Baker Hughes LNG Demand Defies Higher Rates as AI Power Needs Grow

Baker Hughes says LNG demand has not slowed despite higher borrowing costs, because AI data centers and contracted energy projects still require dependable power.

Chairman and CEO Lorenzo Simonelli made that case at the Gastech conference in Bangkok. He argued that long-term purchase contracts continue to make large energy developments financeable, even as expensive debt raises construction costs.

The tension extends beyond Baker Hughes. Cloud companies want electricity quickly, while utilities, regulators, and developers must decide which generation sources can arrive on time. Natural gas currently offers speed and dispatchability, but a large LNG supply wave, volatile prices, and emissions concerns complicate the investment case.

Baker Hughes Sees Energy Projects Moving Forward

The important change is not a new LNG project. It is Baker Hughes reporting that financing pressure has not stopped the projects already filling its order pipeline.

Simonelli told CNBC that Baker Hughes had not seen a slowdown in major energy investment. He said project bankability rests on contracted offtake and the long-term outlook for energy demand.

An offtake agreement commits a buyer to purchase future output under defined terms. That commitment gives lenders greater confidence that a project will generate revenue after construction.

These contracts matter because an LNG export plant requires years of planning, permitting, equipment procurement, and construction. Higher rates can raise its financing burden, but they do not automatically erase contracted revenue.

Simonelli linked that resilience to population growth, industrial activity, and the relationship between data centers and energy supply. His argument was direct: AI infrastructure cannot expand without physical generation and fuel infrastructure behind it.

The company reported just over $37 billion in backlog during the interview, including work associated with gas infrastructure, LNG, and data-center power. Backlog represents contracted work that Baker Hughes has not yet completed or recognized as revenue.

Its latest reported results reinforce that order-book picture. Baker Hughes ended the second quarter of 2026 with $40.1 billion in remaining performance obligations, including a record $37.1 billion within its Industrial and Energy Technology segment.

The company also recorded $10.5 billion in quarterly orders, including $7.1 billion for that segment. Those figures do not prove that every planned energy project will proceed, but they show substantial committed activity.

The quarterly results also showed momentum in power systems and LNG. Management specifically identified power generation as a source of strength.

This makes Baker Hughes an informative, though interested, observer. The company sells compressors, turbines, services, and other equipment used across gas and power infrastructure. It benefits when developers approve more capacity.

That exposure also gives Baker Hughes early visibility into customer spending. If developers began cancelling equipment packages or delaying final investment decisions, the change would eventually appear in orders and backlog.

For now, the company’s account is that customers remain focused on demand years ahead, not only today’s interest-rate environment. The central claim is therefore about project durability, not immunity from higher costs.

That distinction matters. A project can remain economically viable while becoming more expensive, taking longer, or requiring stronger contractual protection.

The Gastech interview suggests that contracted demand still outweighs the financing headwind. The next question is why AI has become so important to that calculation.

AI Data Centers Turn Electricity Into the Bottleneck

AI spending creates value for energy suppliers only when planned computing capacity becomes a real, continuous electricity load.

Modern AI facilities concentrate thousands of processors within a limited area. Those chips require electricity, cooling, networking, and backup systems throughout the day.

The result is a location-specific infrastructure problem. A region can have sufficient generation in aggregate while lacking transmission capacity at the site where a developer wants to build.

That constraint increases interest in behind-the-meter generation. The term describes power produced near a facility and consumed without relying entirely on the wider transmission grid.

Baker Hughes supplies equipment for this type of distributed generation. Simonelli highlighted Southeast Asia, where constrained grids can push data-center operators toward dedicated or local power systems.

The same pressure exists in the United States. The International Energy Agency expects data centers to account for nearly half of American electricity-demand growth through 2030.

Globally, the IEA projects data-center electricity consumption to more than double, reaching about 945 terawatt-hours in 2030. AI represents the most important driver, alongside other digital services.

The agency’s energy outlook expects renewables and natural gas to lead the response. It projects natural gas generation serving data centers to expand by 175 terawatt-hours through 2035.

Natural gas currently supplies more than 40% of electricity consumed by American data centers, according to the IEA. The agency expects it to provide the largest increment of US data-center generation through 2030.

Gas plants attract developers because they can produce power whenever workloads require it. Wind and solar depend on weather unless paired with storage or other firm resources.

That does not make gas the only answer. Renewables are expected to supply nearly half of the additional electricity used by data centers worldwide through 2030.

Nuclear developers are also pursuing long-term agreements with technology companies. Small modular reactors, geothermal systems, batteries, and grid upgrades all compete for part of the opportunity.

However, several low-emission alternatives face construction, permitting, or commercialization timelines that extend beyond the delivery dates of current AI campuses. Gas can therefore function as a bridge between immediate computing plans and slower grid expansion.

This timing gap strengthens Baker Hughes LNG demand because the company participates in several parts of the gas chain. It can sell equipment for liquefaction, pipelines, power generation, and industrial services.

Yet data centers do not consume LNG directly in most cases. They consume electricity, while LNG connects distant gas supplies with importing markets.

The AI-to-LNG mechanism contains several steps. More computing raises electricity demand. Gas plants meet part of that demand. Greater gas consumption supports production, pipelines, import terminals, and export facilities.

Each step introduces uncertainty. A planned data center can be delayed. A grid operator can restrict connections. A utility can select renewables, nuclear power, or demand management instead.

The connection is still commercially meaningful because infrastructure decisions depend on peak requirements and reliability, not only annual energy consumption. A generator that runs less often can remain valuable if it covers periods when other sources cannot.

For cloud companies, the immediate pressure is clear. Announcing new computing capacity is easier than securing the electricity needed to operate it.

For energy developers, the pressure runs in the opposite direction. They must avoid building long-lived assets around AI forecasts that fail to become durable demand.

Baker Hughes LNG Demand Depends on Contracts, Not AI Hype Alone

Long-term purchase commitments, rather than optimistic AI forecasts, provide the financial structure supporting many LNG investments.

A liquefaction project converts natural gas into a chilled liquid that occupies far less volume. Specialized vessels can then transport it between regions without pipeline connections.

These facilities require large upfront investments. Developers generally seek long-term contracts before making a final investment decision, often called an FID.

The International Energy Agency found that projects representing more than 80 billion cubic meters of annual US liquefaction capacity reached FID during 2025. That was a record for the American LNG industry.

More than 130 billion cubic meters per year of LNG contracts were signed worldwide during that year. The United States accounted for about half of the contracted volume.

The IEA’s 2026 gas review described 2025 as the second-strongest year for LNG investment decisions after 2019. This occurred despite broader macroeconomic uncertainty.

Those commitments help explain why higher rates have not produced an immediate halt. Many projects entered the financing process with customers already promising to purchase output.

Buyers accept long contracts because they value security and predictable access. Sellers use those commitments to support construction financing and reduce exposure to unpredictable spot prices.

The structure does not eliminate risk. A contract can contain flexibility, price formulas, destination rights, and credit conditions that change its economic value.

It nevertheless shifts the debate. Developers are not simply betting that AI will increase gas demand next year. They are building around multi-year commitments from utilities, traders, governments, and industrial buyers.

AI adds another source of expected power consumption to that foundation. It strengthens the demand narrative, but it does not carry the entire financing case.

Baker Hughes expects installed global LNG capacity to approach 900 million metric tons per year by 2035, according to Simonelli’s interview. The company previously presented a higher internal outlook of 950 million tons per year for that date.

Management also said it expected to exceed its 2024 through 2026 target for LNG projects reaching FID. By the end of 2025, projects totaling 83 million tons per year had already reached that milestone.

These are Baker Hughes forecasts, not independent guarantees. They reflect management’s view of customer plans, global energy use, and the equipment market in which the company competes.

The independent outlook supports growth, but with more restraint. The IEA expects global gas demand to rise about 9% between 2024 and 2030 under its base case.

Asia Pacific accounts for roughly half of that increase. Power generation represents more than one-third of projected global gas-demand growth over the period.

That distinction prevents the article’s central claim from becoming an AI-only story. Population growth, industrial consumption, energy security, and declining domestic production in some markets also support LNG trade.

AI matters because it adds concentrated demand in markets that were already investing in gas. It also brings technology companies, utilities, and infrastructure investors into the same capital-allocation debate.

The primary opponent is therefore not gas versus renewables. It is contracted long-term demand versus the immediate cost of financing energy infrastructure.

So far, contracted demand appears to be winning. Baker Hughes’ order book and the industry’s FID activity both support that conclusion.

The result can change if buyers stop signing agreements, delay projects, or refuse the higher delivered price needed to cover financing and construction. That possibility is visible in the supply outlook.

The Coming LNG Wave Tests the Demand Story

Baker Hughes’ confidence faces a basic stress test: the industry is adding substantial supply while several major consuming markets remain sensitive to price.

The IEA expects roughly 300 billion cubic meters of new annual LNG export capacity to become available worldwide by 2030. The United States and Qatar account for most of that expansion.

A larger supply base should improve energy security and reduce prices. It could also pressure project returns if capacity arrives faster than buyers can absorb it.

The IEA’s LNG market analysis says long-term contracts lasting at least ten years represented 75% of volumes contracted since 2022. That provides developers with protection, but it does not remove market-wide oversupply risk.

Lower prices can stimulate demand among price-sensitive Asian buyers. They can encourage switching from coal to gas and make imports affordable for new markets.

However, the response is not automatic. Import terminals, pipelines, power plants, and local market reforms must exist before lower-priced LNG can become electricity.

The agency has also warned that weak economic growth, delayed gas infrastructure, and contractual restrictions can limit demand. If those obstacles persist, additional supply can weigh on utilization and prices.

Its 2025 World Energy Outlook presented an even sharper challenge. Under the stated-policies scenario, available LNG supply exceeds demand by 65 billion cubic meters in 2030.

That supply overhang would challenge the assumption that every sanctioned project earns attractive returns. It would not necessarily cancel equipment orders already under contract.

Baker Hughes operates one step removed from commodity ownership. It sells technology and services to asset developers, which can provide revenue before an LNG plant proves profitable over its full life.

Still, lower project returns can eventually affect future orders. Developers facing weak prices may defer another train, reduce expansion plans, or demand lower equipment costs.

AI demand also contains its own forecasting risk. Data-center developers routinely announce projects before securing land, grid connections, permits, chips, or customers.

Some proposed campuses will never reach full scale. Others will operate below their designed capacity, improve efficiency, or shift workloads toward regions with different generation mixes.

Technical progress can reduce electricity consumed per AI task. Falling costs can then stimulate more usage, partly or fully offsetting those efficiency gains.

Regulators can also intervene. Grid authorities may pause connections, require dedicated generation, impose water limits, or make developers fund network upgrades.

The US Energy Information Administration reported that Texas paused new data-center grid connections while authorities reviewed proposed projects. Yet the agency still expected the region to deliver a large share of national electricity-sales growth.

Its September forecast projected the West South Central region to contribute nearly 20% of nationwide growth in 2026 and almost 40% in 2027.

That example captures the central uncertainty. Demand can remain strong while individual projects face meaningful delays.

Environmental constraints add another layer. Burning gas emits carbon dioxide, while production and transportation can release methane, a greenhouse gas with strong near-term warming effects.

The IEA estimates that extracting, processing, and transporting natural gas accounts for nearly 17% of its average life-cycle emissions. Actual performance varies widely among supply chains.

Technology companies have made climate commitments that can conflict with rapidly rising electricity needs. Long-term dependence on unabated gas generation can make those targets harder to meet.

Carbon capture can reduce some emissions, but performance, cost, and deployment remain project-specific. Renewable generation, batteries, nuclear power, and geothermal energy will keep competing for data-center contracts.

None of these risks invalidates Baker Hughes’ current backlog. They show why “no slowdown yet” should not be interpreted as “no slowdown is possible.”

The company’s claim is strongest when applied to projects protected by contracts and already approaching construction. It becomes less certain when extended to every proposed LNG plant or AI campus through 2035.

Gas Gains Because Grids Cannot Move at AI Speed

The deeper mechanism behind Baker Hughes LNG demand is a mismatch between data-center construction schedules and the slower expansion of power grids.

AI developers often plan computing capacity on a timeline measured in months or several years. Major transmission lines can require longer periods for approval, land acquisition, construction, and connection.

That mismatch changes purchasing behavior. A data-center operator can wait for the grid, choose another region, or arrange local generation.

Behind-the-meter gas turbines offer one response. They can serve a campus directly and later operate alongside a stronger grid connection.

This architecture attracts developers because it offers greater control over timing and reliability. It can also create duplicated infrastructure and greater exposure to fuel prices.

Utilities face a difficult decision. If they build generation and transmission for projected AI loads, consumers can inherit costs when those loads fall short.

If utilities move too slowly, high-value projects can relocate. Regions then lose construction, tax revenue, and associated industrial development.

Gas equipment suppliers sit near the center of that timing conflict. They sell assets that can provide firm generation before newer nuclear designs or major transmission projects become available.

Baker Hughes has already raised its three-year target for data-center-related orders. In prepared remarks for its fourth-quarter 2025 results, the company increased that goal from $1.5 billion to about $3 billion for 2025 through 2027.

Management also described resilient power supply as a major bottleneck. That language is commercially significant because it identifies grid scarcity, not merely chip demand, as the sales driver.

The mechanism reaches LNG when additional gas-fired power demand influences expectations for regional gas balances. Importing nations can respond with new terminals or longer supply contracts.

Exporters then need liquefaction equipment, compressors, monitoring systems, and services. Pipeline developers require another layer of machinery and long-term maintenance.

This chain gives Baker Hughes several opportunities to earn revenue from the same demand trend. It also means weakness at one point does not affect every business line equally.

A delayed data center can reduce near-term turbine demand without cancelling an LNG export project supported by other buyers. Conversely, a weak global gas market can hurt future liquefaction orders while local generation demand remains strong.

That portfolio effect helps explain management’s confidence. Baker Hughes does not rely on one AI customer, one power market, or one LNG terminal.

Competitors share parts of the opportunity. GE Vernova and Siemens Energy supply major power-generation equipment, while engineering and service companies support LNG construction and offshore production.

Equipment scarcity can strengthen supplier pricing and order visibility. It can also slow projects when manufacturing capacity cannot expand quickly enough.

Simonelli said Baker Hughes was increasing capacity to serve expected demand. That creates its own execution test, because added manufacturing capacity needs sufficient orders over several years.

The industry must also prevent the pursuit of speed from locking customers into poor long-term choices. A gas system built for an urgent AI campus can operate for decades.

Developers can reduce that risk through modular designs, efficient turbines, lower-methane supply, carbon controls, and contracts that accommodate a changing generation mix.

The practical lesson for enterprise buyers is broader than energy procurement. AI plans depend on physical constraints that software teams cannot solve alone.

Teams evaluating AI projects should document power availability, site schedules, regulatory exposure, and supplier commitments alongside models and chips. A searchable knowledge base can help preserve those cross-functional decisions.

Baker Hughes benefits today because energy infrastructure has become part of the AI stack. The durability of that advantage depends on real utilization, disciplined project selection, and credible contracts.

Three Signals Will Show Whether the Confidence Holds

Orders, signed LNG commitments, and actual data-center electricity use will determine whether Baker Hughes has identified durable demand or a temporary investment surge.

The first signal is Baker Hughes’ Industrial and Energy Technology order intake. Investors should compare new orders with revenue, cancellations, and remaining performance obligations.

A backlog can remain large while weakening underneath if new bookings fall below completed work. Continued growth in power systems and LNG orders would strengthen management’s argument.

The composition matters as much as the total. Equipment orders tied to projects with permits, financing, and contracted customers provide stronger evidence than early-stage announcements.

The second signal is the next round of LNG final investment decisions and long-term purchase contracts. These agreements reveal whether buyers still accept multiyear commitments after accounting for financing costs and prospective new supply.

Large contract volumes would support the view that energy security and Asian demand can absorb additional capacity. A sustained slowdown would weaken the Baker Hughes LNG demand thesis before it necessarily appears in reported revenue.

Readers should also watch contract quality. Duration, buyer credit, destination flexibility, and pricing terms shape whether a project can secure financing.

The third signal is actual electricity consumption at operating data centers. Announced gigawatts measure ambition, while metered demand measures the load that utilities and fuel suppliers must serve.

Grid-connection data, utility sales, and regional generation additions can show whether computing plans are becoming physical demand. Persistent growth would reinforce the case for gas infrastructure.

Cancelled campuses, falling utilization, or extended connection pauses would challenge it. So would rapid deployment of lower-emission generation that reduces the role assigned to gas.

These signals should be evaluated together. Strong AI electricity demand cannot support an LNG project without infrastructure and buyers, while firm LNG contracts do not prove that AI caused the demand.

Baker Hughes has credible evidence for its near-term confidence: a large order book, record segment obligations, active LNG investment decisions, and growing power requirements.

The uncertainty sits farther out. The market must absorb a major supply wave while managing geopolitics, emissions, affordability, and competition from other generation sources.

For technology leaders, the immediate question is whether their AI roadmaps include the energy contracts, grid connections, and operating constraints needed to become real. For energy investors, it is whether those commitments justify assets designed to operate long after the present AI spending cycle.

Baker Hughes LNG demand will remain a useful indicator at that intersection. Watch whether contracted orders keep rising, whether buyers continue approving LNG capacity, and whether data centers consume the power their developers have promised.