Research
The AI build-out is driving an infrastructure super-cycle requiring roughly $3 trillion in investment by 2030, but electricity grid constraints — not GPU supply — have become the primary bottleneck, with over 2,060 GW of US generation capacity stuck in interconnection backlogs and Ireland already imposing a moratorium on new data center connections. Meanwhile, AI talent remains stubbornly concentrated in a handful of clusters, salaries have surged past $200,000 with 500,000+ positions unfilled globally, and Huawei's Ascend 910C is closing the capability gap with Nvidia faster than export-control policy assumes. The central unresolved tension is whether Western-style permitting reform and market signals can relieve these bottlenecks faster than state-directed coordination models in China, India, and the Gulf — a question the briefing raises but cannot settle, partly because the underlying evidence skews heavily toward sources outside the market-oriented and structural-inequality traditions that would most sharply contest each other's answers.
Soaring demand for AI-ready data centers offers many opportunities for companies and investors across the value chain.
AI is fueling high demand for compute power, spurring companies to invest billions of dollars in infrastructure.
AI could consume as much as 20% of the world's energy by 2025. 2.9 watt-hours of electricity, fallen to 0.24-0.3 watt-hours,
The United States and China control 90 percent of the computing power needed to develop and deploy frontier AI. They own all 50 of the top-ranked AI foundation ...
AI workloads could represent 50% of all data center capacity by 2030, compared to approximately 25% in 2025. JLL anticipates a critical ...
In a country with abundant hydro- and wind power, as well as plenty of water, rising electricity needs pit industries against each other.
Investors claim that massive investments in energy generation and grid infrastructure are needed to power data center development while ...
The Gulf states are betting big on artificial intelligence. Their motive is simple: While hydrocarbons will remain part of the energy mix ...
Fundamentally, it restricts the export to China of cutting-edge graphics processing units (GPUs) used for AI applications.
The report benchmarks regions based on their core AI assets and capabilities as they relate to three pillars of AI readiness: talent, innovation, and adoption.
Thirty-five percent of all AI engineers in the United States work within a 40-mile radius of San Jose. Another 23% are in Seattle.
In the realm of Generative AI, Silicon Valley has built a concentrated lead over U.S. ecosystems due to its talent and resource advantage.
AI city clusters across Asia Pacific can be broadly classified into three categories (Figure 6). 'Mega Clusters' offer large-scale talent pools and strong city ...
The AI age is expected to require scaling data centers, grid capacity, and supply chains. Deloitte's 2025 AI Infrastructure Survey examines gaps and solutions.
Still, renewables dominated US capacity growth, accounting for 93% of additions (30.2 gigawatts) through September 2025, with solar and storage ...
A comprehensive 2025 CSIS study of the impact of export controls in SME reported that Chinese chipmakers were designing out US equipment. The ...
The Commission for Regulation of Utilities (CRU) put an effective moratorium on new data center grid connections in Dublin, requiring new ...
France emerges as Europe's AI infrastructure leader with Mistral Compute and sovereign cloud providers.
Continued delays only hinder U.S. energy, economic and national security ambitions. Interconnection Queues are Hindering Clean Energy.
Beyond TSMC, most Taiwanese semiconductor firms either do not make products that meet export control thresholds, or else they have received ...
The U.S. still leads in producing top AI models—but China is closing the performance gap. In 2024, U.S.-based institutions produced 40 notable AI models, ...
America's loss could be the EU's gain as inclusive work cultures and a desire for responsible AI can woo global talent from Silicon Valley.
Explore organizational barriers in using AI as a TA tool, how they can be addressed, and how AI adoption can benefit the TA function in the future.
AI engineer base salaries average $206,000 in 2025, with a further 7% increase tracked in Q1 2026 (MRJ Recruitment) and senior specialists ...
Yes, data centers consume water and can impact local supplies – but many are working to minimize that by using recycled water and efficient ...
Supply constraints from HBM shortages and CoWoS packaging bottlenecks are expected to persist through at least H1 2027. NVIDIA's Blackwell ramp ...
Capacity factors are a comparison of net generation with available capacity. See the technical note for an explanation of how capacity factors are calculated.
Singapore commits $1.6B government funding + $26B tech investments to become Southeast Asia's AI hub, generating 15% of NVIDIA's global revenue.
These factory-built reactors promise compact footprints, high-energy density, and predictable operating costs.
NVIDIA commands approximately 80-90% of the AI accelerator market by revenue as of 2025, generating over $100 billion annually from data center ...
In 2025, venture capital (VC) investments in AI firms globally made up over half (61%, USD 258.7 billion) of all VC investment (USD 427.1 ...
This paper addresses this gap by providing a replicable methodology and a comparative analysis of AI job vacancies and AI talent across European countries.
Rising geopolitical tensions and a surge in critical materials like tungsten are reshaping semiconductor supply chain risk and AI supplies ...
The move reflects the increasing strategic importance of India in Progress' global AI innovation, talent and growth plans and underscores the ...
Microsoft CapEx went from $63 billion in 2025 to a guide of over $100 billion for 2026; Google (Alphabet) went from $91 billion to a range of ...
By 2060, DNV estimates ~80% of data centre electricity demand will come from AI, and the sector reaches 11% (6,400 TWh) of final electricity ...
Australia will also aim to use data center development to promote investment in renewable energy. Although several data center projects ...
In Canada, Hydro Québec anticipates that high-tech AI data centres will boost the demand for hydro power by 14% from 2023 to 2032.2; Globally, ...
However, in July 2023, TSMC delayed the facility's opening until 2025, citing a “shortage of specialist workers” as the root cause. This has ...
In general, these laws require data to be stored and processed domestically, with the ultimate aim of enhancing sovereign control over citizens' data. These ...
Addressing water scarcity and improving water management will be immensely important for ensuring the region's stability, sustainability, ...
In brief. VC-backed companies raised over $80 billion in Q1 2025, nearly a 30% increase over an already robust Q4 2024. While a $40 billion AI deal doubled ...
As of the end of 2025, over 2,060 gigawatts (GW) of total generation and storage capacity were actively seeking connection to the grid. However, most projects ...
On a single-chip basis, a performance gap persists; for example, Huawei's Ascend 910C performs at approximately 60% of the level of Nvidia's H100. However ...
AWS is ideal for breadth, maturity and a vast ecosystem; Azure shines for enterprise and hybrid deployments; Google Cloud excels in AI/ML and ...
New Berkeley Lab article documents how deployment of new electric generation is being constrained by current interconnection processes.
To accomplish this, Vestas used Microsoft Azure high-performance computing, Azure Machine Learning and help from Microsoft partner minds.ai, ...
Brazil's advantages for AI infrastructure are compelling. The country possesses nearly one-fifth of the world's water reserves, critical for both hydropower ...
Use our in-depth salary data to learn about competitors' compensation practices. Talent Pool. Hiring? The easiest way to hire top senior tech talent.
Different critical minerals and elements are utilized at each step of the production process – both in “front-end” wafer fabrication, and “back- ...
Sign up to read the full research briefing
Sign up