Is AI infrastructure growth accelerating?

This blog post has been written by the person who has mapped the AI infrastructure market in a clean and beautiful presentation

AI infrastructure spending has shifted into hyperdrive with 97% year-over-year growth in dedicated compute hardware in the first half of 2024 alone.

The numbers tell a story of unprecedented acceleration: global spending reached $47.4 billion in just six months, while hyperscalers committed $320 billion in 2025 capex. And if you need to understand this market in 30 minutes with the latest information, you can download our quick market pitch.

How much did global spending on AI infrastructure grow in 2024, and what are the main drivers?

Global spending on AI-dedicated compute and storage hardware exploded by 97% year-over-year in the first half of 2024, reaching $47.4 billion.

Data center capital expenditures rose 44% for the full year 2024, while cloud infrastructure services spending climbed from $267.7 billion in 2023 to $321.3 billion in 2024—a 20% increase driven primarily by AI workloads.

The primary growth drivers were accelerated servers with embedded AI accelerators, which accounted for 70% of AI server spending in the first half of 2024 and grew 178% year-over-year. Hyperscalers prioritized training hardware as they expanded capacity across AWS, Azure, and Google Cloud platforms. Cloud and shared environments commanded 72% of total AI server spend, reflecting the concentration of investment among major cloud providers.

Storage systems supporting large AI datasets grew 18% year-over-year in the first half of 2024, driven by both training requirements for large language models and inference needs for deployed AI applications. This storage growth represents the infrastructure backbone needed to support the massive data requirements of modern AI workloads.

The acceleration was most pronounced in the third quarter of 2024, when hyperscale data center capex surged 82% year-over-year, with AI servers driving much of the increase.

How has AI infrastructure investment been trending in 2025, and how does this compare with previous years?

AI infrastructure investment has maintained its aggressive trajectory in 2025, with tech giants collectively committing $320 billion in AI capex, up 30% from $246 billion in 2024.

IDC forecasts cloud infrastructure spending to grow 33.3% in 2025 to $271.5 billion, significantly outpacing the 26.1% growth rate seen in 2024. Meanwhile, non-cloud infrastructure is expected to grow 25.8% in 2024 before slowing to low-single digits in 2025, indicating a clear shift toward cloud-based AI infrastructure.

The individual commitments from major hyperscalers reveal the scale of investment: Amazon plans $100 billion, Microsoft $80 billion, Google $75 billion, and Meta $60-65 billion for 2025. These figures represent not just increased spending but strategic positioning for long-term AI infrastructure dominance.

Compared to the double-digit annual growth pattern established since 2019, 2025 spending represents a sustained acceleration rather than a temporary spike. The growth has been consistent across quarters, with cloud and hyperscaler capex leading the charge while traditional enterprise on-premises deployments remain more cautious.

Need a clear, elegant overview of a market? Browse our structured slide decks for a quick, visual deep dive.

If you want updated data about this market, you can download our latest market pitch deck here

What are credible forecasts for AI infrastructure growth in 2026, and what key factors will influence that growth?

UBS projects global AI spending across applications and infrastructure to reach $480 billion in 2026, representing a 33% increase over 2025 levels.

MarketsandMarkets forecasts the AI infrastructure market specifically to grow from $23.7 billion in 2021 to $79.3 billion by 2026, achieving a compound annual growth rate of 27.3%. Dell'Oro Group anticipates worldwide data center capex to maintain a 24% CAGR through 2028, driven primarily by AI-related infrastructure needs.

The key factors influencing 2026 growth include continued expansion of generative AI and large-language-model training, which requires increasingly sophisticated infrastructure. Hyperscalers' sustained capex commitments and diversification beyond the "Big 4" cloud providers will drive broader market competition and infrastructure buildout. Advances in custom accelerators, including ASICs and NPUs, alongside new GPU architectures will enable more efficient and powerful AI processing capabilities.

Regulatory developments will play a crucial role, as governments balance innovation promotion with data sovereignty and security concerns. The implementation of AI governance frameworks in major markets will influence where and how infrastructure investments are deployed, potentially creating regional advantages or constraints for specific providers.

What does current data suggest about AI infrastructure demand over the next 5 years and 10 years?

McKinsey estimates $6.7 trillion in global data center capex by 2030, with $5.2 trillion specifically supporting AI workloads—representing 78% of all data center investment.

Precedence Research projects the AI infrastructure market to expand from $60.2 billion in 2025 to $499.3 billion by 2034, achieving a compound annual growth rate of 26.6%. This nearly 8x growth over nine years indicates sustained demand well beyond current market cycles.

The International Energy Agency forecasts electricity demand from AI-optimized data centers to more than quadruple by 2030, reaching 945 TWh—over twice current consumption levels. This energy demand projection serves as a concrete measure of physical infrastructure growth, as power requirements directly correlate with computational capacity and facility scale.

Long-term demand drivers include the proliferation of AI across industries beyond tech, with sectors like healthcare, finance, and manufacturing requiring dedicated AI infrastructure. The shift from training-focused to inference-heavy workloads will create different infrastructure requirements, driving continued investment in specialized hardware and optimized data center designs throughout the decade.

Where geographically is most AI infrastructure growth happening, and what regions are projected to lead?

The United States dominated AI infrastructure spending in the first half of 2024, accounting for 59% of global investment, followed by China at 20%, Asia-Pacific & Japan at 13%, and Europe, Middle East & Africa at 8%.

Looking forward, IDC expects Asia-Pacific/Japan to lead growth with a 20% CAGR through 2028, followed by the US at 16%, EMEA at 13%, and China at 11%. This shift reflects both the maturation of US markets and the rapid digitization of Asian economies requiring new AI infrastructure.

Precedence Research identifies North America as maintaining the largest market share in 2024 while Asia-Pacific emerges as the fastest-growing region over the next decade. The geographic distribution reflects regulatory environments, with China's slower projected growth partly attributed to export restrictions on advanced semiconductors and AI hardware.

Regional advantages are emerging based on energy costs, regulatory frameworks, and talent availability. Countries with abundant renewable energy sources and favorable AI governance policies are attracting disproportionate infrastructure investment, creating geographic clusters of AI capability that will influence global competitiveness for years to come.

Which specific segments of AI infrastructure are seeing the fastest growth?

Accelerated servers equipped with AI GPUs represent the fastest-growing segment, with 178% year-over-year growth in the first half of 2024 and commanding 70% of total AI server spending.

If you want clear information about this market, you can download our latest market pitch deck here

What are the biggest hurdles that could slow AI infrastructure growth?

Supply chain constraints represent the most immediate threat to AI infrastructure growth, with chip shortages, packaging bottlenecks in HBM3 and GPU assembly, and limited land and power availability for fabrication facilities slowing capacity expansion.

Regulatory fragmentation creates significant complications, as divergent AI rules between regions—including the EU AI Act and the patchwork of US regulations—complicate global deployments and increase compliance costs. Different data sovereignty requirements force companies to build redundant infrastructure across jurisdictions.

The specialized skills gap affects 62% of IT leaders who report infrastructure staffing shortages, particularly in AI-specific hardware management, optimization, and troubleshooting. This talent shortage constrains deployment speed and operational efficiency even when hardware is available.

Energy and sustainability constraints are becoming critical factors, as data centers' share of global electricity demand could reach 1-2% by 2030, prompting efficiency mandates and renewable energy requirements. Grid capacity limitations in key markets are already forcing delays in data center construction and expansion projects.

Wondering who's shaping this fast-moving industry? Our slides map out the top players and challengers in seconds.

What quantitative evidence differentiates real infrastructure growth from hype?

Cloud infrastructure capacity utilization climbed to 85% in 2023 and peaked at 95% in late 2024 before moderating, indicating genuine demand tightening rather than speculative investment.

Real infrastructure requirements are evidenced by hyperscalers now demanding up to 90 MW per data center campus versus the 30 MW typical in 2023—a threefold increase in concrete power infrastructure needs. These power requirements represent physical constraints that cannot be inflated or misrepresented, providing clear evidence of actual computational demand.

The contrast with market reactions to events like DeepSeek's model launch reveals the difference between hype and fundamentals. While stock prices fluctuated dramatically, actual AI infrastructure requirements remained multitiered across evolving workloads, demonstrating that real infrastructure needs persist regardless of market sentiment.

Revenue concentration provides another reality check: accelerated servers with embedded AI accelerators accounted for 70% of AI server spending, showing that investment is flowing to proven, production-ready infrastructure rather than speculative technologies. The 178% year-over-year growth in this segment reflects actual deployment rather than experimental spending.

Which companies are leading the AI infrastructure build-out, and what's their pace of growth?

Amazon leads AI infrastructure investment with $100 billion planned for 2025, focusing on AWS cloud and AI training clusters, followed by Microsoft Azure at $80 billion for AI platforms and GPUs.

If you want fresh and clear data on this market, you can download our latest market pitch deck here

What role is cloud hyperscalers' capital expenditure playing in accelerating AI infrastructure growth?

Hyperscalers now account for approximately 60% of global hyperscale data center capacity, with their collective $320 billion AI capex in 2025 underpinning much of the sector's growth trajectory.

Their aggressive investments set industry benchmarks and accelerate broader infrastructure build-out through several mechanisms. First, they create massive demand for specialized hardware, driving semiconductor innovation and production scaling. Second, their procurement volumes enable cost reductions that benefit the entire ecosystem. Third, their infrastructure standards become de facto industry requirements.

The hyperscaler model also drives geographic expansion of AI infrastructure, as these companies build presence in multiple regions to serve global demand while complying with data sovereignty requirements. Their capex commitments provide visibility and confidence for supply chain partners to invest in expanded production capacity.

Beyond direct infrastructure spending, hyperscalers are investing in custom silicon development—Google's TPUs, Amazon's Graviton processors, and Microsoft's AI chips—creating alternative hardware ecosystems that reduce dependence on traditional semiconductor vendors while optimizing for specific AI workloads.

Looking for the latest market trends? We break them down in sharp, digestible presentations you can skim or share.

How are energy requirements and sustainability considerations shaping future AI infrastructure build-out?

Energy requirements are becoming the primary constraint on AI infrastructure expansion, with data centers' electricity demand projected to more than quadruple by 2030 to 945 TWh according to the IEA.

Optimized cooling technologies offer the most immediate efficiency gains, with liquid immersion, direct-to-chip, and two-phase cooling systems providing up to 90% efficiency improvements over traditional air cooling. These systems are essential for high-density AI workloads that generate significantly more heat than traditional computing applications.

Renewable integration is driving data center location decisions, with facilities increasingly positioned near wind and solar installations. Power purchase agreements (PPAs) and battery storage systems enable data centers to meet sustainability commitments while maintaining operational reliability. This geographic constraint is reshaping the AI infrastructure landscape, favoring regions with abundant renewable energy resources.

AI-driven efficiency improvements include predictive maintenance systems and real-time energy management that reduce power usage effectiveness (PUE) ratios. These systems optimize cooling, power distribution, and workload scheduling to minimize energy consumption while maintaining performance requirements.

Sustainability considerations are also influencing hardware design, with chip manufacturers developing more energy-efficient processors specifically for AI workloads and data center operators implementing circular economy principles for hardware lifecycle management.

What emerging technologies are most likely to unlock the next wave of AI infrastructure growth?

Chiplets and 3D stacking technologies promise higher compute density by enabling multiple specialized processors to work together efficiently, while neuromorphic computing offers brain-inspired efficiency specifically optimized for inference workloads.

• Gate-All-Around FETs: Enable continued transistor scaling beyond current limitations, providing more computational power per chip while reducing energy consumption
• Advanced Packaging (CoWoS): Addresses critical bottlenecks in HBM memory and GPU assembly, enabling faster data access and processing speeds
• AI-Driven Cooling: Auto-tuned thermal management systems that optimize cooling in real-time based on workload demands and environmental conditions
• Optical Interconnects: Replace electrical connections with optical ones for faster, lower-power data transmission between processors and memory
• Quantum-Classical Hybrid Systems: Combine quantum processing units with traditional AI accelerators for specific computational tasks

These innovations collectively promise to multiply available AI compute capacity by 10x by late 2027, according to AI 2027 forecasts. The convergence of these technologies will enable new classes of AI applications while reducing the physical footprint and energy requirements of AI infrastructure.

Advanced packaging technologies are particularly critical for addressing current bottlenecks in memory bandwidth and processor interconnection that limit AI system performance. The development of chiplet ecosystems will enable more modular and efficient AI system designs.

Planning your next move in this new space? Start with a clean visual breakdown of market size, models, and momentum.

Conclusion

Sources

1. IDC AI Infrastructure Spending Report
2. Dell'Oro Group Hyperscale Capex Analysis
3. Canalys Cloud Infrastructure Report
4. IDC Cloud Infrastructure Forecast
5. AI News Tech Spending Analysis
6. UBS AI Market Forecast
7. McKinsey Data Center Investment Study
8. Precedence Research AI Infrastructure Market
9. IEA Energy Demand Forecast
10. IDC Geographic Growth Analysis
11. Dell'Oro Group Infrastructure Forecast
12. Flexential AI Infrastructure State Report
13. Goldman Sachs Power Demand Analysis
14. Sunbird Cooling Technology Innovations
15. AI 2027 Compute Forecast