NVIDIA highlights wave power as an energy bridge for AI using digital twins and accelerated computing

NVIDIA is making a direct case that the next bottleneck for artificial intelligence may not be chips, but power. In a blog post published June 22, the company pointed to Eco Wave Power, a startup in NVIDIA’s Inception program, as an example of how energy systems could become more “intelligent” as AI infrastructure grows. The focus is on converting ocean wave energy into electricity using marine infrastructure already built along coastlines, and using AI-driven software to model, plan, and operate those systems more efficiently.

NVIDIA said global electricity demand is rising rapidly as AI use spreads across AI factories, agentic AI, industrial AI, edge computing, and physical AI such as robotics and autonomous systems. In many regions, the company argued that expanding grid infrastructure to meet new load can take years due to permitting, transmission upgrades, land acquisition, and capital investment. That timing gap, NVIDIA said, is reshaping how companies and governments think about energy supply for AI.

Eco Wave Power’s approach, according to NVIDIA, targets wave energy deployment that can sit closer to where power demand is growing, including ports and industrial zones. NVIDIA described the concept as noninvasive floating infrastructure, “floaters,” attached to breakwaters or sea walls. It said seawater’s density is roughly 800 times that of air, which allows wave energy generation to extract large amounts of energy using smaller devices than typical wind turbines.

A key engineering claim in the post is that Eco Wave Power keeps expensive computing and conversion hardware on land rather than inside the marine floaters. NVIDIA said earlier wave projects faced a bottleneck at the power-management and distribution stage because hardware mounted on the floater could be exposed to potential damage during rough currents. Eco Wave Power, the company said, places computers, sensors, hydraulic conversion, and electrical parts at onshore centers, keeping that equipment dry and out of storms.

NVIDIA also linked the system to simulation. It said digital twins of wave patterns and floating infrastructure, built using NVIDIA Omniverse libraries (a platform for creating and running 3D simulations of real-world systems), can model wave conditions, structural behavior, deployment configurations, and operational scenarios before construction. In NVIDIA’s framing, those virtual environments help optimize engineering decisions, reduce deployment risk, and speed planning.

At the operational layer, the blog said NVIDIA accelerated computing and AI technologies can support real-time optimization of wave energy systems. The post cited predictive analytics, anomaly detection, environmental forecasting, and predictive maintenance as capabilities enabled by AI models that continuously analyze ocean conditions, equipment performance, and energy generation patterns. NVIDIA also said AI can align energy-intensive computing workloads with periods of stronger renewable generation and dynamically optimize power usage across distributed systems.

The energy focus is backed by project examples. NVIDIA said Eco Wave Power operates projects in Jaffa Port in Israel created with EDF Power Solutions and the Israeli Energy Ministry, and at the Port of Los Angeles developed with AltaSea and Shell. The company also said new projects are in development in Portugal at the Port of Leixões, in Taiwan at Suao Port, and in Mumbai, India, with Bharat Petroleum.

NVIDIA then pivoted from renewables to data centers, arguing that wave power could be relevant to electricity-hungry AI infrastructure because many data centers are moving toward coastal locations for cooling and water access, which can place them near ports. The post described “pilots” at the Port of Los Angeles intended to showcase wave energy as the sole power source for a data center without tapping into existing grid energy, and it described AI software as the control layer for that pilot. In that described system, NVIDIA said the software can monitor and predict when waves will be stronger during the week based on weather patterns and then allocate more intensive compute tasks during those periods.

NVIDIA also framed the broader timing and intermittency argument. It said wave energy is among the largest renewable energy sources and pointed to a U.S. Energy Information Administration estimate that wave energy could produce more than 60% of annual energy consumption, in the U.S. alone. On intermittency, the post quoted Eco Wave Power cofounder and CEO Inna Braverman saying wave energy can generate “around the clock,” contrasting that with solar’s daily and seasonal variability.

Several details remain unclear from the blog post itself. NVIDIA did not provide technical specifications, deployment capacity figures, cost comparisons, or performance results from the Los Angeles data center pilot beyond describing the concept of predictive task scheduling. The post also does not quantify how much compute utilization or uptime the approach delivers, or how permitting, grid interconnection, and safety requirements are handled in practice for each port site. What to watch next, analysts and industry observers may want to see are measurable pilot outcomes, any published engineering benchmarks for wave-to-power reliability, and whether additional public partners expand beyond the named collaborations.