SPECIAL FEATURE | AI data center boom: Jobs machine or energy sinkhole—and should Asia follow?
For policymakers in Asia, this expansion is often framed as a model to emulate. Data centers promise investment, digital credibility and proximity to the AI economy.
Shot of Asian IT Specialist Using Laptop in Data Center Full of Rack Servers. Concept of High Speed Internet Connections with Blue Visualization Projection of Binary Data Transfer (Shot of Asian IT Specialist Using Laptop in Data Center Full of Rack S
The United States is building AI data centers at a scale that now resembles heavy industry rather than digital infrastructure. Hyperscale campuses consuming as much electricity as small cities are spreading across Virginia, Texas and the Midwest, driven by artificial intelligence workloads that demand uninterrupted, high-density compute.
For policymakers in Asia, this expansion is often framed as a model to emulate. Data centers promise investment, digital credibility and proximity to the AI economy. But the U.S. experience, when examined closely, offers a cautionary tale—especially for energy-constrained, climate-vulnerable and labor-sensitive economies like the Philippines.
This is not a question of being “pro” or “anti” AI. It is a question of whether Asia copies a model optimized for American grid capacity, capital abundance and corporate scale—or designs one grounded in regional realities.
In the U.S., data centers already consume about 4 percent of total electricity, with credible projections pointing to a doubling by the end of the decade as AI inference and training expand. Utilities are racing to build new substations, transmission lines and generation capacity, often fast-tracked to meet hyperscaler demand.
The economic upside is real but narrow. Construction phases generate short-term employment for engineers, electricians and contractors. Equipment suppliers benefit. Local tax bases see an initial bump. But once operational, AI data centers run lean. Automation and remote monitoring mean that facilities drawing hundreds of megawatts often employ only dozens of permanent staff.
This is where the first myth collapses: data centers are not long-term job engines. They are capital-intensive, not labor-intensive. They reward specialized skills, not broad employment.
At the same time, the AI systems housed in these centers are explicitly designed to automate white-collar work—customer service, content moderation, coding assistance and back-office functions. The net employment impact is therefore ambiguous at best and negative at worst. Even in the U.S., the question is no longer “How many jobs do data centers create?” but “Who bears the displacement costs enabled by the compute they host?”
In the U.S., the data center boom is colliding with grid constraints and residential electricity prices. In Asia, the collision would be more severe.
The Philippines, for example, already struggles with high power costs and fragile grid resilience. The national grid, operated by National Grid Corp. of the Philippines, remains vulnerable to outages, weather disruptions and supply-demand imbalances. Large-scale AI data centers would not merely be another industrial load; they would become priority customers whose demand reshapes grid planning.
This creates a structural risk. When utilities upgrade infrastructure to serve hyperscale clients, the costs are often socialized. Households and small businesses end up paying higher rates, while data center operators negotiate long-term, preferential power contracts. This dynamic is already visible in parts of the U.S. and would be even more politically volatile in economies where electricity prices are already among the highest in the region.
Water use compounds the problem. Advanced cooling systems still require large volumes of water, even with efficiency gains. In Southeast Asia, where urban water stress is rising and climate variability is intensifying, allocating millions of liters per day to data centers raises unavoidable equity questions.
Singapore’s decision to pause new data center approvals for several years is instructive. It was not anti-tech posturing, but a recognition that energy efficiency, water use and carbon intensity had to be addressed before expansion could continue. That pause alone signals how constrained even the most infrastructure-ready ASEAN economies really are.
Proponents frequently argue that data centers create “future-ready” jobs. This claim deserves scrutiny.
First, the number of permanent jobs per megawatt is extremely low. Second, the required skills—power engineering, advanced HVAC, data center operations—are narrow and globalized. Without deliberate local training pipelines, these roles are often filled by a small pool of specialists who move between projects and countries.
Third, and most critically, AI-enabled automation threatens far more jobs than data centers create. In the Philippine context, this includes business process outsourcing, shared services, content moderation and entry-level tech roles—sectors that have absorbed millions of workers over the past two decades.
If AI data centers accelerate automation while contributing little to employment absorption, the long-term labor equation turns negative. This is not speculative. It is already visible in how generative AI tools are being deployed across global service industries.
Asia should not reject data centers outright. But it should reject the assumption that scale alone equals progress.
A smarter approach is conditional expansion. Data center approvals should be tied to hard requirements, not voluntary pledges. These include guaranteed renewable energy integration that does not cannibalize residential supply, transparent reporting on water use, and clear commitments to local workforce development.
For the Philippines, this also means aligning data center policy with national energy planning, not treating it as an isolated investment issue. The grid cannot be retrofitted on demand without long-term consequences. Neither can water systems or labor markets.
Crucially, Asia must ask what kind of AI economy it wants. Hosting infrastructure for foreign hyperscalers may generate rent and prestige, but it does not automatically build domestic AI capability. Without parallel investment in local research, SMEs and applied AI for agriculture, logistics and public services, data centers risk becoming extractive digital infrastructure—consuming local resources while exporting most of the value.
AI infrastructure will be built somewhere. That is inevitable. The real question is who sets the terms and who bears the costs.
The U.S. experience shows what happens when speed and corporate scale outrun public planning. Rising energy demand, modest employment gains and growing local resistance are not bugs; they are features of a model optimized for hyperscalers.
Asia, and especially the Philippines, still has a choice. It can design a data center strategy that is smaller, stricter and aligned with public interest—or it can import an American template that assumes cheap power, deep grids and labor slack that simply do not exist here.
The mistake would not be saying no to data centers. The mistake would be saying yes without conditions—and discovering too late that the promised digital future arrived with higher power bills, fewer jobs and less room to maneuver.
————————————————————————-
WATCH TECHSABADO ON OUR YOUTUBE CHANNEL:
WATCH OUR OTHER YOUTUBE CHANNELS:
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE CHANNEL.
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE CHANNEL.
