Nvidia had already spent two years funding more than 100 AI startups, including OpenAI, Anthropic, a reported $1 billion deal with Poolside and a considered $2.5 billion into Mira Murati's Thinking Machines Lab, a company that had separately committed to using at least a gigawatt of Nvidia's Vera Rubin systems. In September 2026 it confirmed the largest of those moves, buying Hugging Face for $12.93 billion, with $11.9 billion to shareholders and $1 billion in equity to keep staff on. Jensen Huang said co-founder Clement Delangue came to him, and Nvidia says the platform, home to over 3 million models and more than 18 million developers, will stay open to competing frameworks and cloud providers. Hugging Face gains funding and infrastructure that open source AI has wanted for years, under an owner with an obvious interest in what runs where.
August 26, 2026 · Technology & Capability
Amazon orders two million more Nvidia chips
Amazon burned through a one million GPU commitment in under five months and has ordered twice as many again. AWS and Nvidia said on August 26, 2026 that they will deploy 2 million additional Blackwell Ultra, Rubin and Rubin Ultra GPUs across AWS infrastructure in 2027 and 2028, bringing planned Nvidia capacity at AWS past 3 million chips. The deal also brings Nvidia Vera CPUs to AWS and includes AI factories for the US government with 100,000 GPUs on secure infrastructure, with AWS CEO Matt Garman and Nvidia CEO Jensen Huang both saying demand is running ahead of forecasts.
August 24, 2026 · Technology & Capability
Nvidia puts its $20 billion Groq buy into production
The chip startup that was supposed to be Nvidia's fastest rival is now a product inside Nvidia's own racks. Nvidia announced at Hot Chips 2026 on August 24 that Groq 3 LPX, an inference accelerator built from the assets of its $20 billion Groq purchase in December, has entered full production, with 256 chips per rack and racks going online this year. In Artificial Analysis testing it produced 3,400 output tokens per second running the open model Gemma 4 31B at a 100,000 token context, though Nvidia's figure came from a private pre-release endpoint while competing numbers came from live services. Nebius is the first cloud provider to commit to it.
August 22, 2026 · Technology & Capability
Toronto startup raises $90M to build robot training worlds
Instead of teaching robots by running them in warehouses and factories, a new company wants to raise them inside a video-game-like simulation where they can fail thousands of times for free. Veeda AI, a Toronto startup founded by former Nvidia researchers Sanja Fidler (CEO), Zan Gojcic and Huan Ling, has raised more than $90 million in early funding from Radical Ventures and Khosla Ventures, according to a report by The Logic citing filings and people familiar with the deal. Fidler says the real world is too slow, costly and dangerous for robots to learn in, and pitches simulated reality as the "critical infrastructure layer" for physical robotics.
August 14, 2026 · Power & Politics
Nvidia turns out to own $21 billion of SpaceX
The company that makes the chips behind the AI boom is also quietly one of the biggest owners of the companies buying those chips. A required quarterly disclosure to the US Securities and Exchange Commission showed Nvidia holding about $21 billion of Elon Musk's SpaceX, which went public in June, plus roughly $30 billion of Intel. Those two stakes were over 80 percent of Nvidia's $63 billion stock portfolio, and the SpaceX shares came from Nvidia's $10 billion investment in Musk's AI company xAI, which SpaceX then absorbed in February.
August 13, 2026 · Power & Politics
Anthropic spending the most it ever has for a single purchase, to buy Israeli chip efficiency startup Decart for $6-7 Billion
The company that makes Claude is about to spend more on one acquisition than it ever has, and the target is not a model lab but a firm that squeezes more work out of existing chips. Anthropic is in talks to buy Israeli startup Decart, founded in 2023 by Dean and Orian Leitersdorf and Moshe Shalev, for roughly $6 billion to $7 billion, with Decart's team expected to join Anthropic's inference and performance organization. Bloomberg first reported the talks on August 13, 2026, Calcalist reported days later that the parties were exchanging advanced drafts with most of the payment in Anthropic shares, and both companies declined to comment; Nvidia had previously been in talks for the same company.
August 6, 2026 · Technology & Capability
AI writes working virus genomes from scratch
A computer program has designed living viruses from scratch, and 16 of them actually worked in the lab. Researchers led by Brian Hie of Stanford University and the Arc Institute, with PhD student Samuel King and collaborators at NVIDIA and UC Berkeley, used AI models called Evo 1 and Evo 2, trained to predict DNA letters the way chatbots predict words, to write full genetic blueprints for bacteriophages, viruses that infect bacteria and cannot infect people. About 300 designs were built and tested, 16 came alive, and some killed E. coli better than the natural virus they were modeled on.
July 27, 2026 · Power & Politics
Nvidia bets $5 billion on Sutskever's secret lab
Secretive AI lab Safe Superintelligence (SSI), co-founded by Ilya Sutskever (formerly OpenAI's top scientist) and Daniel Levy, just got a massive injection of computing power and cash, which matters because it is now one of the best-funded AI efforts on the planet despite having released nothing publicly yet. SSI announced a long-term partnership with chipmaker NVIDIA that includes a reported $5 billion investment and priority access to NVIDIA's next wave of AI processors. The company has now raised $7 billion in total at a $32 billion valuation, backed by major investors including Andreessen Horowitz, Alphabet, and Sequoia Capital.
July 26, 2026 · Power & Politics
Nvidia and SoftBank plan Ohio's biggest data center
The largest data center ever built could be coming to Ohio, and the money and power behind it stretches from Tokyo to Washington. Nvidia is in talks to help finance up to $250 billion so OpenAI can lease the facility, which is being developed by SoftBank's energy arm, SB Energy, about 50 miles south of Columbus; combined with separate chip financing Nvidia is discussing with OpenAI, the total could top $500 billion. The electricity supply for the site is controlled by the U.S. government, with Commerce Secretary Howard Lutnick deciding who gets access, and Japan is partly funding the project through a trade deal linked to a $33 billion natural gas investment.
July 24, 2026 · Technology & Capability
Singapore startup harvests daily human movement for robots
Teaching robots to move and act like humans requires massive amounts of footage of humans actually living their lives, and a Singapore startup just built a device to capture exactly that at scale. Ropedia, co-founded by three Nanyang Technological University graduates, raised $22 million from angel investors connected to Google, NVIDIA, and Amazon, plus Asian venture funds, to grow its head-mounted wearable called HOMIE, which records synchronized video, depth, eye-gaze, and body-movement data to train physical AI and robotics systems. The company also released a free public dataset of over 10,000 hours of recorded human experience, which crossed two million downloads within five days of going live.
July 24, 2026 · Technology & Capability
US and South Korea lock in $700 billion AI chip deals
South Korea and the United States locked in over $700 billion in AI memory and chip supply commitments in a single day, reshaping who controls the hardware that runs the world's largest AI systems. At an AI Summit in San Francisco on July 24, 2026, Nvidia and SK Group signed letters of intent for a partnership valued at more than $500 billion, covering SK Telecom's construction of a 2-gigawatt AI data center using Nvidia's Vera Rubin platform and a long-term co-development agreement between Nvidia and SK hynix on HBM4 high-bandwidth memory. Separately, Samsung Electronics and Broadcom signed a $200 billion memorandum of understanding for Samsung to supply HBM and manufacture Broadcom chips on 2-nanometer and below process nodes through 2030, with South Korean President Lee Jae-myung and executives including Broadcom CEO Hock Tan and Samsung Foundry President Jinman Han in attendance.
July 22, 2026 · Power & Politics
White House Genesis Mission expands to 278 AI science programs
The Genesis Mission expands to 278 selected scientific programs with $5 billion in federal commitments. OSTP Director Michael Kratsios announced the expansion of the Genesis Mission, a White House-led initiative established by a Trump Executive Order in November 2025 to accelerate AI-driven scientific research across 20 federal agencies. The expansion selected 278 projects from more than 5,000 applications spanning healthcare, energy, infrastructure, and national security, with 33 total challenges now involving agencies including DOE, HHS, NASA (led by Jared Isaacman), and the Department of War. Private sector contributions of over $500 million in compute and software access came from Genesis Mission Consortium members including Nvidia, AWS, Microsoft, IBM, and AMD, with more than 340 institutions participating overall.
July 20, 2026 · Technology & Capability
Blacklisted Chinese lab builds 1-gigawatt AI data center on domestic chips
A blacklisted Chinese AI company just finished building one of the largest data centers any Chinese lab has ever built, using only domestic chips. Z.AI, formerly known as Zhipu, brought a 1-gigawatt facility, enough power for roughly 750,000 homes, into partial operation to train its GLM models on Huawei's Ascend chips. Those chips deliver less compute per watt than Nvidia's, so Z.AI is compensating by deploying more of them.
July 15, 2026 · Power & Politics
US firms quietly switch to Chinese AI models
U.S. companies are quietly abandoning American AI models for cheaper Chinese ones to control costs. Lindy.ai founder Flo Crivello moved all of his company's traffic to DeepSeek-V4 after Anthropic bills topped what he paid two dozen employees. Uber burned through its entire annual AI budget in a single quarter, and Airbnb, Perplexity, and Nvidia have all turned to Alibaba's Qwen model instead.
July 8, 2026 · Power & Politics
Warren demands Pentagon and tech firms reveal classified AI deals
Senator Elizabeth Warren wants to know if classified military AI contracts allow mass domestic surveillance or autonomous weapons. She sent letters to the Pentagon and seven companies, including Google, SpaceX, Nvidia, and Microsoft, demanding disclosure of their classified AI deals following the Pentagon's deployment of frontier AI on classified networks in May. Warren set a July 20 deadline and separately pressed Defense Secretary Hegseth on how a contract went to Reflection AI, a startup backed by a venture firm where Donald Trump Jr. is a partner.
July 7, 2026 · Power & Politics
DeepSeek builds its own chips after Nvidia ban
U.S. export controls blocked DeepSeek from Nvidia's chips, so the Chinese startup is now quietly building its own. People familiar with the matter say DeepSeek has spent about a year expanding its chip engineering team and talking to design firms, foundries, and memory suppliers. The push comes as DeepSeek raises its first outside funding round, expected to bring in $7 billion at a valuation near $55 billion.
July 6, 2026 · Power & Politics
Nvidia and Palantir bring AI into classified government networks
Nvidia and Palantir just teamed up to put AI models inside classified government networks. The partnership deploys Nvidia's Nemotron models inside Palantir's Sovereign AI Operating System, letting agencies run fully air-gapped, customizable deployments on their own data while keeping ownership of the resulting models. Nvidia CEO Jensen Huang has cited $1 trillion in visible revenue from its chip lines, with government deals like this one a key channel for that hardware.
June 24, 2026 · Power & Politics
Warren calls for antitrust rules to cover AI pricing
Senator Elizabeth Warren delivered remarks at the Open Markets Institute's 2026 conference on June 24, calling on Congress to strengthen antitrust enforcement tools, apply them to AI companies and algorithmic pricing systems, and appoint judges with deeper knowledge of antitrust law. Warren argued that corporations should not receive exemptions from competition law because their pricing or business decisions rely on algorithms, and she specifically cited AI-driven rent pricing tools as an example warranting regulatory intervention. She also called for greater transparency in merger reviews and referenced the Trump administration's approval of major consolidations as evidence of weakening enforcement.
June 22, 2026 · Technology & Capability
Reflection locks in $6.3 billion SpaceX compute deal
Open-source AI startup Reflection secured a $6.3 billion compute deal with SpaceX to access Nvidia GB300 chips in the Colossus 2 data center.
June 12, 2026 · Power & Politics
Nvidia tops first agentic AI hardware benchmark
NVIDIA Blackwell GPUs lead the first round of results on AgentPerf, a new industry benchmark from Artificial Analysis designed to evaluate the performance of agentic AI systems. The results underscore NVIDIA's continued dominance in providing the hardware infrastructure necessary for the next generation of autonomous AI agents.
June 10, 2026 · Technology & Capability
Google DeepMind releases text-diffusion model four times faster
Google DeepMind launches DiffusionGemma, an experimental open-source model that utilizes a novel text-diffusion architecture to generate text up to four times faster than traditional autoregressive models. By moving away from token-by-token generation, the model explores a more efficient path for real-time AI interactions. NVIDIA simultaneously announces optimized support for DiffusionGemma on its RTX GPUs and DGX Spark systems, accelerating the deployment of these high-speed local AI capabilities.
May 1, 2026 · Power & Politics
Pentagon signs eight tech firms for classified AI use
The Department of Defense announced agreements with eight technology companies, including SpaceX, OpenAI, Google, Microsoft, Nvidia, Amazon Web Services, Oracle, and Reflection, to deploy their AI tools within classified Pentagon networks for "lawful operational use." Anthropic was excluded from the announcement, as the Trump administration had blacklisted the company over its refusal to permit unrestricted military use of its Claude models, including for autonomous weapons and domestic surveillance. The Pentagon stated the agreements would advance its goal of becoming an "AI-first fighting force."
January 27, 2025 · Power & Politics
DeepSeek tops US charts wiping out Nvidia's value
DeepSeek's chatbot app surpasses ChatGPT as the most-downloaded free app on the U.S. iOS App Store. Nvidia's stock loses roughly $600 billion in market cap in a single day, the largest single-day loss in U.S. stock-market history, widely called a "Sputnik moment" for American AI.
November 17, 2023 · Power & Politics
OpenAI board fires Sam Altman
OpenAI's board fires CEO Sam Altman, citing a "lack of candor."
February 20, 2018 · Power & Politics
Musk leaves OpenAI board over Tesla conflict
Elon Musk departs the OpenAI board, citing conflict of interest with Tesla AI.