Skip to content

Let's Data ScienceLEARN • BUILD • STAY AHEAD

News
Blog
Pricing
Contact

© 2026 Let's Data Science

Advertise|Terms|Privacy||Image Rights

Home
News
AI Infrastructure

Topic24 articles · Updated April 24, 2026

AI Infrastructure News

News on the plumbing that powers modern AI: compute and GPUs, inference and training platforms, cloud partnerships, and the supply chain behind the models practitioners use every day.

ai-infrastructure
6.8
Apr 24, 2026
SK Group Expands AI Ecosystem Partnership in Vietnam
SK Group has formalized a multi-party partnership in Vietnam to accelerate AI and energy research, combining corporate investment, provincial cooperation, and university collaboration. The initiative includes an initial USD 300,000 research commitment through SK South East Asia Investment PTE. LTD. to VinUniversity's Center for Environmental Intelligence, and separate MOUs with the Nghe An provincial government and the Vietnam National Innovation Center. The program targets AI-driven energy systems, electric mobility infrastructure, and the groundwork for AI data center planning that leverages SK's energy technologies. This aligns with SK's broader push to build AI data centers across Southeast Asia and to couple compute expansion with efficient cooling and power solutions.
Quick Read+1
intel
6.9
Apr 24, 2026
Intel Sees Revenue Rise Driven by CPU Demand
Intel returned to growth as demand for central processing units (CPUs) strengthens amid a shift toward agentic AI. The company reported revenue growth of 7.2% and signaled next-quarter revenue will exceed market expectations. Executives attribute the change to rising need for CPUs to handle orchestration, control-plane duties, and data management in agentic systems, naming Openclaw and Anthropic's Claude Code as examples of the new workloads. Industry peers, including Nvidia, have echoed the view that the bottleneck is shifting from raw GPU compute to context and coordination, and Nvidia has signaled its own CPU push. Intel also benefits from an ambitious new chip-factory project and a reported 10% U.S. stake. For ML practitioners and infrastructure teams, the story highlights a material rebalancing of compute stack economics and procurement signals that could affect architecture choices for agentic and orchestration-heavy applications.
Quick Read
data-centers
6.6
Apr 23, 2026
Populist Movements Target AI Data Centers
There is a rising trend of populist outrage directed at physical AI infrastructure, with data centers becoming visible symbols of perceived elite accumulation. Investor David Friedberg characterized the data center as the modern "temple of the wealthy," a focal point for anger about uneven economic gains. Opinion writers like Ben Shapiro frame these attacks as the political expression of economic populism, distinct from older moral populism. For practitioners, this is not only rhetoric: physical attacks and heightened opposition create real operational, security, and regulatory risks for AI deployments that depend on centralized compute, power, and fiber connectivity.
Quick Read+1
brookfield
6.9
Apr 23, 2026
Brookfield Builds Backbone for AI Infrastructure
Brookfield Corporation is positioning itself as a critical AI infrastructure provider through three linked plays: selling renewable power to hyperscalers, developing and owning AI data-center real estate, and offering compute via its new subsidiary, Radiant. The company is also packaging AI-themed investment products through Brookfield Asset Management to capture institutional demand. Brookfield's scale, long-term contracts, and balance-sheet capacity let it absorb capital intensity and lock in customers, creating an infrastructure moat rather than competing on software. Key financial context: market cap $103.41B, recent share price about $45.80, PE 95.45, and trailing revenue growth at -12.48%. For practitioners, Brookfield's moves matter because they affect capacity, energy sourcing, and procurement options for large-scale model training and inference workloads.
Quick Read
google-cloud
7.2
Apr 23, 2026
Google Cloud Expands Nvidia GB300 Access for Thinking Machines
Google Cloud has signed a new agreement to expand Thinking Machines Lab's access to NVIDIA Blackwell GB300 GPUs via A4X Max on the Google Cloud AI Hypercomputer. Thinking Machines will run GB300 NVL72 instances and early tests show a 2X improvement in training and serving speed versus prior-generation GPUs, enabled by Google Cloud's Jupiter network for rapid weight transfers. The deal pairs high-throughput compute with Google Cloud services like GKE, Spanner, Cluster Director, Cloud Storage, and Anywhere Cache to support continuous training alongside production serving. The partnership highlights hyperscaler efforts to deliver rack-scale, NVIDIA-optimized infrastructure for frontier model research and reinforcement learning workloads.
Quick Read+2
agentic-ai
8.1
Apr 23, 2026
Google Expands Gemini Enterprise To Power Agentic Workflows
Google launched the Gemini Enterprise Agent Platform at Google Cloud Next 2026, repositioning Gemini as an end-to-end system for building, orchestrating, and governing agentic AI at enterprise scale. The platform converges Vertex AI capabilities into a single offering with a low-code Agent Studio, the Gemini Enterprise app for discovery and deployment, and first-class access to 200+ models through Model Garden, including third-party models. Strategic partnerships with NVIDIA, Accenture, and systems integrators will couple new infrastructure - notably A5X and Rubin GPU scale claims - with professional services to accelerate production agents. The product emphasizes multi-step reasoning, tool integrations, connectors to enterprise systems, and governance controls to mitigate sprawl. For practitioners, this is a platform bet: invest in agent design patterns, observability, and governance if you plan to deploy goal-oriented automation across business workflows.
Quick Read+6
memory-chips
7.1
Apr 23, 2026
SK Hynix Posts Record Q1 Profit on AI Demand
SK Hynix delivered a record first-quarter performance, propelled by surging memory prices and elevated AI datacenter demand. The company reported revenue of 52.58 trillion won and operating profit of 37.61 trillion won, with HBM remaining the strategic growth driver. Strong, sustained purchases from cloud and AI infrastructure customers and longer-term contracts lifted pricing across DRAM and NAND, triggering double-digit to triple-digit quarter-over-quarter price moves in several market reports. SK Hynix retains a leading 57% HBM market share, benefits from close supplier relationships with major AI chipmakers, and faces an industry-wide supply tightness that analysts warn could persist into 2028-2030. For practitioners, the immediate implication is continued upward pressure on AI compute bills and a tighter procurement environment for HBM-equipped GPUs.
Quick Read+4
microsoft
7.7
Apr 23, 2026
Microsoft Expands Australia AI Infrastructure with A$25B Investment
Microsoft is committing A$25 billion (USD 18 billion) to expand Azure AI supercomputing and cloud infrastructure in Australia by the end of 2029. The package includes expanded in-country data centre capacity, deeper national cyber-defence cooperation via the Microsoft-ASD Cyber-Shield, and a major workforce plan to train three million Australians with AI skills by 2028. The move is Microsoft's largest-ever investment in Australia, aligns with the Australian government's data-centre expectations, and positions Australia as a tier-one market for low-latency AI services and sovereign compute. Expect faster access to high-performance Azure resources for local enterprises, stronger public-sector security integrations, and an enlarged AI talent pipeline.
Quick Read+8
world-it-show
6.6
Apr 23, 2026
World IT Show Spotlights AI-driven Innovation and Commercialization
The 2026 World IT Show opened in Seoul April 22, convening 460 companies from 17 countries across 1,400 booths to demonstrate AI moving from research into production. The show emphasized "physical AI" and sovereign infrastructure, with telecom leaders pivoting into full-stack AI providers. SK Telecom staged an 864 m2 pavilion showcasing A.X and the 500B-class A.X K1, AI-RAN, GPU-as-a-Service "Haein," agent services, digital twins and robot training platforms. Samsung, LG, KT and others displayed AI-enabled devices and industrial applications. The event included business matchmaking-50 overseas buyers met 190 Korean firms-and awards recognizing AI startups, including a presidential prize for Gaudio Lab's audio AI platform. For practitioners, the show signals intensified national investment in sovereign models, carrier-driven AI infrastructure, and pragmatic deployments across consumer and industrial stacks.
Quick Read+4
ai-infrastructure
7.2
Apr 23, 2026
Gartner Forecasts IT Spending Growth Driven by AI Infrastructure
Gartner projects worldwide IT spending will reach $6.31 trillion in 2026, a 13.5% increase from 2025, led by massive investment in AI infrastructure and data center expansion. Spending on data center systems is expected to surge 55.8%, approaching $788 billion, while IT services will remain the largest category at over $1.87 trillion. Software spending, buoyed by generative AI development, is set to grow 15.1%. The forecast highlights supply-driven price inflation in high-bandwidth memory, which is reshaping device replacement cycles and margins. The report describes a multi-speed IT market: hyperscalers and AI-focused segments outpace traditional IT categories, creating strong demand for GPUs, accelerators, memory, and high-throughput interconnects.
Quick Read
us-iran-conflict
6.7
Apr 22, 2026
US-Iran Conflict Escalates; Oracle and OpenAI Expand AI Infrastructure
The United States and Iran are experiencing an escalation in conflict, and the Wall Street Journal reports 400 US military personnel have been reported. Meanwhile, Oracle and OpenAI are advancing AI infrastructure, indicating continued enterprise investment in large-scale model deployment.
Quick Read
google
8.0
Apr 22, 2026
Google Commits Massive AI CapEx as Bitcoin Rallies
Alphabet is committing an unprecedented $175 billion to $185 billion in 2026 capital expenditure to expand AI data centers, proprietary silicon, and global AI infrastructure. CEO Sundar Pichai framed the spending as essential to the AI arms race; Google Cloud CEO Thomas Kurian said Gemini AI will power a more capable Siri for Apple by late 2026. Markets responded: Bitcoin jumped 4.4% to about $79,100, driven by roughly $1.54 billion of ETF inflows over six days and a short squeeze that erased about $200 million in bearish positions. Separately, the US is reportedly close to a $500 million bailout of Spirit Airlines, potentially giving the Treasury up to 90% equity. The package of tech investment, crypto flows, and policy moves creates immediate demand signals for chips, data center capacity, and cloud services.
Quick Read
marvell
7.6
Apr 22, 2026
Marvell Expands Role in Non-GPU AI Infrastructure
Marvell Technology has moved from a networking specialist to a central supplier for non-GPU AI infrastructure. Two developments are driving the re-rating: Nvidia's $2 billion strategic investment and reports that Google is in talks to co-develop a memory processing unit (MPU) and an inference-optimized TPU with Marvell. Those moves position Marvell to capture XPU attach spend, high-speed interconnect demand, and optical networking upgrades as hyperscalers scale inference deployments. The market is pricing a multi-year earnings trajectory, Marvell already reports record fiscal-year revenue near $8.2 billion and 18 active ASIC designs, but the commercial outcomes hinge on contract timing, production partners, and design wins versus incumbents like Broadcom.
Quick Read+5
vast-data
7.6
Apr 22, 2026
Vast Data Secures $1B Funding at $30B Valuation
Vast Data closed a $1 billion Series F that values the company at $30 billion, with participation from Nvidia, Drive Capital, Access Industries, Fidelity, and NEA. The raise includes primary proceeds and a secondary tender that lets employees and early investors sell shares. Vast's software-first AI infrastructure, branded as an AI Operating System built on the DASE architecture, powers large-scale GPU fleets for customers including CoreWeave, Google Cloud partners, and the U.S. Air Force. The company reports more than $4 billion in cumulative bookings and over $500 million in committed annual recurring revenue. Nvidia's participation underscores chip-to-stack strategic investing as storage and data fabrics capture outsized attention from investors supporting GPU-driven AI growth.
Quick Read+4
samsung
6.8
Apr 22, 2026
Samsung Memory Division Prioritizes AI Over Smartphone Business
Samsung's memory unit is shifting capacity toward AI customers, creating tight supplies and higher prices for mobile components. TM Roh, who leads Samsung's Device Experience (DX) and Mobile Experience (MX) units, warned that the MX business could run an annual deficit because DRAM and NAND flash prices have risen and memory allocation is skewing to data-center/AI customers. Samsung is favoring faster LPDDR5 and LPDDR5X parts over cheaper mobile variants, reducing available inventory for Galaxy S26 production and pushing handset costs up. Reported AI accelerators are consuming orders at scale, one next-generation GPU cited needs 1.5TB of LPDDR5X, amplifying shortages. The shift improves short-term margin opportunities in AI infrastructure but raises supply-chain, product-cost, and market-share risks for Samsung's smartphone business.
Quick Read
ai-infrastructure
7.3
Apr 22, 2026
I/ONX Delivers Symphony SixtyFour Cutting AI Infrastructure Costs
I/ONX launched the Symphony SixtyFour, a high-density AI platform that supports up to 64 accelerators per node and claims to eliminate the legacy "host tax." The company says the design recovers 30kW of wasted support power per rack, reduces rack-scale deployment costs by up to 70%, and delivers 30-50% total cost of ownership savings across inference and fine-tuning lifecycles. Symphony SixtyFour collapses what would be an eight-node cluster into a single node, draws roughly one-fourth the power of traditional 64-device clusters for inference, and removes the need for liquid cooling in inference-only deployments. The announcement targets enterprise AI teams seeking power- and cost-efficient production inference, but independent benchmarks and software ecosystem support will determine real-world impact.
Quick Read+1
bybit
6.9
Apr 22, 2026
Bybit AI Launches MCP for Multi-Agent Trading Infrastructure
Bybit launched an official Model Context Protocol (MCP) that standardizes infrastructure for multi-agent trading. The MCP enables integration of AI agents such as Claude, ChatGPT, and Cursor with Bybit services without custom API work, exposing real-time market data, order book snapshots, candlestick aggregations, fee schedules, portfolio visibility, and trading capabilities that include spot, perpetual futures, conditional orders, stop-loss, and take-profit. The protocol is designed for scalable single-agent to multi-agent architectures and supports discovery and interoperability for MCP-compatible applications. For traders and developers this reduces integration friction, accelerates deployment of automated strategies across the data-to-execution path, and shifts exchange infrastructure toward AI-native workflows.
Quick Read
energy-costs
6.8
Apr 22, 2026
UK Firms Shift AI Infrastructure Abroad Over Energy Costs
Rising electricity prices and constrained grid capacity are pushing British AI teams to run models and training workloads outside the UK. A CUDO Compute study, using a Censuswide survey of 700 senior AI decision makers, finds one in five UK firms have already moved AI workloads overseas and 32% of AI-first companies are considering relocation because of power costs. At the same time about 30% of organizations prioritize sovereign or regionally controlled compute despite higher expense. The mismatch between policy ambitions for domestic AI leadership and the realities of local energy and grid constraints is acute, prompting government moves on electricity contracts and levy changes to blunt gas-price-driven spikes.
Quick Read
gartner
7.6
Apr 22, 2026
Gartner Forecasts IT Spending Surge Driven by AI
Gartner projects global IT spending will rise 13.5% in 2026 to $6.31 trillion, driven primarily by hyperscaler datacenter builds and AI infrastructure. The firm sees data center systems as the fastest-growing category, with spending up 55.8%, and expects AI-related investment to approach half of total IT budgets by 2027. Consumer device growth and some enterprise categories lag, but hyperscaler purchases of AI-optimized servers, high-bandwidth memory, and IaaS are creating a multi-speed market. Geopolitical pressure is also accelerating sovereign-cloud investment in many regions, reinforcing regional capacity and compliance-focused deployments. For practitioners, this means stronger demand for AI-capable hardware, tighter component supply dynamics, and continued focus on ROI for enterprise AI initiatives.
Quick Read+3
cpu-prices
6.9
Apr 22, 2026
AMD and Intel Raise CPU Prices Amid AI Demand
AMD and Intel are raising consumer and server CPU prices as AI infrastructure demand tightens semiconductor capacity. Consumer CPU pricing climbed 5-10% in March while server CPUs rose 10-20%, and supply-chain sources expect another round of hikes through Q3 2026 with cumulative increases reaching 16-17% for some SKUs. Drivers are the renewed role of CPUs in emerging "Agentic AI" workloads, continued memory and component scarcity, and capacity constraints at advanced process nodes such as 2nm and 3nm. Shortages are forecast to persist through 2026 and into 2027, forcing OEMs, ODMs, and cloud buyers to replan procurement, budgets, and architecture trade-offs between CPUs and accelerators.
Quick Read
sk-hynix
6.1
Apr 22, 2026
SK Hynix Builds $13B Packaging Fab to Scale HBM
SK Hynix will invest KRW19 trillion (about $12.9 billion) to build a new advanced packaging and test facility in Cheongju, South Korea, aimed at expanding production of high-bandwidth memory for AI datacenter accelerators. Construction of the facility, named P&T7 in company briefings, begins in April with a targeted completion by end-2027. The plant focuses on advanced packaging processes that are critical to HBM stacking and module test, addressing a tight supply environment that has driven DRAM and NAND contract prices sharply higher. The move increases SK Hynix's capacity for HBM and co-packaged modules, eases constraints for GPU vendors and hyperscalers, and reinforces the industry shift toward specialized packaging as the bottleneck in AI memory supply.
Quick Read+8
hbm
7.1
Apr 22, 2026
Samsung Postpones HBM5E Production After D1d Yield Shortfall
Samsung Electronics has indefinitely postponed mass production of its next-generation HBM memory, citing insufficient yields from its D1d (7th Gen 10nm) DRAM. The company received pre-production approval but is postponing mass production until yields meet internal targets; concerns have been raised about the ROI of initiating a trial run. Samsung says it will re-examine the process roadmap and allocate resources to improve yield, while continuing investment in a large Onyang fabrication and packaging facility. The delay threatens timelines for HBM5E adoption in high-bandwidth AI accelerators and may shift market share toward competitors that reach production readiness sooner.
Quick Read
ai-infrastructure
7.2
Apr 21, 2026
Core Scientific Raises $3.3B to Build AI Data Centers
Core Scientific is selling $3.3 billion in high-yield junk bonds to finance a strategic pivot from crypto mining into AI infrastructure. The company plans to develop six data center facilities that are contracted to be leased to CoreWeave under a 12-year agreement, a move Bloomberg and other outlets estimate could generate roughly $10 billion in revenue. The financing via junk bonds signals elevated execution and refinancing risk but accelerates capacity expansion for GPU-heavy workloads. For practitioners, the deal matters because it will increase demand for commodity GPUs such as H100-class accelerators, intensify competition for power and colocation capacity, and create more specialized AI rack inventory on the market.
Quick Read+2
space-tech
7.1
Apr 21, 2026
SpaceX Debt Jumped to $23 Billion Last Year
SpaceX debt rose to 23 billion dollars last year. Apollo Global Management is nearing 3 billion in chip funding tied to XAI.
Quick Read

Showing 24 articles

See the full news feed