AI buildouts move into the channel

An already booming AI market is poised for further expansion as cloud providers, AI model makers and software companies continue to invest hundreds of billions of dollars in the technology.

Worldwide AI spending is expected to grow 44% year over year to $2.52 trillion in 2026, according to a January Gartner report. The analyst firm added roughly $500 billion to its September forecast, which estimated the market would settle in at just over $2 trillion this year, after shooting up nearly 50% from 2024 to 2025.

Infrastructure will account for the lion’s share of the market, consuming more than half of AI spend in 2026, just as it did last year, when investments in data centers, servers and other hardware components totaled nearly $1.8 trillion. AI services, the category where the bulk of enterprise spend resides, will remain the second largest bucket, growing 40% year over year to capture $588 million in revenue.

Just a little over three years since OpenAI unleashed ChatGPT, roughly half of all software sold will include a generative AI component as agentic automation spreads across the industry, Gartner Distinguished VP Analyst John-David Lovelock told Channel Dive.

“The speed of AI is breaking things — breaking market norms,” Lovelock said.

Despite AI’s rapid proliferation and steep growth curve, many organizations have stalled in the early stages of adoption, facing governance, security and workforce preparedness challenges.

Use of AI agents declined during the second half of 2025, according to an Accenture report. More than half of the 3,350 employees surveyed at organizations using AI experienced productivity losses tied to low quality and misleading AI outputs, Accenture found.

A gap between outcomes and expectations has opened on the agentic front. While more than 7 in 10 organizations deployed AI agents last year, only 11% of use cases moved into production, Camunda found. The process orchestration platform provider commissioned Coleman Parks to survey 1,150 senior IT decision makers in September and October 2025.

“Exercising caution with agentic AI means many organizations can’t move beyond pilots or isolated use cases,” Kurt Petersen, SVP of customer success at Camunda, said in a release accompanying the report.

Adoption efforts have also clashed against a lack of tangible returns on AI investments.

“As far as expectations for AI go, they won't get lower than they are this year,” Lovelock said.

Generative AI traversed its peak of inflated expectations stage in Gartner’s hype cycle over the last two years and is now mired in a trough of disillusionment, Lovelock noted in the report.

Rather than investing in “moonshot projects” with questionable ROI potential, enterprises will primarily purchase AI through their incumbent software provider in the months ahead, he said.

Spending on AI software, a category that includes generative AI chatbots and autonomous agents, is expected to increase nearly 60% year over year to $452 billion in 2026.

An acquisitive environment

Rapid growth coupled with subdued enterprise enthusiasm feeds an environment conducive to consolidation, according to Lovelock.

“When these two things coincide, it generally means that point solution providers get acquired by solution providers, solution providers get acquired by suite providers, suite providers get acquired by platform providers and then platform providers start to look for better partnership arrangements,” Lovelock said.

On the cusp of the new year, Microsoft added agentic AI data engineering platform Osmos to its cloud portfolio, Meta purchased AI agent startup Manus and Snowflake signed an agreement to acquire AI observability platform Observe.

GPU giant Nvidia entered into a non-exclusive licensing agreement with AI inferencing chipmaker Groq to secure access to the startup’s technology and talent. The deal cost Nvidia roughly $20 billion, according to a CNBC report.

The transaction was indicative of where the market and the money is headed, according to Lovelock.

“We're going to see companies being bought for technology that's pre-revenue,” said Lovelock. “We're going to see companies being bought just for the customers that they're already in front of and just for the talent that they have.”

As Nvidia’s quarterly revenue neared $60 billion in September, the company said it would invest up to $100 billion in OpenAI as part of an AI compute partnership. The company poured $5 billion into CPU provider Intel just weeks before the OpenAI deal was announced.

The size, scope and speed of the spending stoked fears of a potential AI bubble in October 2025, as economists saw investments in the technology outpace near-term revenues.

Lovelock is more sanguine about AI’s revenue-generating potential, although he does anticipate some vendor volatility this year.

“I can't answer if there's a stock market problem, but I do know that there's not a problem with the revenue that's going to flow into AI,” Lovelock said. 

There are, however, more AI providers than are likely to survive into 2027, when Gartner expects global AI spending to grow another 30% and surpass $3 trillion.

“This year we’ll have a consolidation, and we’ll have companies that don't make it — that’s a normal part of a new technology’s lifecycle,” Lovelock said. “The market won’t support the hundreds of players that there are now. We’ve had a thousand flowers bloom and now it’s time to prune the garden.”

Channel partners and telcos are fielding demand for higher bandwidth connections to power proliferating cloud- and AI-based applications in the enterprise.

Telco carriers have preached over the last two years that AI could lead to a renaissance in margins as enterprises demand faster, lower-latency connectivity. Data from research firm Omdia, a Channel Dive sister company, and anecdotes from partners lend credence to those claims. Gigabits per second is growing for the average business retail circuit, in some cases pushing toward 100 Gbps — a rate usually reserved for data center connections.

An Omdia survey of 419 large enterprises found that 18% have more than ten 100 Gbps ports across their offices and data center exchanges, and that the ratio will move to 35% in one year’s time. 

Avant Communications just processed a customer order for four 400 Gbps connections into a data center site.

“That is just the start, as they have another 12-plus locations to do that around the world,” Chip Hoisington, Avant VP of engineering connectivity, colo and wireless, said in an email. “In fact, their ask was for 1.6 Tbps connections for each site, but that does not even exist yet.”

Hoisington said 800 Gbps will soon become widely available as optical technology providers Ciena and Nokia-owned Infinera complete trials of their new equipment.

“I know they are testing those electronics now, and we are likely to have 1.2 or 1.6 Tbps per connection available in the next 24 months,” Hoisington said. “It is interesting that we can stack multiple 400 gig connections just like we did in the old days with Double T1s and Double T3s.”

Retail bandwidth growth

The bandwidth frenzy is not confined to the data center.

Service providers have been providing 100 Gbps since the 2010s, and hyperscalers have been buying it since 2019, but 100 Gbps is hitting the mainstream in B2B retail, Brian Washburn, Omdia Chief Analyst of B2B telco solutions, wrote in a report.

“The technology is reaching large enterprises to upgrade Ethernet ports in the data center, and occasionally, network access. Large enterprises may not be close to filling 100 Gbps of traffic yet. But many see rapid growth in data center and cloud connect traffic,” Washburn said.

Trust Technology Consultants recently sold a 5 Gbps circuit to one of its clients in northwest Louisiana. 

“For our region, that's bonkers. That's the most anybody's ever bought,” Co-owner Mack Slaughter told Channel Dive.

Bandwidth is also up in terms of quantity. Slaughter said he’s seeing more tertiary circuits — often through satellite or fixed wireless providers — tacked onto the primary and secondary fiber circuits and managed through software-defined wide area networking. Slaughter said businesses can’t afford any downtime in their applications.

ACS Cloud Partners President Eric Asquino said AI is the chief driver of increased bandwidth requirements. Businesses are adding cloud-heavy, AI-heavy applications at their offices.

“At this point, there's pretty much nothing left on-prem, so they're connecting to everything. Their apps are getting more sophisticated, and they're using more of them,” Asquino said. “If they're doing any kind of AI in their own environment, then that's going to eat up some bandwidth as well.”

End of the race to zero?

Elevated bandwidth requirements are just one part of an AI-fueled telco renaissance. The quiet hope of carriers and the partners that sell them is that prices will increase, or at least snap out of their death spiral.

“It's been a long, long, long time since we've seen an increase,” Asquino said. 

On the contrary, prices have consistently plummeted over the years. 

“The prices are just rock bottom from those guys, and it's a real danger to a lot of people,” Slaughter said.

None of the partners Channel Dive spoke to said they’ve seen an uptick, but they did see prices largely level off from their descent in 2025.

“Pricing is starting to level out and is no longer a race to the bottom. When 100 Gbps connectivity demand rose, we did see pricing on 10 Gbps connections come down,” Hoisington said. “Similarly, we are seeing 100 Gbps pricing starting to decrease as more providers come out with their 400 Gbps offerings.”

Asquino said he’s skeptical that the price per gigabit would rise, as cable companies and local exchange carriers battle for customers.

“For some reason bandwidth is the only thing that doesn't follow inflation,” he said.

Partners have been working to offset their shrinking circuit businesses by targeting other parts of the IT stack. Unified communications, cybersecurity, and cloud infrastructure are among the cross-sell plays at their disposal.

“As an agent, where can we capture more wallet share to make sure we're still paying the light bill?” Slaughter said.

Sponsored content

By VERTIV

 AI has moved from experimentation to enterprise priority. But as organizations shift from cloud-based pilots to production-scale deployments, a critical question emerges: what should the infrastructure actually look like? 

The answer isn't as simple as "cloud vs. on-premise." Enterprise AI spans an enormous range of use cases, each with different demands on compute, data control, latency, and reliability. A generative AI copilot doesn't need the same infrastructure as a real-time quality inspection system on a factory floor. Yet without a structured approach, many organizations default to convention, risking fragmented infrastructure, stranded capacity, or systems that can't scale when it matters most. 

That's the gap this report was built to fill. 

A Decision Framework Built on AI Archetypes 

Developed by Vertiv and STL Partners, a digital infrastructure research firm, this report introduces a practical decision framework that helps enterprise IT leaders move from AI strategy to infrastructure deployment with clarity and confidence. 

At the core of the framework are six AI archetypes, each defined by the workload requirements and data constraints of real-world enterprise use cases: 

1. Model Development and Experimentation: Building or adapting models using proprietary datasets 
2. Operational Retraining and Recalibration: Updating deployed models on a recurring basis 
3. Generative and Interactive: Copilots, content generation, and decision support tools 
4. Pattern Recognition and Anomaly Detection: Interpreting images, video, sensor data, or unstructured documents 
5. Predictive: Forecasting, risk assessment, and scenario modeling 
6. Real-Time Control and Autonomous Systems: AI decisions that trigger physical actions in milliseconds 

These archetypes map to four physical infrastructure models, from lightweight edge AI clusters to purpose-built, high-density GPU environments, giving organizations a clear path from workload classification to infrastructure specification. 

What you'll learn inside 

The report walks through a structured decision-making process that most enterprise IT leaders can complete in under an hour: 

• How to calculate deployment scale across three maturity stages, from pilot to enterprise-wide platform 
• How to score workloads against four critical constraints: workload criticality, data sensitivity, latency, and data gravity 
• How to classify use cases into the archetype that best fits their operational profile 
• How to determine the right deployment model, whether cloud-centric, hybrid, enterprise-owned, or edge 
• How to translate those decisions into specific physical infrastructure requirements, including power, cooling, rack density, and connectivity 

The framework also addresses a reality many organizations underestimate: facility readiness. Power availability, cooling capacity, and network infrastructure often become the gating factor in AI deployment timelines. The report provides guidance on assessing facility constraints before they become bottlenecks. 

Why this matters now 

Enterprise AI is scaling fast. By early 2026, Deloitte found that 37% of organizations were consuming 1 to 10 billion tokens per month, while 30% exceeded 10 billion. As adoption grows, token usage scales nonlinearly, making infrastructure planning and cost forecasting increasingly complex. 

The report also reveals a critical economic inflection point: when peak demand increasingly resembles steady demand, managing owned infrastructure often becomes more economical than consumption-based cloud pricing. Organizations that take a reactive, case-by-case approach risk cost overruns, governance exposure, and systems that can't keep pace with demand. The framework in this report supports a portfolio approach, enabling organizations to right-size infrastructure to each category of workload and scale deliberately as AI matures across the business. 

A phased roadmap for action 

Beyond the framework itself, the report outlines a four-phase deployment roadmap: Assess, Select, Deploy, and Validate. This structured sequence helps organizations make the right decisions in the right order, from initial workload classification through production deployment and ongoing optimization. 

Whether you're planning your first on-premise AI deployment or scaling an existing hybrid environment, this report provides the analytical foundation to move forward with confidence. 

Download the full report to access the complete decision framework, detailed archetype profiles, infrastructure model specifications, and Vertiv's proven reference designs.

Amazon, Microsoft and Google together shelled out $110 billion in capital expenditures as the buildout to handle high-capacity AI workloads continued during the three months ended March 31.

AWS parent company Amazon led the pack with a reported $43.2 billion in capex, primarily to meet customer demand for generative AI, SVP and CFO Brian Olsavsky told investors during a Q1 2026 earnings call.

“The faster AWS grows, the more short-term capex we will spend,” the company’s President, CEO and Director Andy Jassy said, noting the investments covered “cash for land, power, buildings, chips, servers and networking gear … We've been through this cycle with the first big AWS growth wave and like the results.”

Microsoft’s quarterly capex was $31.9 billion, two-thirds of which went toward GPUs, CPUS and other short-lived assets, EVP and CFO Amy Hood said during a Q3 2026 earnings call. With its Azure and other cloud services segment revenues growing 40% year over year, the company plans to increase capital investments to $40 billion next quarter.

Today’s data center investments are earmarked for tomorrow’s workloads, as Microsoft and its competitors continue to add generative AI features to core offerings.

“We want to make sure we are getting the capex to get the capacity in time for those increases in usage,” Microsoft Chairman and CEO Satya Nadella said. “If you think about even Agent Mode in Excel, it sort of kind of didn't work until it started working. And that's just because the model showed up … you have to be ready for those opportunities.”

Google SVP and CFO Anat Ashkenazi gave investors a breakdown of the company’s $35.7 billion capex during the company’s Q1 2026 earnings call. The overwhelming majority went toward technical infrastructure to support AI across the company — 60% for servers and 40% for data centers and networking equipment.

Cloud’s immense appetite for data center resources is far from sated.

Google upped its projected capex for the year by at least $5 billion to $190 billion, inclusive of costs related to the acquisition of data center and energy infrastructure provider Intersect, which closed in March 2026.

The magnitude of the capex blitz is, to date, supported by substantial growth in the global market for cloud services, according to Synergy Research Group. Enterprise cloud spending increased by more than $35 billion year over year to $129 billion during the first three months of 2026, marking the ninth consecutive quarter of the growth rate has increased, the analyst firm reported.

“The numbers are huge, but so is the financial strength of these companies,” SRG Chief Analyst John Dinsdale told Channel Dive. “Our forecasts are clear that cloud revenues will keep growing faster than capex.”

An ongoing run on semiconductors has far-reaching implications across the tech sector and channel, straining component supply chains and driving up costs for PCs, servers and related network hardware.

More than half of the nearly $10 billion Microsoft plans to add to its capex next quarter will be eaten up by higher component prices and finance leases, Hood said. The company's sales of Windows devices exceeded expectations as its OEM and channel partners responded to inflationary signals by stockpiling memory chips.

Gartner expects memory costs to more than double and storage prices to increase 243% this year. Constraints aren’t likely to ease until 2027, the analyst firm said.

As AI moves deeper into production, channel partners may need to shift the customer conversation from which model to use to where inference should actually run.

That’s according to an Akamai report, which ties AI inference performance directly to business results — including customer satisfaction, cost to serve and revenue per visitor. The edge computing provider surveyed 200 architects, engineers and technical leaders involved in AI inference decisions in March.

Enterprises have a wide range of AI use cases in production, including personalization, customer support, employee copilots and document processing, Akamai found. But the infrastructure supporting those workloads is starting to show strain. More than 4 in 5 respondents said their most important AI use cases require end-to-end response times of 500 milliseconds or less. For nearly two-thirds, the requirement was 250 milliseconds or less.

The performance bar opens a familiar, but potentially bigger-than-ever, door for partners: customers need help making AI work reliably in the real world, not just proving that a use case can run.

“AI infrastructure is undergoing the same shift the web itself went through, from centralized to distributed,” Ari Weil, Akamai cloud evangelist, told Channel Dive. “The industry has treated inference as a hyperscaler workload, but that model is breaking under real-world demand.”

As a result, Weil said, customers are running into “latency walls, egress cost surprises, sovereignty constraints and capacity limits that no amount of operational ingenuity can engineer around indefinitely.”

Partners who recognize this early on, he added, “have a meaningful window to lead their customers through it.”

Many enterprises know that proximity matters, but their architecture has not kept pace. Three in 5 of the respondents viewed proximity to end users and decision points as important or critical. Yet, less than half currently run inference in a single centralized cloud region. Similarly, 45% of those surveyed expect to stick with that centralized cloud approach for their most important use cases over the next one to two years.

Infrastructure gaps give partners a chance to take a strategic advisory role around AI architecture. The opportunity is not only to guide AI application deployments, but to help customers decide which workloads need lower latency, which can remain centralized, how traffic should be routed and how performance, cost and governance should be managed.

Adopting an advisory role will require partners to expand their fluency beyond AI products and hyperscaler offerings. Weil pointed to “AI-optimized infrastructure” evolving toward distributed inference, citing Akamai’s Inference Cloud and Nvidia’s AI Grid as examples of the architectural pattern partners should understand.

“The pattern matters: AI-tuned compute deployed close to where users, agents, applications and data actually live,” Weil said.

Many enterprises are compensating for infrastructure limitations with operational workarounds. Nearly two-thirds of respondents classified traffic steering as very important or critical. More than 70% said they need to enact a rollback within an hour, and nearly half said 15 minutes is the acceptable window.

Those fixes can keep workloads moving, but they add burden.

Akamai found that integration complexity tops the list of reasons AI has not scaled further, followed by infrastructure gaps, security and compliance risk and data quality and availability. Unclear ROI ranked far lower.

Cost visibility remains another pressure point. Here, Akamai found that 77% of organizations lack consistent unit-level economics tracking for inference. They can’t easily tell whether AI workloads are scaling efficiently or are becoming more expensive to operate.

Taken together, the data give partners yet another opportunity. Customers who don’t need to be educated about AI’s capabilities often need help operating applications without adverse latency, risk or cost impacts.

Weil said that creates three imperatives for partners:

• Broaden the conversation beyond “which model” and “which hyperscaler” to include where inference runs
• Build practices that connect AI workloads with network, security and edge competencies
• Help customers move away from short-term workarounds toward distributed architectures built for high-level AI experiences at a global scale

“What I'd most like to see is partners treating this as a learning moment, not a positioning moment,” Weil said. “The shift to distributed AI will reshape how customers buy and deploy compute over the next several years. Partners who invest now in understanding the architecture … will be the ones their customers turn to when the workarounds finally stop working.”

An AI-fueled data center buildout is driving unprecedented demand for energy. By 2050, data centers will account for up to one-third of commercial building electricity use in the U.S. As the decade closes, global electricity consumption by data centers is expected to grow 15% per year, more than four times faster than all other sectors.

The electrical equipment market is experiencing a concurrent boom. By 2030, the sector is projected to surge from $20 billion to $65 billion, according to research and consulting firm Wood Mackenzie.

Data centers will likely define the future of the electrical equipment industry, just as AI is reshaping compute architectures.

Just six years ago, data centers captured under 2% of the U.S. electrical equipment market, per the report. By 2030, Wood Mackenzie researchers estimate the massive, energy-intensive facilities will seize up to 40% of total demand.

“We're going to see this rapid influx of data center capacity coming online over the next two years or so,” Ben Boucher, senior analyst at Wood Mackenzie, told Channel Dive. “That will put pressure on the underlying supply chains, but also on power infrastructure, electricity generation and transmission and distribution as well. It's a perfect storm of constraints predominantly driven by data centers.”

Record market expansion threatens to trigger a supply chain crisis.

Transformers, which adapt power from the utility grid for servers and data center infrastructure, are a critical bottleneck for the electrical equipment industry. Lead times have stretched up to three years for large-scale power transformers, Boucher said, and prices have skyrocketed by 50% for large transformers and over 100% for smaller distribution transformers used for everyday consumption.

The scale of electrical equipment required to meet the planned data center boom is staggering, Boucher added. Transformer demand will multiply sixfold by 2030, while panel board and switch manufacturing will also need to ramp up exponentially.

Labor shortages and capital requirements are other looming constraints. Boucher said that fewer young people are entering manufacturing, creating a skilled worker deficit. New manufacturing facilities also require billions of dollars in investment and several years to become operational.

At the same time, some manufacturers are concerned whether demand for data center gear will endure.

“Some of the original equipment manufacturers have concerns whether the demand levels are going to be sustainable,” Boucher said. “Is this data center boom going to be a short-lived bubble?”

Electrical components distributor Arrow Electronics saw its global components segment revenues increase 39% year-over-year to $6.6 billion in this year’s first quarter, Jon Watters, a spokesperson for Arrow, told Channel Dive over email. 

Watters tied the growth to boosts in “supply chain activity to furnish electrical equipment supporting AI data center buildout and expansion.” 

For channel partners, the traditional commercial and utility customer base will likely continue seeing growth as aging infrastructure modernizes, Boucher said. The challenge will be navigating supply constraints and potential price pressures as data centers compete for the same limited equipment pool.

Available power for AI infrastructure hit record lows even as global data centers brought more capacity online in Q1 2026. Total inventory surged 33% in North America year-over-year, according to CBRE research.

"A data center without power is just a physical structure,” Gordon Dolven, head of data center research at CBRE, told Channel Dive.”The servers won't be energized, they won't be operational. We used to say in the olden days that it’s just a cold refrigerator.”

The decline in accessible power, measured by vacancy rates, is extending construction timelines and complicating site selection for data center projects across the U.S. In Northern Virginia’s “data center alley,” vacancy rates dropped to 0.3%, effectively selling out the world’s largest data center market. In Atlanta, another key market, vacancy rates sit at just 1%.

Strong demand for AI infrastructure from startups, neocloud providers and hyperscalers that eat up grid availability is driving the record-low vacancy rates.

“The robustness of demand continues to astound me,” Dolven said. “We have all-time low vacancies driving all-time high prices, even with record increases in inventory. Demand continues to be incredibly resilient across the globe.”

The challenge for hyperscalers building out AI infrastructure is multifold, Dolven said. Long utility wait times, electrical equipment bottlenecks, community opposition to construction and, most of all, ongoing power scarcity are pushing up project costs and timelines.

Five years ago, a data center proposal might have entailed five megawatts of IT power. Now, massive facilities built to support mounting AI reliance could require 500 megawatts and 500 acres of land, according to Dolven.

“It's a critical component of every development to put the power procurement first,” Dolven said. “I would say we remain in a power-constrained market.”

As the power supply tightens, landlords that build and lease data center facilities to tenants, including hyperscalers, are in a strong position at the negotiation table, Dolven said.

The result is booming rental rates. In the U.S., rental pricing is increasing steadily — with the steepest gain reaching almost 15% year-over-year in Chicago at prices ranging from $200-$230 per kilowatt-hours per month, according to the report.

Data center infrastructure companies like Equinix are expanding their colocation spaces to provide rentable facilities to multiple customers, including partners working on behalf of enterprises.

For enterprises and partners seeking data center space in tight markets, long-term planning is essential, Dolven said.

“We’re seeing occupiers and tenants across every small, medium and large business having to talk about their IT strategy with their CTO or their CIO, 24 to 36 months in advance of a lease expiring at a colocation facility,” he said.

The power crunch is also driving another challenge: community pushback. As data centers become bigger and more energy-intensive, residents near the sprawling projects are raising concerns about electricity costs, environmental impact and noise and visual pollution. In some markets, city councils and planning committees are a major roadblock for data center development.

In the age of AI, hyperscalers, enterprises and partners alike will have to weigh growth alongside power availability, pricing constraints and the communities they serve.

As AI demand skyrockets, equipment shortages and limited electric grid availability have constricted the data center buildout. Despite record construction activity, the North American data center market is grappling with finite power infrastructure capacity, according to CBRE's H1 2026 North American Data Center Trends report.

New data center construction rose by almost 25% in the first half of 2026 across major U.S. markets, including Northern Virginia, Atlanta and Dallas-Fort Worth. Total power grew by a third year-over-year — but vacancy rates, representing the percentage of power capacity available for lease, fell to a record low 1.4%, per the report.

“The scale at which we're building today is significantly larger than what it was five years ago, and completely different than what it was ten or 15 years ago,” Pat Lynch, global executive managing director for the data solutions group at CBRE, told Channel Dive. “It is unlike any other industry in my lifetime, where you can triple supply, yet still have vacancy rates going down.”

Amid declining vacancy rates, data center landlords like Equinix and Digital Realty are leasing space faster than developers can build it with necessary power infrastructure. Net absorption, which measures change in leased space, increased by almost 12% in the first half of 2026, CBRE found.

Rental rates are also rising, reflecting how limited power availability is boosting prices for major tech giants and smaller enterprises looking to lease data center space for AI operations, according to Gordon Dolven, head of Americas data center research at CBRE.

“Rental pricing and vacancy rates have been perfectly inversely correlated in our report for the last few years,” he said.

For their clients looking to acquire data center space, Dolven recommends looking at leasing options three years ahead of schedule and being flexible about location. Data center space in Northern Virginia’s “Data Center Alley” may be unavailable, but Atlanta and Columbus, Ohio could have openings.

Northern Virginia remains the largest market by inventory, but Atlanta has emerged as the top market for total construction in North America. Data center development in the region surged over 50% year-over-year in the first half of 2026, surpassing Data Center Alley for the first time, according to the report.

“You’re seeing Reno, Nevada compete with Silicon Valley,” Dolven said. “You're seeing Atlanta, Georgia compete with Northern Virginia. You're seeing so many new markets pop up because they're trying to seek out that available power capacity.” 

Amid the massive data center buildout, concerns about overbuilding, land use and water availability have plagued the tech industry, spawning anti-data center sentiment among activists, communities and politicians.

CBRE is AI-optimistic. “We’re applying our industry to AI on a daily, hourly basis,” Lynch said. “If we think back on history, there were thoughts that the automobile would ruin industry. There are so many doom-and-gloom periods if you look back, and these technologies redirected employment in some ways, but they also made broad steps in improving our livelihood. I think AI will be the same thing.”

Still, having lived through the telecom bubble of the early 2000s, Lynch said limitations to the data center buildout might help investors.

“I could make the argument that limited power, which has been a bit of a throttle [to the data center market], could potentially limit an overbuild situation,” he said.

Enterprises looking to scale AI are grappling with power and resource constraints, and Hammerhead AI has a solution that it wants channel partners to deploy. 

The software platform provider specializes in converting stranded or underutilized data center power into AI-ready capacity through its Orchestrated RL Control Agents software. In September 2026, the company tapped global IT distributor TD Synnex to bring the technology to the channel. 

Many data centers use less energy from the grid than is allocated. ORCA coordinates electrical, cooling and compute to convert existing electrical headroom into deployable capacity.

“We are unlocking the power that makes compute available,” Rahul Kar, CEO and founder of Hammerhead AI, told Channel Dive. “Think of us as power as a SKU, and then give it to anyone who's trying to build data centers, and then all the rest of the things can plug into that SKU.”

TD Synnex will provide solutions, integration and resale capability for ORCA software, leading go-to-market efforts funneled through its network of more than 150,000 resellers worldwide, according to a press release.

Limited power availability has emerged as the most significant bottleneck for the massive global data center buildout. Available capacity in North American data centers plunged to a record-low vacancy rate during the first half of the 2026, dropping to 1.4% as surging AI demand strained the grid. 

The demand for powered space to support AI expansion is expected to continue, according to a Goldman Sachs analysis. The investment bank expects U.S. power needs to more than double by 2027.

Hyperscalers and large-scale vendors such as Nvidia, Microsoft and Google have ramped up capital investments to overcome resource challenges, funneling spending into power supplies, land use, cooling solutions, electrical equipment and chip production. 

Hammerhead is banking on TD Synnex’s massive partner ecosystem to power AI operations for enterprises and neoclouds.

“What we are doing is becoming the front end for customers looking for compute and tokens,” Kar said. “We are giving the power to TD Synnex essentially for all their customers, so it becomes a very easy channel play for us to distribute the power, and then they take care of everything else related to the ecosystem: bringing together the parties, the hardware, the MSPs and then ongoing services alongside that.”

Unlocking AI power

TD Synnex’s partnership with Hammerhead is a significant move in addressing AI infrastructure challenges, according to Alastair Edwards, chief analyst at Omdia, a Channel Dive sister company.

Edwards said that by making technologies like Hammerhead available to resellers, TD Synnex is giving partners a way to capitalize on the data center boom. 

“Enterprise AI is a very complex stack that you're having to build, and I think TD has the opportunity to bring all the relevant vendors together that partners may not have access to today,” he said. “So the distributors who are going to win are going to be the ones that bring these new technologies in and expose them to their channel partners to help them succeed.”

Guy Wood, senior director analyst at Gartner, said the Hammerhead collaboration means TD Synnex is betting big on data center expansion. 

“There’s this whole concept of optimizing various aspects of a data center run to make it as cost-effective, as power-effective, as compute-effective as it can be,” Wood said. “Partnering with companies like Hammerhead, who are able to package optimization software, suggests TD Synnex sees massive growth in the number of data centers, and not just a few big ones run by AWS or Meta, but a massive growth in data center expansions, requiring pre-packaged types of software to optimize the build and run process.”

Execution will determine success for TD Synnex’s collaboration with Hammerhead, according to Gartner Senior Director Analyst Anne McClelland. 

“My immediate thought goes to what will be the training and enablement to get the resellers, the individual sellers, comfortable with conversations around land use, power management or cooling management,” McClelland said. “These are not conversations that these partners and salespeople are necessarily having today. They're talking about different kinds of feeds and speeds. So this is new turf.”

McClelland added that TD Synnex will have to focus on preparing partners to sell Hammerhead technology. Resellers will need technical expertise to have detailed infrastructure conversations with buyers.

Market pressures

Hammerhead’s technology is one approach to address power constraints. The problem's scale extends far beyond a single solution.

The IT market is growing at its fastest pace on record, according to Gartner research. The firm expects more than 80% of enterprises to deploy industry-specific AI agents by 2030, up from just 10% in 2025. In 2026, Gartner forecasted global AI infrastructure spending to reach nearly $1.5 trillion.   

The amount of power it will take to support AI is expected to spike alongside data center buildouts, according to Gartner research. Then there’s the matter of securing and governing all AI activities, and extracting return on investment. Many channel partners are feeling overwhelmed, McClelland said.

“The rapid buildout of AI is a bit of a cacophony, from the standpoint of what traditional channel players are feeling right now,” McClelland said. “The channel can be a bit of a laggard community when it comes to picking up new technology, new trends, just because of their investment capability. Because they work on very slim margins, they have to place their chess moves, their chess pieces, very carefully on the chess board. I believe that they're waiting to see how things will actually settle out and what their roles should be.”

Partners should cultivate skills and hire employees focused not just on cloud, networking and security, but on power management, cooling and automation, McClelland added. 

Channel's new role

Channel firms are already shifting to AI infrastructure hardware and services, according to Edwards.

In its initial stages, data center buildouts were centered around cloud giants AWS, Microsoft, and Google. Nvidia and a handful of other large hardware vendors reaped the rewards through direct sales of AI servers, storage and networking. 

A shift started in 2026, according to Edwards. 

“We've seen a big shift towards the channel, and there's a number of reasons for that,” Edwards said. “The buildout has expanded beyond just the big hyperscalers into regional neocloud, enterprise and public sector customers. We've seen this kind of skewing of growth in the channel towards those big data center deals, and investment that has been focused in the U.S. is broadening out into the rest of the world.” 

Some channel firms, including Computer Center, WWT, SHI and Ahead, have already won neocloud and enterprise data center deals, Edwards said.

The impetus to build is mounting. Companies that started with AI pilots in the public cloud face high costs and lock-in contracts as they scale projects, driving many to weigh the benefits of on-premises environments and colocation facilities. 

“We’ve seen tremendous growth in the channel for data center hardware, compute, networking, and storage,” Edwards said. “Distributors like TD Synnex, which have historically served channel partners, are now actually playing in those big enterprise data center buildups, as well.”