
Almost two hundred years after the famous California gold rush, we’re currently witnessing a similar phenomenon: a mad dash by businesses looking to secure AI-related resources. In the “AI gold rush,” data center space is an especially hot commodity. This dynamic is transforming the world of colocation, with both supply and demand soaring higher than ever before.
The data center land grab produces challenges, but it also presents opportunities for businesses and Sales Partners alike. On a recent SE webinar, Mike Davis, Michael Ratta, and Cloud & AI Programs Director Bhu Virdi discussed the most fascinating and impactful developments.
We’ve summarized the main points from that presentation below.
How the Exploding AI Market Is Remaking Colocation
Artificial Intelligence (AI) is the defining technology of our era, and its growth is radically changing the colocation market. This is hardly surprising given the sheer scale of AI investments. The AI infrastructure market is projected to exceed $2 trillion in hardware revenue by 2033.
The Insatiable Demand for AI Compute
AI programs, such as Large Language Models (LLMs), utilize massive amounts of computational power. Case in point: Training a model like ChatGPT-4 requires as much energy as powering 160 American homes for a year.
These needs are only increasing. OpenAI’s models saw a 300,000-fold increase in required compute power in just the past 6 years. This creates compelling market dynamics. Computational power has become a super valuable commodity, but it remains a finite resource.
How Data Centers Serve AI Workloads
AI workloads are especially intensive, requiring GPUs, TPUs, and high-speed networking (InfiniBand, 400 Gbps Ethernet, etc.). Data centers don’t typically offer these components themselves, but they do provide an ideal environment for AI workloads.
To serve the needs of AI, data centers provide:
- Physical space
- Advanced cooling capabilities
- Power
The Current State of the AI Data Center Market
There’s one word that best describes the current AI data center market: hot. Businesses are snapping up data center space even faster than the centers can be constructed.
Here are some notable characteristics of the data center market today.
Decreased Vacancy
The vacancy rate for primary data centers fell to a record-low of 2.8% in H1 2024, down from 3.3% a year earlier. This is largely due to a recent “land grab,” with companies securing more data center space both for their current needs and because they expect to require even more space in the future.
Increased Supply
In response to surging demand, data center construction is exploding nationwide – both in primary and secondary markets. In H1 2024, this construction boom produced a 10% increase in data center supply within primary markets.
But even as the number of data centers increases, vacancy is expected to remain low. Why? Because demand is so high that 80% of the data center space under construction has already been pre-leased.
Lease Pricing on the (Gradual) Rise
The cost of leasing space in North American data centers is continuing to rise, but the increase has recently slowed. This is likely due to the rapid construction of data centers across North America.
Regional Hotspots
While the rapid construction of data centers is a national phenomenon, it’s especially pronounced in certain regions. In Atlanta, for example, 2024 saw a 76% increase in data center construction. Parts of Texas experienced even faster growth, with Austin and San Antonio’s combined construction activity quadrupling YoY.
Colocation: The Nerve Center of the AI Revolution
Colocation is central to the AI revolution for a simple reason: It’s the ideal option for AI deployments.
Here are the key advantages of using a colocation data center:
- Adaptable Scalability. Data centers are being built at a massive scale – meaning organizations will have access to the additional power, space, and cooling that are necessary when scaling quickly.
- Security and Compliance. Data center operators understand how to navigate security/ compliance requirements, including PCI, HIPAA, high trust, ISO, and SOC.
- Cost-effectiveness. Most businesses lack (a) the institutional knowledge to effectively house their own AI workloads and (b) the staff to manage such a facility – making colocation a more efficient option.
- Direct Cloud On-Ramps. Data centers can act as “hubs,” allowing organizations to more easily connect with other data centers and public cloud infrastructure.
- Speed to Market. Leasing colocation space is significantly faster than building a privately-owned facility.
Thanks to these advantages, colocation represents a perfect solution for organizations in the midst of a strategic shift towards AI.
How 4 Key Verticals Are Deploying AI Workloads in Data Centers
It’s not just tech giants like OpenAI and Google that are using data centers for AI. Here’s how businesses in four main verticals are deploying AI workloads.
*Note – Data center usage is not limited to these verticals. AI is becoming standard across industries, which means a wide variety of organizations could be interested in accessing colocation facilities for AI workloads.
Healthcare
Healthcare facilities are using AI for workloads like:
- Medical Imaging Analysis
- Drug Discovery & Genomics
- Predictive Analytics for Patient Outcomes
- Real-time Patient Monitoring
- Automated Database Management
Colocation centers are ideal for these workloads because operators understand how to maintain HIPAA compliance.
Financial Services
Because of the sensitivity of the data they handle, financial services providers are especially likely to leverage data centers rather than cloud-based solutions.
Common AI workloads for financial services companies include:
- Algorithmic Trading / High-Frequency Trading (HFT)
- Fraud Detection
- Risk Modeling & Management
- Customer Engagement
- Operational Efficiency
Media & Entertainment
Media and entertainment companies, including video game companies, often use a hybrid infrastructure of cloud and edge-based computing.
These businesses might deploy the following AI workloads to data centers:
- Personalized Content Recommendations
- Content Creation & Production
- Real-time Streaming Optimization
- Gaming
Manufacturing & Industrial Automation
Manufacturing businesses, especially those that use robots and other forms of industrial optimization, often prefer deployments that are physically close to their facilities.
Common AI workloads include:
- Predictive Maintenance
- Quality Control
- Robotics & Automation
- Supply Chain Optimization
Future Challenges
As AI continues to increase in popularity, the demands on data centers will only grow more strenuous. Data centers will consume even more power, and rack densities are sure to soar.
All this output creates excessive heat, which can overpower traditional “crack units” for cooling and endanger the operation of the facility. Luckily, operators are already employing novel cooling solutions, with “liquid cooling” becoming the new industry standard.
This cooling dilemma (and its fascinating solution) reflects the dynamism of the current moment. AI is transforming the colocation landscape, producing new norms and expectations along the way.
Leveraging Colocation’s Role in the AI Gold Rush
In recent years, AI’s total CapEx has been more than $300 billion – and that total will only increase as more organizations adopt agentic AI and other AI solutions. This creates a market opportunity around colocation that you simply can’t overlook.
Make sure you’re having regular conversations with customers about where they’re deploying AI workloads. Given the advantages of colocation, they’ll likely end up wanting to put their AI workloads into data centers.
More Resources
For even more insights, watch the complete SE Webinar: Navigating the New Colocation Frontier. Connect with your Intelisys Regional SE for additional guidance.
Also, be sure to check out our full stable of colocation resources: