AI data center electricity demand is projected to nearly double to 950 TWh by 2030, forcing Microsoft, Google, Amazon and other technology companies to evaluate nuclear power, solar energy and natural gas for reliable, affordable 24/7 electricity.
The global artificial intelligence boom is creating a new challenge for technology companies: securing enough electricity to operate AI data centers without facing escalating energy costs, power shortages or delays in grid connections.
Nuclear power promises dependable, low-carbon electricity. Solar offers competitively priced renewable energy but requires storage or backup for continuous operations. Natural gas provides dispatchable electricity but exposes operators to fuel-price volatility and carbon emissions.
The financial implications are significant. For a 100 MW AI data center operating at 90 percent average utilization, every $10 per MWh increase in electricity costs adds approximately $7.9 million to annual operating expenses.
With AI infrastructure investments reaching hundreds of billions of dollars, electricity pricing and availability are becoming critical factors in data center profitability.
AI Data Center Electricity Demand to Reach 950 TWh by 2030
The International Energy Agency’s 2026 energy and AI analysis projects global data center electricity consumption to increase from approximately 485 terawatt-hours in 2025 to 950 TWh in 2030.
This represents an increase of nearly 96 percent within five years.
A hypothetical 100 MW data center operating at 90 percent average utilization consumes approximately 788,400 MWh annually.
Its annual electricity expenditure would vary substantially depending on the delivered electricity price.
| Delivered electricity price | Annual electricity cost |
| $40/MWh | $31.5 million |
| $60/MWh | $47.3 million |
| $80/MWh | $63.1 million |
| $100/MWh | $78.8 million |
| $120/MWh | $94.6 million |
A difference of $60/MWh could increase annual electricity expenditure by $47.3 million for a single 100 MW facility.
For operators developing multiple gigawatt-scale AI infrastructure projects, these differences could translate into billions of dollars over the operating life of their facilities.
However, the cheapest electricity generation technology does not necessarily provide the cheapest reliable electricity supply.
Nuclear Power: Reliable Electricity but High Investment Costs
Nuclear power is emerging as an important option for technology companies seeking continuous, low-carbon electricity.
Unlike solar generation, nuclear reactors can produce electricity throughout the day and night, subject to planned maintenance and occasional outages.
Existing nuclear facilities are particularly attractive because their generation infrastructure and grid connections are already established.
Google and Constellation announced a 20-year agreement in October 2026 to add 890 MW of nuclear generating capacity through upgrades at 11 existing reactors.
Constellation plans to invest more than $4.3 billion, with the first additional capacity expected in 2028.
The agreement also includes a separate 15-year arrangement covering 2,700 MW of electricity from Constellation’s existing nuclear generation fleet in the PJM market.
The Google–Constellation agreement demonstrates how technology companies are securing long-term electricity supplies while supporting investment in existing nuclear infrastructure.
However, new nuclear projects face significant challenges.
Construction requires substantial upfront investment, regulatory approvals and long development periods. Financing costs can increase sharply when projects experience delays.
Small modular reactors could eventually provide another option, but their commercial competitiveness depends on successful deployment, manufacturing scale and regulatory progress.
Nuclear’s biggest advantage is continuous low-carbon generation. Its biggest challenge is securing new capacity at an acceptable cost and within the required timeframe.
Solar Power: Low Generation Costs but Expensive 24/7 Reliability
Solar energy offers an attractive opportunity for AI data center operators seeking to reduce electricity expenditure and carbon emissions.
Solar power purchase agreements can provide predictable generation prices without direct exposure to fuel-price fluctuations.
However, solar generation declines during cloudy conditions and stops after sunset.
AI data centers require electricity continuously, making solar generation alone insufficient for round-the-clock operations.
Battery energy storage systems can shift electricity generated during the day into evening hours.
But battery capacity, storage duration, conversion losses and replacement costs affect the economics of continuous renewable electricity.
Consider a 100 MW AI data center operating at full load.
Its electricity requirements would be:
| Operating duration | Electricity required |
| 1 hour | 100 MWh |
| 4 hours | 400 MWh |
| 8 hours | 800 MWh |
| 12 hours | 1,200 MWh |
| 24 hours | 2,400 MWh |
A 100 MW battery system designed for four hours of discharge provides 400 MWh of nominal storage capacity.
That is insufficient to independently power the facility for an entire day.
The system would require additional generation, longer-duration storage or dependable grid electricity.
The International Renewable Energy Agency’s research on 24/7 renewable electricity examines the economics of combining solar, wind and batteries to provide more dependable electricity.
Solar can significantly reduce electricity costs when favorable generation conditions, storage and complementary power sources are available.
The key distinction is between cheap solar electricity and affordable 24/7 solar-based power. They are not the same product.
Natural Gas: Reliable Power With Fuel-Price Risks
Natural gas offers another option for AI data center developers seeking electricity that can be generated when required.
Combined-cycle gas turbines can deliver efficient electricity generation, while other gas-based technologies can support backup and flexible operations.
Gas generation is particularly relevant where grid connections are delayed or renewable electricity cannot meet continuous demand.
However, operating costs depend heavily on natural gas prices.
Consider a power plant consuming seven MMBtu of natural gas to generate one MWh of electricity.
| Natural gas price | Fuel cost per MWh |
| $3/MMBtu | $21 |
| $4/MMBtu | $28 |
| $6/MMBtu | $42 |
| $8/MMBtu | $56 |
| $10/MMBtu | $70 |
These figures represent fuel expenses only. They exclude construction, financing, maintenance and other operating costs.
An increase in natural gas prices from $4 to $8 per MMBtu doubles the illustrative fuel expense from $28 to $56/MWh.
Gas also faces carbon-emission concerns, methane leakage risks, environmental permitting requirements and dependence on reliable fuel infrastructure.
Gas-fired plants are dispatchable but not immune to equipment failures or fuel-supply interruptions.
Natural gas can be competitive when reliable generation is needed quickly, but long-term operating costs remain exposed to commodity markets.
Natural Gas vs Solar: Which Is Cheaper for AI Data Centers?
A March 2026 analysis by S&P Global Market Intelligence examined the economics of powering a hypothetical 627 MW data center in West Texas.
The analysis compared dedicated combined-cycle natural gas generation with solar electricity supported by ten-hour battery storage.
The results highlighted substantial differences in modeled electricity costs.
| Financial and operating metric | Natural gas | Solar plus battery |
| Modeled electricity cost | $61/MWh | $130/MWh |
| Modeled 20-year cost | $2.9 billion | $6.2 billion |
| Data center demand | 627 MW | 627 MW |
| Generation configuration | Combined-cycle gas | 2,705 MW solar |
| Battery requirement | Not applicable | 903 MW battery |
The modeled solar-plus-storage system required considerably more installed generation capacity than the data center’s electricity demand.
The estimated electricity cost for solar with batteries was approximately 113 percent higher than natural gas under the study’s assumptions.
The modeled 20-year cost difference was approximately $3.3 billion.
However, the S&P Global electricity cost comparison reflects specific assumptions about location, technology, financing and system design.
The renewable scenario also excludes potentially available tax credits.
The findings should therefore not be interpreted as proof that natural gas is cheaper than solar-plus-storage in every market.
Both configurations require careful evaluation of reliability, backup capacity and outage risks.
Nevertheless, the comparison highlights why data center developers must evaluate the total cost of dependable electricity rather than headline renewable generation tariffs.
Nuclear vs Solar vs Natural Gas: Key Advantages and Risks
| Power source | Major advantage | Major limitation | Best application |
| Nuclear | Continuous low-carbon generation | High investment and construction risk | Long-term firm electricity |
| Solar | Competitive renewable generation | Intermittent output | Reducing average energy costs |
| Solar with batteries | Renewable electricity beyond daylight hours | Storage costs and duration limits | Renewable supply with evening coverage |
| Natural gas | Dispatchable electricity | Fuel-price exposure and emissions | Firm generation and flexible supply |
| Grid-connected hybrid | Diversified electricity sources | Grid charges and coordination complexity | Large AI data center campuses |
There is no universal winner.
Existing nuclear generation can be attractive where long-term contracts are competitively priced.
Solar can reduce average electricity expenditure, especially in locations with strong sunlight and favorable power purchase agreements.
Natural gas can provide dependable generation where fuel infrastructure is available and emissions requirements permit its use.
Hybrid systems combining multiple electricity sources can reduce dependence on any single technology.
Grid Connection Delays Could Be a Bigger Problem Than Electricity Prices
Electricity generation costs are only one part of the challenge facing AI data center developers.
Grid connectivity is increasingly influencing project schedules and investment decisions.
The IEA has highlighted more than 2,500 GW of generation, storage and large-load projects worldwide awaiting grid connections.
Transmission capacity, substations, transformers, permitting and network upgrades can delay electricity delivery even when generation projects are available.
A data center developer may secure an attractive long-term electricity agreement but still face delays before power can reach the facility.
These delays can increase financing expenses, postpone revenue generation and undermine AI infrastructure investment plans.
On-site electricity generation and battery storage may reduce some grid-related risks, but they introduce additional investment and operational responsibilities.
Operators also require uninterruptible power supplies, power-quality management and tested backup systems to maintain availability during disturbances.
For AI data centers, electricity delivered on schedule can be more valuable than cheaper electricity that arrives years late.
India: Solar Tariffs Alone Cannot Determine AI Data Center Power Costs
India’s growing cloud computing and AI infrastructure market presents opportunities for renewable electricity procurement, battery storage and grid investment.
Solar power purchase agreements can help data center operators manage energy expenditure and sustainability commitments.
However, the actual cost of electricity delivered to a facility depends on several factors, including state electricity tariffs, open-access charges, transmission costs, banking regulations and grid availability.
These conditions vary across Indian states.
A low solar generation tariff does not automatically translate into the lowest delivered electricity cost for an AI data center operating 24 hours a day.
Developers must evaluate hourly electricity demand, renewable generation patterns, storage requirements and backup arrangements.
Grid connectivity and the availability of dependable power should also influence site selection.
For Indian operators, the most competitive electricity strategy may involve a combination of grid supply, renewable PPAs, battery storage and backup generation.
What Should Microsoft, Google, Amazon and Other AI Investors Prioritize?
AI infrastructure developers should evaluate electricity projects against three commercial requirements: cost, reliability and delivery schedule.
The most important questions include whether electricity will be available when the facility opens, how much dependable power will cost and whether the arrangement can meet long-term emissions commitments.
Contract terms also matter.
Electricity buyers should examine availability guarantees, fuel-price exposure, tariff escalation, transmission responsibilities, battery replacement costs and construction delays.
A project offering lower generation costs may become financially unattractive if grid upgrades or equipment shortages postpone commercial operations.
Investors should therefore compare complete delivered electricity costs, not simply advertised generation prices.
Final Verdict: A Hybrid Power Strategy May Offer the Best Economics
Nuclear, solar and natural gas each address different requirements of AI data center electricity supply.
Nuclear offers firm low-carbon electricity but faces investment and construction challenges. Solar can reduce generation costs and emissions but requires complementary resources for continuous supply. Natural gas provides dispatchable electricity but introduces fuel-price and emissions risks.
For many large AI data centers, combining firm electricity contracts, renewable generation, appropriately sized battery storage and reliable backup systems may offer a stronger commercial solution than depending on one technology.
The winner in the AI electricity race will not necessarily be the technology with the lowest generation cost. It will be the solution that delivers reliable power at the lowest sustainable total cost, on the date the data center needs it.
SHAFANA FAZAL
