The Jenbacher and Waukesha engine maker says AI power demand and grid bottlenecks drove its equipment backlog to $8.0 billion
AI data centers have a power problem, and some of them have stopped waiting on the grid to fix it.
INNIO N.V. (Nasdaq: INIO), the company behind the Jenbacher and Waukesha gas engine brands, says its equipment order backlog climbed to $6.6 billion after the second quarter of 2026. That figure stood at $3.3 billion a year earlier. By the end of the third quarter, it had reached $8.0 billion.
The buyers driving that growth are not traditional utilities. They are large AI data centers in the US that want electricity on site, on their own schedule, without waiting in line for a grid connection.
Gas engines for the AI age
Data centers are the main reason. Orders linked to them made up around 60% of INNIO’s equipment revenue in the third quarter of 2026.
One customer stands out. An unnamed US data-center developer has placed orders totaling approximately 1.72 GW. That came in two pieces: a 1.1 GW order in July 2026, followed by a 620 MW follow-on in September 2026.
INNIO also secured a separate 450 MW order from a US energy company. That equipment is meant for data centers in North America, with deliveries scheduled by 2028.
Why the grid became the bottleneck
INNIO CEO Dr. Olaf Berlien has pointed to two forces behind the surge. One is the demand for electricity created by AI workloads. The other is grid bottlenecks, which can leave large facilities waiting for connections that power their operations.
There is a statistical wrinkle here, too. AI data centers increasingly rely on on-site power plants that official energy statistics do not count. A growing slice of electricity production may be happening outside the numbers that policymakers and analysts usually watch.
Scaling up the factory floor
INNIO currently has manufacturing capacity of 3.5 GW per year.
INNIO plans to expand capacity toward 10 GW annually by 2030.
What this means for energy and AI infrastructure
For investors, INNIO’s numbers are a data point in a broader trend: AI’s power needs are spilling beyond utilities and into equipment makers that serve decentralized generation. The 60% revenue share from data centers shows how fast that customer group has become central to INNIO’s business.
Concentration is the risk worth watching. When one unnamed developer accounts for approximately 1.72 GW of orders, the backlog depends heavily on that customer’s plans holding steady.
For energy planners, the bigger question is measurement. If more AI facilities generate their own electricity on site, and official statistics do not capture that output, then the public picture of how much energy AI consumes may be incomplete.
