Remote compute is cheaper but for whom?#
For many people, “the cloud” is magic: you click, and something runs somewhere far away, invisibly. But that “somewhere” is a real place: a data center. Think of using a VPS or SSHing into a remote lab. It seems effortless. But behind that click is a chain of investments: hardware, power, cooling, backup systems, fiber networks, and staff. These facilities have real costs, heat, electricity, water consumption and many of those benefits rarely benefit the local population. In fact, much of the value flows elsewhere and and much of the burden stays locally. In the next decade there will be a massive surge of AI data centers and increased costs for people who live near them.
Most of that capital expenditure (CapEx) is not spent locally. The GPUs come from NVIDIA, the CPUs from Intel or AMD, networking gear from global vendors. The money for those flows to a few large firms, often in other countries or regions. If a government offers land on cheap long leases, or power at a discount, that’s often the main “local incentive” they give. But that incentive is largely captured by the company operating the data center, not by the average citizen.So when someone says “the data center will boost local GDP,” you need to ask: how much of the money really circulates locally? How many local people get well-paid, sustained jobs? And how much of the negative externalities fall on local people?
Overpromises and employment illusion#
A common promise is that a data center brings many jobs. That’s only true during the construction phase. Once built, modern centers need surprisingly little staff. According to industry sources, for very large data centers maintain only “several dozen” permanent onsite workers. (datacenterknowledge.com ) In some anecdotal accounts, after build-out, a facility may have just 4 to 10 staff handling maintenance, security, and occasional hardware swaps. Others estimate that a 12 MW data center might require about 20 to 22 operational staff. (optrium.co.uk ) That’s not negligible, but it’s small relative to the scale of power, cooling, and infrastructure it absorbs. So the “jobs” story is often overstated when pitched to local governments the long-term, high-impact jobs tend to cluster upstream (hardware, software, design) or elsewhere. Meanwhile, the local economy may only see small gains: extra demand for security, facilities maintenance, property taxes, cooling services, etc. And those gains are often dwarfed by the subsidies, tax breaks, and infrastructure costs the operator enjoys.
Yet many locality-level studies warn that promises are oversold. For example: data centers often absorb a lot of energy yet create only modest permanent jobs. (chmura.com ) The advanced “Cloud Next Door” study shows how communities near dense data center clusters (e.g. Northern Virginia) face environmental, infrastructural, and socioeconomic stress, while benefits accrue to remote corporate actors. (arxiv.org )
Costs: power, heat, and environmental externalities#
Data centers consume vast amounts of electricity. As demand for AI computation grows, global capital needs for compute infrastructure are projected to exceed $6.7 trillion by 2030. (mckinsey.com ) Much of that cost is in power, cooling, and redundancy systems (backup generators, UPS, cooling loops). (encoradvisors.com )
Where that electricity comes from can matter enormously. If it’s from fossil fuels, the carbon emissions burden is local (or regional). The waste heat must be expelled. Some datacenters try to reuse heat or locate in cold climates but that only works in special settings. More often, heat is dumped, and local cooling and power systems feel higher demand. As energy consumption rises, electricity prices in heavily loaded grids tend to increase. In some regions with dense data center presence, utilities and regulators worry that “everyone else pays for their expansion.” (en.wikipedia.org ) Also, water use (for cooling) and infrastructure strain (roads, power lines) may impose costs on the community.
As a summary the costs are as below:
- Electricity Costs
- Locals worry that increased power demand from massive data centers could drive up regional electricity prices. Some fear they’ll effectively subsidize corporate energy use through higher utility bills.
- Property Values & Affordability
- Large-scale industrial projects may inflate property values or rents, potentially pricing out middle- and lower-income residents.
- Resource Consumption
- Data centers use substantial electricity and water, stressing local grids and water systems.
- Pollution & Environmental Impact
- Noise, diesel emissions from backup generators, and construction-related disturbances can affect nearby neighborhoods.
Conclusion: the real cost, and policy implications#
Data centers and inference hubs are indispensable to modern digital infrastructure. But the myth that they are a miracle local economic engine is overstated. Much of the value hardware margins, software IP, high-end engineering accrues elsewhere. Locals often inherit the burden: higher electricity demand, heat, environmental stress, and limited long-term jobs. If a government gives away cheap land or decades-long leases, it should demand stronger local return: binding requirements for local hiring, power cost sharing, heat reuse schemes, infrastructure cost offsets, and environmental safeguards. Local governments (city councils and town halls) across the country are holding meetings on data center buildouts, zoning approvals, and infrastructure development. These meetings are often open to the public, and investigative teams or researchers are attending them to understand community impacts firsthand. But in many third world countries, these safe guards are completely absent.