What are the odds in the race for future fuels?

It takes several years to get a data centre up and running, so sometimes it’s easy to feel like you need a crystal ball at the planning stage. A question I am often asked is what will the fuel mix be like for data centre loads in a few years. I thought it might be useful to set out the current options for future fuels and the factors that will go into making the right fuel choices.

Ben Pritchard – CEO, AVK-SEG

For now pretty much all prime power comes from the grid, and indeed 95% of our current work at AVK is for traditional build data centres that have a grid connection and standby assets that run on liquid fuel. But time is not on your side if you don’t have your prime power contracted. The queues for connection are very long (8-10 years in London, and don’t mention Amsterdam or Dublin) and will get longer. In the EUDCA’s excellent recent State of European Data Centres 2025 report there is a widespread recognition that self-generation is coming. 4% of colocation and hyperscale operators interviewed said that microgrids are already a realistic option for them, 14% said they will be in two years, and a further 36% said they will be in five years. That is over half the industry.

With up to 25 gigawatts of new capacity expected by 2030, it is fair to say that a significant proportion of new builds will be taking care of their total generation needs. Looking beyond 2030 it seems extremely likely that the majority of data centre load will be served by private behind-the-meter generation. So, in reviewing future fuels, we should be looking at prime as well as backup.

This is not a comprehensive list, it is just intended to cover the likeliest contenders for a broad role in data centre energy design. It does not cover non-variable renewables such as tidal or geothermal, as they are less fuels and more energy transfers. Nor does it cover less likely but very exciting possibilities such as nuclear fusion. And there is a great deal of potential for blends and mixes of these fuels where they are compatible.

Lane 1: Natural gas

Natural gas creates emissions, but is the cleanest fossil fuel alternative, emitting 50 percent less CO2 than coal. It is the primary non-renewable source of power generation today, generating around 20% of EU power and 28% of UK power. Some advantages of natural gas are that it can be switched off and on easily, and can run at capacity factors exceeding 80 percent, plus it has an extensive delivery grid.

Lane 2: Diesel

Diesel is a fossil fuel derived from crude oil, making it a non-renewable resource with a high carbon footprint. It also contains impurities such as aromatics, sulfur, and metals, which can lead to engine deposits and increase emissions further. However, it is reliable, plentiful and competitively priced, and has been the traditional fuel for data centre backup generators.

Lane 3: Hydrotreated vegetable oil (HVO)

In place of diesel, many data centres are now using biodiesel alternatives such as hydrotreated vegetable oil (HVO) which can reduce GHG emissions by up to 90% compared to diesel. HVO is not as widely available as diesel and has 7% less lower power density, but it has a longer shelf-life and it prolongs the life of expensive equipment. The HVO name is used to cover several varieties, some of which are actually diesel blends, only HVO100 guarantees 100% pure HVO, with no diesel added.

Lane 4: Hydrogen

Hydrogen fuel cells convert the chemical energy of hydrogen into electricity, producing water and heat as byproducts. Some operators have introduced hydrogen-powered backup fuel cells for their data centres. The European fuel cell market has a capacity of 200-300 MW and is growing. However, not all hydrogen is green, and the use of hydrogen as a primary power source is still in its early stages.

Lane 5: Biomethane

Biomethane or biogas is produced via biomethanation systems that mimic natural processes to recover waste from farms, the food industry or sewage treatment plants. It can be used exactly like natural gas, without any modifications to your appliances or gas system. In 2024, Europe’s biomethane production capacity grew by 20% to 6.4 billion cubic meters per year, around 7% of the capacity provided by natural gas.

Lane 6: Methanol 

Methanol is liquid, low-impact fuel for data centres, particularly as a replacement for diesel in backup. It is being widely used for shipping now, and can be produced from renewable sources (green or e-methanol as opposed to grey methanol), and can use existing infrastructure for storage and transportation. But greater production capacity is needed to improve cost-effectiveness.

Lane 7: Ethanol

Cellulosic ethanol is produced from cellulose, a component of plant cell walls, which is abundant in agricultural waste and non-food crops. It is derived from low-cost, readily available materials such as agricultural waste, and it promotes the circular economy by turning waste into a valuable resource. It has been piloted as a data centre fuel but is not yet widely available.

When it comes to deploying these fuels of the future for data centres, there are a lot of factors and variables that have to be considered. The four key ones are:

1. Emissions – targets and regulations

The industry has made significant commitments to reduce and ultimately eliminate emissions, primarily CO2, and this is now regulated via national and transnational laws, such as the EED. But with rising power demand from AI, industry emissions have started to rise sharply. Power demand will continue to rise for the foreseeable future, and this is driving major investment in new energy sources. The question is what priority do operators (and their customers in the case of colocation) place on emissions reduction when the lack of grid capacity means emissions targets are undermining the industry’s capacity to expand? And also how far will regulators hold them to account during this transitional period? Another variable is government-driven. For instance, many politicians oppose high levels of investment in new nuclear infrastructure which will take 10-15 years to complete, potentially replacing greener infrastructure investments in the run-up to 2050. The way this debate plays out in different markets will inevitably have a knock-on effect on new fuels for primary data centre power.

2. Applications – standby or baseload?

Different fuels suit different use cases. If you’re running for a shorter period of time, you can have on-site storage. Liquid fuels like diesel or its renewable substitute HVO (if you have engines that can run on it) are very popular for standby, as they are easy to replenish, dense (take up less space) and liquid fuel combustion engines start up fast. But hydrogen cells also work well here.

Then, if there isn’t a grid connection, you would look at a prime fuel application. Liquid fuel is not ideal for this because of the need for storage and the size of the storage tanks. When things are running constantly, you need a permanent supply that is going to fuel these systems. At the moment, without a grid connection, natural gas is used across every generation technology, whether it’s a fuel cell or a reciprocating engine or a gas turbine. However, those fuels could potentially change, and you need flexibility built into your power delivery technologies in case this happens. Medium-speed engines work well for prime power, where they are going to power long-term base load, and the same is true for turbines. 

3. Geography – urban or greenfield?

We don’t know the future anatomy of AI at the moment. Will it be predominantly hyperscale, or will it be more of a mix of hyperscale and smaller, more urban facilities? 

This impacts fuel selection a great deal. For example, fuel cells may be most suitable in busier locations, like cities where there are tighter restrictions around emissions, and more limited space. However, cells currently have a limit of 10 to 15 megawatts. Also, their power density is a lot less than reciprocating engines. Each little fuel cell block is 100 kilowatts, so when you start to look at the scale of land that’s needed in a busy city area where land is expensive, once you require 15 or 20 megawatts, you’re putting a more expensive asset in that’s actually taking up more land. This may not be so tempting for operators. There are always going to be special local considerations. For instance, we have a customer currently that needs high speed standby, but because the facility is in a Marine Conservation Area, liquid fuel is not an option due to the risk of spillage. So in those circumstances, gas is the correct choice. 

4. Commercial factors – cost and time to market

Commercial volume and cost-efficiency is still a major factor. Gas prices in Europe have fluctuated hugely in recent years for geopolitical reasons. For new fuels, the commercialisation process takes time and requires a solid market to scale effectively. Until hydrogen or fuel cell technology gets truly taken up in a major industry, where it’s mass produced, prices will generally not be driven down. For instance we’ve seen that with battery storage, prices were extremely high, but they have now decreased significantly, partly due to the demand from the automotive industry.

Availability and speed to market are also key. The ethanol production process, for instance, is complex and requires advanced technology, and the infrastructure for distributing and storing cellulosic ethanol is not yet fully developed, which can limit its accessibility. 

A few years ago, everyone was saying hydrogen is the new thing. Then about a year later it seemed to stop being a thing. At the moment, when you say hydrogen, people raise their eyebrows. The point is not that hydrogen will not succeed at scale, it’s simply that the technologies and the market move so quickly that whatever you say now will probably be wrong a few years down the line.

Broadly speaking, for prime self-generation power natural gas looks good currently, integrated where possible with fast-growing renewables and with nuclear on the horizon. For backup HVO and fuel cells are excellent contenders, but factors like cost and geographic or regulatory circumstances are important and vary by market. 

Luckily, AVK is technology agnostic. We have no vested interest in specific fuels, we simply provide what we believe are the best power delivery mechanisms with a future carbon-free fuel map built in wherever possible. So our advice in such a fluid environment where time to market and data centre uptime are both critical would be to cover all the bases wherever possible. Grow your renewables steadily, particularly as they come online on the local grid. And for the rest, make sure that your equipment will run on as many low-to-no carbon fuels as possible. Getting the mix right and staying flexible is the key.  

We don’t want our customers to back the wrong horse; spread your bets, keep your emissions as low as possible and keep an eye on emerging future fuels. In the long run, there will surely be surprises in store.