The artificial intelligence industry is entering a new phase of development. Over the past two years, market attention has focused primarily on model capabilities and chip supply, such as the surge in demand for NVIDIA GPUs, competition for HBM (high-bandwidth memory) supply, and advancements in packaging technology. However, as AI models continue to scale, a more practical issue is emerging: Even if there are enough AI chips, is there sufficient power and infrastructure to support the operation of these computing resources?
New York State’s recent suspension of approvals for new large-scale data centers has drawn market attention in this context.
On July 14, New York Governor Kathy Hochul signed an executive order suspending state-level environmental permits for new hyperscale data centers, with the suspension lasting up to one year. This measure mainly targets large projects that have not yet completed the state environmental permitting process, focusing on data centers of approximately 50 MW and above.
This policy does not mean New York is banning data center development. Instead, it aims to reassess the impact of rapid AI data center expansion on energy, the power grid, the environment, and community infrastructure.
From an industry perspective, the message is clear: AI competition is entering the infrastructure stage.
The future of AI development will depend not only on the number of GPUs but also on the availability of power, land, network connectivity, and data center capacity.
New York Halts Data Center Approvals as AI Infrastructure Pressure Mounts
Historically, data centers have been seen as the backbone of the cloud computing industry.
Enterprise servers, web services, and cloud storage all require significant computing resources, but their overall energy demands have remained relatively stable.
AI data centers are changing this paradigm.
Traditional data centers mainly run CPU servers, while AI data centers deploy large numbers of GPU accelerators. These chips deliver greater computing power but also drive higher energy consumption and increased cooling requirements.
Especially when training large AI models, GPUs must operate under sustained high loads for extended periods. As models grow larger, data centers’ demand for power supply rises rapidly.
Previously, discussions around the AI industry focused on:
- Is there enough GPU supply?
- Is there enough HBM?
- Is there sufficient advanced packaging capacity?
- Now, the industry is asking:
- Is there enough data center space?
- Is there reliable power supply?
- Is there high-speed network connectivity?
These questions are becoming new constraints for AI expansion. New York’s suspension of large data center approvals essentially reflects local governments’ concerns about the pressure from rapid AI infrastructure growth.
Building a data center isn’t just about adding more servers. It’s a complex project involving energy, land, power grid management, and environmental considerations.
Why AI Data Centers Are Becoming Energy Consumption Giants
AI’s rapid development is reshaping global electricity demand. In the past, internet companies grew mainly through software and cloud service expansion. In the AI era, competition relies much more on physical infrastructure. A large AI data center may require tens of megawatts—or even more—of power. Compared to ordinary cloud computing centers, AI data centers have several distinct characteristics.
Higher computing density. To improve AI model training efficiency, companies typically deploy large clusters of high-performance GPUs. These devices consume much more energy per square foot than traditional servers.
Longer operating hours. AI training and inference tasks often run continuously, especially once large model services are live and processing user requests around the clock.
Greater cooling requirements. High-performance AI chips generate substantial heat, requiring more complex cooling systems, which further increases energy consumption.
Therefore, the future of AI data centers is not just a chip supply issue—it’s an energy system challenge. This is why tech giants like Microsoft, Google, Amazon, and Meta have continued to invest in data centers, power procurement, and energy partnerships in recent years.
AI industry competition is expanding from software into the energy sector.
AI Computing Power Competition Is Shifting from Chips to Infrastructure
In recent years, the market has followed a straightforward AI investment logic: AI development → need more GPUs → NVIDIA benefits.
This logic fueled rapid growth for AI chip companies. But as the industry evolves, the market is realizing that GPUs are just one component of the AI infrastructure.
A complete AI system requires multiple elements working together. Chips deliver computing power, HBM enables high-speed data access, high-speed interconnects handle information exchange between computing nodes, and data centers and energy systems ensure the whole ecosystem runs reliably. If any link is missing, AI computing power cannot be fully unleashed. For example, even if a company owns a large number of GPUs, insufficient data center power means chips can’t operate at full capacity. If network connectivity is lacking, GPUs can’t collaborate efficiently, reducing overall computing performance.
Going forward, AI infrastructure competition may resemble past semiconductor industry battles. Companies will need not only cutting-edge technology but also robust supply chains and infrastructure capabilities. This is why the market is now paying attention to high-speed interconnects, optical communications, server manufacturing, and power infrastructure companies. The bottleneck for AI is shifting from "insufficient computing power" to "how to deploy computing power at scale."
Will Slower Data Center Construction Impact AI Development?
New York’s suspension of large data center approvals raises a key question: Will AI infrastructure construction slow down as a result?
For now, this move appears to signal an industry shift toward regulation, rather than an obstacle to AI development. Building AI data centers remains a strategic priority for global tech companies. Microsoft continues to ramp up AI infrastructure investments; Google is accelerating its AI computing capacity buildout; Amazon is advancing cloud-based AI services; Meta is increasing its investment in AI data centers.
These companies all need massive computing resources to support future AI applications. At the same time, local governments and energy departments must address practical concerns.
For example: Will data centers drive up local electricity prices? Can the power grid handle the additional load? Is construction compliant with environmental standards? These issues may affect how quickly data centers come online. In the future, AI infrastructure construction may depend more heavily on energy availability.
Some regions may become new data center hubs due to abundant power resources—such as Texas, Arizona, and other areas with lower energy costs.
Which Industry Chains May Benefit from AI Infrastructure Upgrades?
The growth of AI data centers will drive expansion across several industry segments. In chips, NVIDIA remains the core supplier for AI computing, while AMD is steadily increasing its market share in AI accelerators. In networking and high-speed interconnects, as AI clusters scale up, demand for data transmission is rising rapidly.
Broadcom, leveraging its switch chips and custom ASIC expertise, holds a key position in AI network infrastructure.
Marvell participates in AI infrastructure through high-speed interconnects, optical communications, and data center network solutions.
The importance of network infrastructure is rising, compared to a sole focus on GPUs. In addition, data center operators may become major beneficiaries of the AI boom. For example, data center REITs provide server hosting space, helping tech companies quickly scale up their AI computing capabilities.
Energy companies may also see new growth opportunities. AI data centers require long-term, stable power supply, which could drive power grid upgrades, energy investments, and new electricity partnership models. The future AI industry chain may form a complete ecosystem: chips → storage → networking → data centers → power.
How Will Data Center Competition Evolve in the AI Era?
Over the next few years, data center competition may see several shifts.
Scale competition will intensify. Large AI models demand ever-increasing computing resources, making hyperscale data centers an industry trend.
Energy efficiency will become a core metric. Companies will need not only more data centers, but also lower per-unit computing costs.
Data center location will become more important. In the past, companies focused on network connectivity and proximity to users. In the future, energy supply may be a bigger factor. Regions with stable power resources and robust infrastructure will attract more AI data center investment. This means AI competition is no longer just among tech companies—it’s also between countries, regions, and their infrastructure systems.
How to Track AI Infrastructure Trends with Gate Stock Trading
As the AI industry chain expands, market focus is shifting from single AI chip companies to data centers, power, high-speed interconnects, and the semiconductor supply chain.
Gate Stock Trading covers major stock markets worldwide, enabling investors to monitor developments across different segments of the AI industry chain. From US AI chip companies to Asian semiconductor and infrastructure firms, the global AI industry is forming a more complex investment ecosystem.
Opportunities in the AI era are not just about finding the next chip company, but understanding the broader infrastructure upgrade trend.
As demand for AI computing power continues to grow, power, networking, and data centers may become the next focal points for the market.
Summary
New York’s suspension of large data center approvals does not signal a cooling of the AI industry. Instead, it reflects that AI expansion is entering a new stage. Previously, AI competition centered on models and chips; now, it is extending to power, networking, data centers, and infrastructure construction capabilities. GPUs determine AI’s computing power, but infrastructure determines whether AI can truly scale.
As AI applications continue to expand, global demand for computing power will keep rising. Companies that can overcome bottlenecks in energy, networking, and construction may become the key beneficiaries in the next phase of AI.
The next AI competition may not just be a chip war—it could be a long-term battle centered on infrastructure capabilities.
FAQs
Q1: Why did New York suspend approvals for large data centers?
The main reason is the increased pressure on power supply, grid stability, the environment, and community infrastructure caused by the rapid growth of large AI data centers.
Q2: Does the suspension of data center approvals mean AI development will slow down?
Not necessarily. The policy is more about driving industry regulation than restricting AI technology development.
Q3: Why do AI data centers require so much power?
Because AI model training and inference need large numbers of GPUs running for long periods, and high-performance chips require more complex cooling systems.
Q4: Besides GPUs, what other AI infrastructure areas are worth attention?
High-speed interconnects, HBM, data centers, power supply, and energy infrastructure are all vital components of the AI industry.
Q5: Will the core of AI competition change in the future?
Yes. In the future, competition will depend not only on model and chip capabilities, but also on a company’s ability to build a complete AI infrastructure.
