Getting your news
Attempting to reconnect
Finding the latest in Climate
Hang in there while we load your news feed
Microsoft
Data center news, project activity, and monthly briefings for Microsoft.
Microsoft · Construction & power moves · 5
full tracker →Where Microsoft is securing land and power — each traced to primary filings.
Editor's picks
-
CMA takes first steps to ensure choice in business software
The Competition and Markets Authority (CMA) has proposed designating Microsoft with strategic market status in its business software ecosystem and has opened consultation before making a final decision.
- The CMA says it will gather further evidence on bundling, Copilot/AI choice, security product integration, and cloud licensing, with a roadmap of possible interventions aimed at improving choice for UK businesses and public sector organisations.
- The CMA also provided an update on its cloud services engagement with Microsoft and AWS, saying both companies have made material steps including removing or reducing data transfer fees, enabling interoperability requests, and rolling out AI tools to support cloud switching; a further update is due in Spring 2027.
-
Grid Forward Releases 100th GridPulse Podcast Episode Tackling the Big Questions for Today’s Electric Grid
Grid Forward has announced the release of the 100th episode of its GridPulse podcast.
- The episode, titled “Our Big 100th Episode Tackles the Big Questions for Today’s Grid,” features Jigar Shah, Audrey Zibelman, and Herman Trabish, with Bryce Yonker guiding the discussion over a 50-minute episode.
- The podcast launched in May 2020 and the episode covers AI and distributed energy resources, grid planning, rate structures, public trust, and domestic energy supply chains; the release also notes sponsors and past guests across utilities, technology firms, data center developers, universities, and government organizations.
-
Advancing hardware security with open-source Integrated HSM technology
Microsoft has announced Project Tessera, an open-source evolution of Azure Integrated Hardware Security Module (HSM), at the Open Compute Project Global Summit.
- Microsoft is collaborating with Marvell, Oracle, and the Open Compute Project to make the Tessera Integrated HSM architecture, specifications, protocols, drivers, software, firmware, and reproducible build environments open source.
- The company says the project will progress toward a fully open-source silicon implementation and that Microsoft and Marvell demonstrated HSM partition cloning and resiliency rings between a Tessera Integrated HSM and a Marvell LiquidSecurity 2 HSM; the repository is available at GitHub.
- Microsoft also says the initiative will extend into confidential computing with a minimal unikernel-based firmware stack, leveraging AMD SEV-SNP and OpenVMM.
-
OCP Global Summit: Amid Ecosystem Adoption, NVIDIA Opens Vera Rubin Architecture to Help Scale AI Factories Worldwide
NVIDIA announced a set of new partnerships, open standards, and open-source contributions at the Open Compute Project (OCP) Global Summit in San Jose, focused on scaling AI factory infrastructure.
- NVIDIA contributed Vera Rubin rack and compute tray designs to OCP via MGX and joined partners in advancing open standards for AI factories, including the Open Data Center Specification v0.7 and the OCP Grid Resilience Specification.
- The company also announced new DSX AI Factory open-source software components, as well as networking collaborations with Meta, Cisco, and Google covering Spectrum-6, USFP, and Falcon Transport; Visionbay.ai also unveiled an 800V DC AI factory design.
- This is a new announcement from NVIDIA, presented as event news and partner updates rather than a retrospective; the article also references earlier-created NVIDIA architectures and specifications as context.
-
Deep dive into Azure Maia 300 architecture
Microsoft has announced Azure Maia 300, a third-generation AI accelerator system optimized for LLM inference and cloud-scale agentic workloads.
- The system is co-designed across software, silicon, rack infrastructure, networking, and datacenters, and is intended to maximize tokens per watt and tokens per dollar for Microsoft’s AI products.
- Microsoft says Maia 300 delivers up to 7x greater rack-scale token throughput and more than 2.5x performance per watt versus Maia 200; the design includes 96 accelerators per rack, 12,288 accelerators per superpod, 167 kW racks, and 8 HBM4 stacks with 288 GB capacity and 18 TB/s bandwidth per accelerator.
-
L’Uruguay accelera sulla transizione e sulla logistica
Uruguay has announced and described a broad set of infrastructure, energy, logistics, digital, and industrial initiatives, alongside trade and investment figures with Italy and the region.
- The government’s 2026-2029 development plan foresees $9.63 billion in infrastructure investment, with about 45% for transport and logistics; it also includes rail, port, road, and bridge upgrades, plus Montevideo port rail extension and a new floating dock at Puerto Capurro.
- Separate initiatives include a Montevideo electric BRT due to start in 2027 with a cost of at least $590 million, an IDB CCLIP of $500 million plus $10 million for project preparation, a $1 billion Salto Grande renovation program, and major energy projects including Melo solar park and Paysandú e-fuels with investment above $5.3 billion.
- The article also reports on Uruguay’s IT sector, AI and data center activity near Montevideo, bilateral trade with Italy, MERCOSUR-related infrastructure financing, and a long interview section discussing policy priorities and business opportunities.
Recent news
-
Could the Societal Value of Workloads Shape Data Center Policy?
The article discusses a policy debate about whether data centers or their workloads could face future regulation based on societal value, rather than announcing a new project or rule.
- A panel at Informa’s Data Center Europe event in Vienna (Oct. 13-15) said workload-based scrutiny may emerge, citing Finland’s grid connection rules that prioritize data centers under 1 MW.
- Speakers from ADVANT d.o.o., ITIC, and the European Data Centre Association said current European regulation does not differentiate by workload type; the article frames the issue as speculative for now.
- Background references include AI-driven growth, Goldman Sachs forecasts, and public criticism over power demand, water use, land consumption, and grid impact.
-
Google data center secures Van Buren Township approval, with startup expected in late 2027
Van Buren Township has approved the final site plan for Google’s proposed 1-gigawatt hyperscale data center near Detroit Metro Airport, clearing the way for construction of Michigan’s second hyperscale facility.
- The Planning Commission voted 5-1 after nearly an hour of public comment; Google says Project Cannoli should begin drawing power by late 2027 / December 2027 and ramp to 1 GW within a year.
- Google secured electricity through a 20-year contract with DTE Energy; regulators approved special contracts on Oct. 1, and environmental regulators approved wetland permits in August.
- The contract commits Google to fund 480 MW of energy storage, 1.6 GW of renewable energy, and 300 MW of other power capacity; it also requires payment for at least 80% of claimed power and includes penalties for early termination.
- Local incentives include an estimated $125 million in tax relief over 12 years and a separate $15.4 million payment to the township for infrastructure projects, in exchange for at least 51 full-time jobs.
-
IBM, IonQ, Diraq, and Atom Computing advance to DARPA’s quantum hardware testing stage
DARPA has announced that four quantum computing companies have entered Stage C of its Quantum Benchmarking Initiative and may receive up to $300 million each for independent testing and validation of their hardware.
- The new Stage C companies are IBM, IonQ, Diraq, and Atom Computing; earlier Stage C participants were Microsoft and PsiQuantum through a precursor program.
- IBM says it will continue investing in quantum computing, IonQ says the award will support Superion development, Diraq says the award reinforces its silicon spin qubits path, and Atom says it sees a near-term route to practical quantum computing; the article also cites DARPA, Bob Sutor, and Gartner commentary on the program and industry outlook.
-
Centre To Roll Out Data Centre Policy Amid Power Cost Concerns
The Indian government is preparing a data centre policy, announced by Minister of State for IT Jitin Prasada at a Central Electricity Authority conclave.
- Prasada said the policy is being prepared amid concerns over rising electricity demand from data centres and its effect on consumer power bills; he did not disclose the scope or rollout timeline.
- He linked the policy to India’s push to attract hyperscaler data centres while protecting environmental sustainability and water use; the article also cites protests against Amazon and Google projects and mentions CtrlS Datacenters’s separate ₹10,000 Cr Bihar investment and ₹400 Cr Patna data centre inauguration.
- This is a report about a new policy being prepared, not a detailed government release of the final policy text.
-
With 90% of its developable land utilised, New Noida to come up in over 20,000 hectares
Noida Authority has announced a 20,000-hectare greenfield expansion called New Noida to support industrial growth and phased land acquisition.
- The plan is being presented as a new infrastructure push for Noida, with land acquisition slated to begin in phases and existing areas also being considered for higher FAR to enable vertical redevelopment.
- The authority said the next groundbreaking ceremony will include about 120 industrial proposals with an estimated investment of nearly Rs 35,000 crore across data infrastructure, electronics manufacturing, commercial real estate, logistics, and institutional development; it also said the region already hosts major firms including Samsung, Dixon, TCS, Infosys, HCL, Microsoft, Wipro, Adobe, Sify, and AdaniConneX.
-
Microsoft Pushes Custom AI Chips with Azure Maia 300
Microsoft has announced Maia 300 and said Maia 200 is now in production, while also outlining plans to scale the Azure Maia family to multi-gigawatt deployments in 2027.
- Maia 300 is Microsoft’s third-generation in-house AI accelerator; it is planned for production in 2027 and is designed for inference-first workloads including long-context and memory-intensive AI tasks.
- Maia 200 is already deployed across Microsoft data centers in the US, Europe, and Australia, powers Copilot and other cloud services, and is expected to extend to GitHub Copilot and Microsoft Foundry by the end of the year; Microsoft says it delivers up to 1.7x more tokens per watt than leading GPU infrastructure and Maia 300 could deliver 40% more tokens per megawatt and up to 7x rack-scale throughput versus Maia 200.
-
Trump Honors Microsoft’s CEO After Vance Drops an H-1B Visa Hammer on the Company
The White House honored major technology leaders while Vice President JD Vance separately criticized Microsoft over H-1B visa use and said the company was being suspended from a program for skilled foreign workers.
- JD Vance said Microsoft had abused the visa system, citing 6,000 U.S. worker layoffs last year alongside 6,300 H-1B visas and almost 3,000 green cards; Microsoft said most H-1B applications were to extend or change status for existing employees.
- President Donald Trump awarded the National Medal of Technology and Innovation to Satya Nadella and gave National Medal of Science awards to Elon Musk, Sergey Brin, Jensen Huang, and Lisa Su; he also praised AI and data center investment as helping the U.S. economy.
-
It’s not your model — AI has a data movement problem
The article argues that enterprises scaling AI are often facing network and connectivity problems rather than compute or model failures.
- It cites recent research showing 59% of organizations report bandwidth issues and 53% report latency challenges, and says AI agents can generate up to 450% more network traffic per task than humans.
- The piece references Cisco Live, Azure, and Google Cloud as examples, and emphasizes that performance problems often emerge across Latin America, Asia, and Africa when AI deployments move beyond initial local testing.
-
Data residency is not enough, says Airtel at IMC 2026, as Indian rules still focus on location
Airtel’s Chief Business Officer Abhishek Biswal said at India Mobile Congress 2026 that data sovereignty goes beyond storing data in India and includes access and operational control.
- Biswal said sovereignty has three layers, ending with operational sovereignty over physical infrastructure, logs and telemetry; he also warned that foreign jurisdictions could switch off access or compel disclosure.
- The article is a panel discussion / commentary piece, not a first-time product launch announcement; it references Airtel’s earlier Airtel Cloud launch in August 2025 and examples involving Microsoft, Nayara Energy, and SAP India.
- The piece cites policy context including the RBI’s 2018 directive, CERT-In’s 2022 directions, the DPDP Act, 2023, and the US CLOUD Act of 2018; it also quotes GSMA Intelligence survey figures of 75% and an average premium of 15% mentioned by the speaker.
-
Beyond the Demand Forecast: Six Questions About Powering Data Centers From DPX 2026
POWER magazine has published a conference-based analysis of DPX 2026, highlighting the main questions discussed about building power infrastructure for data centers and AI load growth.
- The article summarizes remarks from Peter Lake, Jehmal T. Hudson, Alexander Hubbard, Matthew Gardner, Nelli Babayan, and Jason Reagan rather than announcing a single new project.
- It includes concrete policy and market details such as Virginia’s GS-5 rate class effective Jan. 1, 2027, Berkeley Lab’s 192 TWh to 649 TWh U.S. data center electricity forecast, and GE Vernova’s planned turbine output ramp to 20 GW in Q4 2026, 24 GW by 2028, and 30 GW by 2030.
-
AMD to ramp up CPU, GPU supply in 2027 as AI demand surges
AMD has announced plans to substantially increase CPU and GPU supply in 2027, with CEO Lisa Su saying the company is already expanding supply and capacity in 2026.
- Su said AMD is planning capacity three to five years ahead and was in Taiwan on October 5-6 to ramp supply for CPU and GPU production.
- She reportedly met Foxconn and was scheduled to meet TSMC; she also visited South Korea and met Samsung Electronics’ Device Solutions (DS) to broaden partnerships.
- The article also notes AMD’s 2027 ramp depends on partners including TSMC, SK Hynix, and Samsung, with constraints tied to HBM4, CoWoS packaging, and foundry capacity.
-
Google enters massive 3.6-GW power deal with Constellation Energy
Google has announced a long-term power procurement deal with Constellation Energy to secure 3,590 megawatts of electricity in the US’ largest power grid, PJM Interconnection.
- The agreement includes a 20-year power purchase agreement for 890 MW from multiple Constellation-owned nuclear plants, plus an additional 2,700 MW long-term supply agreement in PJM not tied to a specific generation source.
- The deal is expected to support more than $4.3 billion of new investment and upgrades to 11 nuclear units in Illinois, Pennsylvania, and New Jersey; first upgraded plant deliveries are expected in 2028. The companies said the arrangement responds to PJM’s proposed “bring your own power” approach for data center customers.
-
Can Data Center Design Choices Change Public Acceptance?
The article discusses how data center operators, designers, and community groups are trying to improve the appearance and siting of data centers to reduce local opposition, with the UK Government announcing a new design competition.
- UK Government announced a national Data Center Design Challenge on June 11 with RIBA to encourage architects, engineers, designers, and communities to collaborate on high-quality design, public engagement, and sustainable outcomes.
- Examples cited include Microsoft using biomimicry at a Noord-Holland, Netherlands site in 2022 and expanding it to more than 20 sites across the US and Europe; Kao Data is prioritizing former industrial brownfield sites for redevelopment.
-
CoreWeave Targets Enterprises with Forge Platform
CoreWeave has announced CoreWeave Forge, an integrated software platform for organizations to develop, run, and continuously improve AI models and agents.
- The platform was announced Sept. 30 during CoreWeave’s Fully Connected event in San Francisco and is available today in Free, Pro, and Enterprise editions.
- CoreWeave said Forge unifies existing products with new software for training, inference, evaluation, observability, and agent development; it also launched a partner network with VAST Data, CrowdStrike, and ClickHouse.
- The company also said Nvidia Vera CPU access is in early phases, Vera Rubin NVL72 is now available on its cloud platform, and Cognition is the first customer running workloads on the hardware platform.
- CoreWeave Aria is now available after being previously in preview; the AI research assistant helps customers analyze data and recommend improvements.