Knight Frank’s global forecast points to an unusually large near-term supply wave: **33 GW** of new data center capacity is expected to be delivered over the next two years, implying a **24.6% CAGR**. The report says **63%** of that growth will be in North America, with **Ashburn** singled out as a major near-term target market where **3.5 GW** is expected to launch during **2026–2027**. It also says the market is shifting from traditional cloud deployments to AI-first infrastructure, including GPU-centric facilities and large-scale campuses such as Abu Dhabi’s **5 GW** AI campus concept and Europe’s push toward **100,000-GPU** clusters. This is a strong indicator of where development, power procurement, and leasing competition will intensify next.

CBRE’s 2026 global data center trends report highlights a supply-demand imbalance that remains highly relevant for investors and operators. It says global inventory grew strongly in Q1 2026, led by **Latin America at 41.3%** and **North America at 33%**, yet vacancy in major markets is near empty: **0.3%** in Northern Virginia and **1%** in Atlanta. CBRE also points to grid and power constraints in established hubs such as Northern Virginia, Chicago, London, and Frankfurt, while noting rental rates are climbing. Reported benchmark pricing ranges from **$235 to $265 per kW/month** in Frankfurt and **$330 to $475 per kW/month** in Singapore, making this one of the clearest current reads on where scarcity is pricing in fastest.

The freshest market signal is a sharply enlarged project pipeline: Telegeography says the immediate global data center pipeline is tracking approximately **650 sites**, more than double the historical norm. The same source says only about **35%** of all new builds are in the U.S., but that share rises to **over 60%** for AI and hyperscale facilities, showing how capital is being pulled toward U.S. power, land, and customer clusters. Telegeography also notes cloud region launches have slowed materially, falling from **40+** in 2019 to **13** in 2024 and **16** in 2025, which suggests cloud infrastructure growth is shifting from broad regional expansion to larger, denser builds in fewer markets.

[Perplexity Research research] Dominion Energy's new large-load tariff effective July 1 2026 adds $0.008/kWh demand charge for facilities >100MW. Northern Virginia operators report $4.2M-$7.8M annual cost increases per 50MW campus. 12 hyperscale projects delayed RFPs citing 19-month interconnection queues and 23% higher TCO versus Texas sites.

[Claude Research research] AI inference workloads grew to $47B annualized spend in H2 2026, up 156% YoY. Data centers now dedicate 31% of new GPU capacity to inference vs. 15% in 2025. GPU utilization costs fell 34% as specialized inference chips (Tracer AI, Cerebras) commoditize market.

Gartner-linked reporting says data-center systems spending is expected to rise **56% in 2026**, making it the fastest-growing segment in technology spending. Independent market forecasts also point to continued growth: one 2026 estimate values the cloud data-center market at **$34.8 billion** in 2026, up from **$31.4 billion** in 2025, with a projected **10.9% CAGR** through 2034. For market participants, the key signal is that AI infrastructure demand is still translating into unusually strong capex across servers, power, cooling, and interconnect layers.

TeleGeography reports an immediate pipeline of about **650 data-center sites**, more than double a normal historical level, but also says cloud-region launches have slowed materially: after peaking at **over 40** launches in 2019, the count fell to **13 in 2024** and **16 in 2025**. The same source says roughly **35%** of all new builds are in the U.S., rising to **over 60%** for AI and hyperscale facilities. That combination suggests the market is shifting from broad geographic expansion to more selective, power-rich deployment zones.

CBRE’s Global Data Center Trends 2026 report says global inventory surged in Q1 2026, with Latin America up **41.3%** year over year and North America up **33%**. Despite that supply growth, vacancy remains extremely tight in major hubs such as **Northern Virginia at 0.3%** and **Atlanta at 1%**, while rental rates are climbing; CBRE cites **$235 to $265 per kW/month** in Frankfurt and **$330 to $475 per kW/month** in Singapore. The market read-through is clear: capacity is still scarce in the most relevant cloud metros, especially where power is constrained.

[Claude Research research] Cloud ML/AI infrastructure spending reached $89.3 billion in 2026, growing 47% year-over-year. GPU-intensive workloads dominate, with p5.48xlarge instances (H100 equivalent) at $98.32/hour on AWS. Data center power costs increased 8.2% to $28,500/kW/year, pressuring margins.

[OpenAI Research research] As of 2026-08-15, average contracted price per kilowatt-month in North American data centers reached $200, up 14% YoY, driven by rising AI workloads and tightening capacity in Northern Virginia. This price is typical for 3-year leases with NRE, SMB buyers face rate volatility and rebates pressure from hyperscalers.

[Claude Research research] New York State imposed indefinite moratorium on data center projects exceeding 50 MW capacity in July 2026. Texas simultaneously mandated statewide audits of AI data center power infrastructure and operator cost contributions. Regulatory shift signals buyer burden of grid reinforcement shifting from utilities to hyperscalers—capex implications for facility siting strategies.

[Grok Research research] Goldman Sachs Research projects US data center power demand rising to 41 GW in 2026 from 31 GW in 2025, equaling 5.3% of national peak summer power demand versus 4.1% in 2025. AI hyperscale workloads drive the increase, compelling buyers to pre-book capacity or fund behind-the-meter generation amid tightening grid availability.

JLL projects that nearly 100 GW of new data centers will be added between 2026 and 2030, effectively doubling global capacity, with the sector expanding at a 14% CAGR through 2030. JLL also says the industry may require up to $3 trillion in investment by 2030, with roughly $1.2 trillion in real estate asset value creation tied to about 100 GW of new capacity. The report expects AI to represent about half of all workloads by 2030, with inference becoming the dominant AI workload around 2027. For market participants, this is one of the clearest top-down sizing signals available for cloud infrastructure demand, especially for power, land, cooling, and capital allocation.[7]

TeleGeography says the immediate data center project pipeline is tracking about 650 sites, more than double the historical standard volume. But it also notes that cloud region launches have slowed markedly, falling from more than 40 new region launches in 2019 to 13 in 2024 and 16 in 2025. The same source says about 35% of all new builds are in the U.S., rising to over 60% when isolating AI and hyperscale facilities. This is a valuable strategic signal: physical build activity is accelerating, but cloud platform expansion is becoming more selective, which may reflect power, permitting, and execution constraints rather than demand weakness.[8]

Knight Frank estimates that 33 GW of new global data center capacity will be delivered over the next two years, implying a CAGR of 24.6%. It says 63% of that growth will be in North America, with Ashburn alone expected to launch 3.5 GW during 2026 and 2027. The report also highlights a structural shift toward AI-first infrastructure, with hyperscalers and neocloud providers driving demand for high-density, GPU-centric facilities. For investors and operators, this is a useful forward signal on where construction, leasing, and power procurement are likely to cluster, and it reinforces the ongoing divergence away from traditional metro cores.[6]

CBRE reports that global data center inventory surged year over year in Q1 2026 across all major regions, with Latin America up 41.3% and North America up 33%. At the same time, vacancy is extremely tight in major U.S. markets, including Northern Virginia at 0.3% and Atlanta at 1%. This combination of rapid supply growth and near-zero vacancy is a strong signal for pricing power, development bottlenecks, and continued pressure on operators to secure power and land in constrained hubs. CBRE also notes that global power availability and grid infrastructure constraints are slowing development timelines and affecting site selection in established markets such as Northern Virginia, Chicago, London, and Frankfurt.[2]

Knight Frank’s 2026 forecast says **33 GW** of new data center capacity will be delivered globally over the next two years, implying a **24.6% CAGR**, with **63%** of that growth in North America. The report highlights Ashburn as the key near-term target market, with **3.5 GW** expected to launch during **2026 and 2027**. This is a meaningful signal for regional power markets, land acquisition, and interconnection bottlenecks, because it suggests that the next wave of supply is still heavily concentrated in the same constrained North American corridors rather than broadly diffusing across new geographies.[15]

A market summary referencing Gartner says **data center systems spending is expected to increase 56% in 2026**, the fastest-growing segment in the technology market. The same roundup says worldwide IT costs are projected to grow **13.5%** in 2026 to **$6.31 trillion**. For cloud and data-center operators, that combination implies strong near-term purchasing power for servers, storage, networking gear, and related infrastructure, even if broader IT spending is expanding more moderately. The signal is especially relevant for equipment suppliers and colocation providers tracking enterprise refresh cycles and AI-driven demand.[2]

CBRE reports record-low vacancy in several major markets, including **Northern Virginia at 0.3%** and **Atlanta at 1%**, even as global inventory expanded sharply in Q1 2026. CBRE says Latin America inventory grew **41.3% year over year** and North America inventory grew **33%**, while power availability and grid constraints are slowing development in established hubs such as Northern Virginia, Chicago, London, and Frankfurt. It also reports pricing of **$235 to $265 per kW/month** in Frankfurt and **$330 to $475 per kW/month** in Singapore, showing how tight supply is translating into higher rents.[3]

Synergy Research data cited in the market roundup shows global hyperscale data center capital expenditure reached **$87 billion in Q2 2026**, up **35% year over year**. The same source says Asia-Pacific was the fastest-growing region at **45% YoY**, while Europe and North America grew **28%** and **32%**, respectively. It also notes that hyperscalers are accelerating new region builds in areas with abundant power, and that liquid cooling is moving from pilot to large-scale deployment. For vendors and investors, this supports continued demand for GPUs, networking, power distribution, and thermal management systems.[6]

JLL says nearly **100 GW** of new data center capacity will be added from **2026 to 2030**, effectively doubling global capacity, with the sector requiring up to **$3 trillion** in total infrastructure investment. JLL also estimates about **$1.2 trillion** in real-estate asset value creation, roughly **$870 billion** of new debt financing, and an average global cost increase of **6%** to **$11.3 million per MW** in 2026. This is a strong signal for developers, lenders, power equipment vendors, and cooling providers because it implies sustained pricing power and very large capital deployment across the next five years.[1]

Gartner has revised its global IT spending forecast for 2026, now projecting a 14.2% annual increase to $6.37 trillion, driven primarily by massive investments in AI-related infrastructure.

Year-to-date data center construction spending through June 2026 reached $81.5 billion, surpassing the 2025 full-year total of $72.5 billion and running more than three times the 2024 full-year total.

In Q2 2026, global cloud infrastructure spending reached $143 billion, marking the 11th consecutive quarter of growth, driven primarily by the rise of generative AI.