A super project covering an area equivalent to 1.5 Manhattans, solely for powering a 1GW data center.
In the past year, the most anxiety – inducing thing globally was finding electricity for AI.
On June 10th, NVIDIA just published a technical blog targeting the supporting energy storage for AIDC. In NVIDIA’s description, energy storage is no longer an optional supporting role but an “essential part” of the power system in AI data centers.
For a company that sells chips to start worrying about batteries for its customers, this fact alone speaks volumes: Ultimately, computing power boils down to electricity.
The International Energy Agency (IEA) has calculated that the power consumption of global data centers will double from approximately 415TWh in 2024 to about 945TWh in 2030 – roughly equivalent to the annual power consumption of the entire Japan today.
Moreover, AIDC has a strict rule: There can be no power outage for even a single second.
So, although “green electricity” has always been highly anticipated, it has not been incorporated into the mainstream power – supply model. The core issue is its instability – once the sun sets, the electricity is gone.
The long – standing consensus in the industry is: While green electricity is good, it can’t handle the task of “base load” which requires all – weather stable supply.
However, a desert in Abu Dhabi is starting to challenge this consensus.
In January 2025, during the Abu Dhabi Sustainability Week, this project was officially announced. Sultan Al Jaber, the chairman of Masdar, made a rather significant statement: Intermittency has been the biggest obstacle for renewable energy for decades. He called it “the moon – landing – level problem of our era.”

Announcement report during the Abu Dhabi Sustainability Week
The project is configured as follows: 5.2GW of photovoltaic power, paired with 19GWh of battery energy storage, with the goal of stably outputting 1GW of power 24 hours a day, and it will be put into operation in 2027.
This is a very worthy project sample for study – how exactly can photovoltaic + energy storage power a data center? How can stability be ensured? What is the cost? Recently, foreign investment bank Bernstein dissected this project and did some calculations.
The conclusion is: In regions like the Middle East with abundant sunlight, low land costs, and sensitivity to gas prices, photovoltaic + long – term energy storage has the opportunity for the first time to truly approach a base – load power source.
What’s even more notable is: Most of the most valuable parts in this calculation involve Chinese companies.
The most counter – intuitive aspect of this project is that it seems wasteful.
5.2GW of photovoltaic power is only used to output 1GW of electricity.
In other words, to obtain a stable 1GW load curve, the project party has increased the photovoltaic installed capacity to more than 5 times.
However, “over – building” is the key. Bernstein estimates that this system can generate approximately 12.45TWh of electricity per year, which averages to about 34GWh per day.
For a 1GW load, only about 24GWh is needed per day. In other words, the power generation is intentionally increased by about 40%.
Where does the excess electricity go? During the sunniest part of the day, part of it is used directly, and part is stored in the battery; when the sun sets, the electricity in the battery is released to meet the demand.
In the past, photovoltaic power generated too much electricity at noon, and the power grid couldn’t handle it, so the excess had to be wasted. Now, this electricity is not wasted but is used at night. In essence, it’s like equipping the photovoltaic system with an oversized “power bank.”
What supports this “store during the day, release at night” model is the 19GWh energy storage – based on a 1GW output, it can last for about 19 hours. Even if there is no sun all night, the battery can handle the power demand on its own.
According to Bernstein’s calculations, the availability of this system can reach about 99.6%, which is approaching the level of a traditional power plant.

The project has achieved a continuous power output of 1GW
There is a key point here: How long the energy storage can last almost determines the success of this project.
Bernstein has made a comparison. Without any energy storage and trying to maintain a constant 1GW load, a large amount of the electricity generated by the photovoltaic system at noon is in excess and has to be wasted. The proportion of electricity that can actually be used will drop from the theoretical about 27% to only about 9%.
With 12 – hour energy storage, the availability can reach about 97.5%; increasing it to 19 hours, it’s about 99.6%. But if you increase it further, there is almost no improvement.
So, 19 hours is not to further improve the reliability but to leave a margin for powering the whole night and the battery’s own losses.

The increase in availability slows down significantly after 13 hours of energy storage addition
Being able to achieve it is one thing, and being cost – effective is another.
Let’s start with the conclusion: It’s really expensive.
The total investment in the project exceeds $6 billion, which is equivalent to more than 40 billion RMB. When spread over a stable output of 1GW, this upfront threshold is not low.
However, photovoltaic and energy storage have an advantage that gas – fired power can’t match: There is almost no fuel cost during operation. The sun is free, and the battery charging and discharging don’t consume oil or gas.
Based on this, Bernstein estimates that the cost per kilowatt – hour of this project is about $97/MWh – converted into a unit we are familiar with, it’s about 0.65 RMB per kilowatt – hour.
If the energy storage is reduced from 19 hours to 12 hours, the cost can be reduced to about $80/MWh, which is about 0.58 RMB per kilowatt – hour in RMB, and the availability can still be maintained at about 95%.
Then when is it more cost – effective than natural gas?
Bernstein’s turning point is: When the natural gas price rises above about $8/MMBtu, photovoltaic and energy storage start to become competitive.
Looking at the current market, the spot gas price in the US is only about $3.7, so photovoltaic and energy storage don’t have an advantage; but the spot price of imported LNG in Asia once reached about $17.5, in such places, photovoltaic and energy storage are more attractive.
So, there is a prerequisite for this: In places where natural gas is cheap and abundant, such as the US, gas – fired power is still a smarter choice.

The higher the gas price, the more cost – effective photovoltaic and energy storage are; the two lines intersect at about $8/mmbtu
What’s more worth pondering is where exactly the $6 billion is spent. Breaking down the accounts, the major expenses are obvious:

Energy storage accounts for almost half of the $6 billion
Energy storage alone accounts for almost half of the cost.
This has rewritten an old perception. In the past, when talking about reducing the cost of photovoltaic power, people focused on modules; but in the case of photovoltaic and energy storage for base – load power, the cost of modules has already been reduced to the extreme.
Now, the price of a photovoltaic module is as low as about $0.09/W; the energy storage system has also dropped to about $130/kWh, both at the lowest levels in the past decade. Whether this system can become even cheaper mainly depends on the battery rather than the photovoltaic panels.
Photovoltaic and energy storage also have an underestimated advantage: Speed.
Bernstein’s comparison shows that a photovoltaic and energy storage project can be completed in about 2 years, while a natural gas – fired power plant has to wait about 4 years due to the shortage of turbines, and a nuclear power plant takes more than 6 years. In a nutshell – the one who gets the power supply first makes money first.
Of course, it also has a drawback: It takes up a lot of space. This project needs to enclose about 90 km² of desert, which is about 1.5 Manhattans in size, and it must be built in places with abundant sunlight and low land prices.
This means that it can’t be replicated everywhere – it can only work in places like deserts and Gobi.
After doing the calculations, it’s time to see who will end up getting a share of this pie.
For domestic readers, the most interesting aspect of this project may not be how much electricity it generates in the desert, but its supplier list:
Let’s start with Sungrow. In May 2026, it signed a contract with Masdar to supply 7.5 GWh of the PowerTitan 3.0 energy storage system, plus 2.6 GW of photovoltaic inverters. Just for energy storage, more than a thousand sets of equipment will be installed.

Sungrow’s promotional picture
For the photovoltaic module part, JinkoSolar won the contract to supply 2GW. It uses the flagship Tiger Neo series, which has been specially adjusted for the high – temperature desert environment.

Picture of JinkoSolar signing a contract with Masdar
The general contracting of the project is entrusted to PowerChina and L&T of India.
Regardless of how the shares are divided, the two biggest beneficiaries named in Bernstein’s report are both Chinese companies: CATL, leading in battery cell and battery technology; Sungrow, leading in system integration and power electronics.
Looking at the bigger picture, this is actually a growing business.
Bernstein predicts that the average annual growth rate of global energy storage demand in the next five years will be about 34%; the global cumulative energy storage installed capacity will increase from about 281 GW in 2025 to about 1,991 GW in 2030.

Outlook for the global demand for power and energy storage systems
However, the bigger market is actually in China
该文观点仅代表作者本人，36氪平台仅提供信息存储空间服务。
36kr Europe (eu.36kr.com) delivers global business and markets news, data, analysis, and video to the world, dedicated to building value and providing business service for companies’ global expansion.
© 2024 36kr.com. All rights reserved.

source