Applied Kinetics, a two-person startup backed by Y Combinator, has built AI agents to automate equipment procurement for data centers and power infrastructure, targeting a phase of the supply chain that founders say has become a critical bottleneck in America's electricity buildout.
The company's agents operate within Microsoft Teams and Slack, connecting to enterprise systems from SAP, Oracle, and NetSuite to track suppliers, reconcile engineering changes as they surface during manufacturing, draft purchase-order amendments, and flag drawing discrepancies for transformers, switchgear, generators, and cooling equipment. Founders Rohit Naras and Adhik Durga, who met at Georgia Tech and worked previously on AI infrastructure at Crusoe and IBM, describe Applied Kinetics as an "AI-native, end-to-end manufacturer" managing the arc from technical requirements through production and delivery.
The timing appears calculated. Equipment procurement has emerged as one of several chokepoints slowing data center construction at a moment when power demand for artificial intelligence workloads is surging faster than many utilities anticipated. Gartner projected in June that global data center electricity consumption will reach 565 terawatt-hours in 2026, up 26 percent from 447 TWh in 2025, with AI-optimized servers accounting for 31 percent of data center power in 2026. "Surging demand for compute-intensive AI workloads is driving unprecedented data center power growth, while AI capacity is now constrained by power availability," said Linglan Wang, a director analyst at Gartner.
Grid Queues Stretch Past Five Years
The U.S. interconnection queue held approximately 8,200 active projects seeking transmission interconnection as of late 2025, totaling 1,312 gigawatts of generation and 749 GW of storage, according to Lawrence Berkeley National Laboratory's "Queued Up: 2026 Edition" published in May. The median time from interconnection request to commercial operation exceeded five years for projects reaching operation in 2025.
ERCOT reported roughly 410 GW of large-load interconnection requests to the Texas Senate Commerce Committee in April, prompting the grid operator to implement a "Batch Zero" process for large loads that state regulators approved on June 18. Goldman Sachs Research in September 2026 raised its outlook for global data center power demand to a potential 170 percent increase by 2030 compared with 2025 levels. The bank had earlier projected U.S. data center power doubling to 66 GW by 2027 from 31 GW in 2025.
FERC Chairman Willie Phillips framed the challenge bluntly in May 2024: "Our country is facing an unprecedented surge in demand for affordable electricity while confronting extreme weather threats to the reliability of our grid and trying to stay one step ahead of the massive technological changes we are seeing in our society."
Carving Out a Niche After the Purchase Order
Applied Kinetics positions its post-purchase-order agent as complementary to a wave of AI tools addressing earlier stages of energy development. Siting, permitting, and interconnection studies have drawn considerable capital and attention, with startups including Basepoint, Paces Agent, Temos AI, Gridleaf, and Concurrent AI building agents to automate market selection, site screening, interconnection applications, and permit submissions for solar, wind, storage, and data center developers.
Enverus acquired interconnection-study specialist Pearl Street Technologies in March 2025, expanding automation in power-system modeling across utilities and grid operators. AWS launched an "Agentic Grid Planning" program with Duke Energy in September 2026. "Utilities are facing a challenge that scaling their manual processes alone cannot solve," said Joseph Santamaria, general manager for energy and utilities at AWS, when the companies announced the collaboration on September 17.
Applied Kinetics carved out the phase after equipment orders are placed, a period that can stretch for months or years on long-lead electrical gear. The company's agent reconciles engineering changes that emerge during manufacturing, notifies suppliers of revised specifications, drafts updated purchase orders, and coordinates milestone reconfirmations across email, enterprise resource planning systems, and product lifecycle management platforms. The founders have not disclosed funding, revenue, or named customers beyond their Y Combinator listing.
Federal Pressure to Accelerate Connections
Federal regulators issued show-cause orders on June 18 directing all regional transmission organizations and independent system operators to justify or reform their rules for connecting data centers and other large loads. The commission described the directive as one of FERC's "most significant actions" to modernize markets for large demand.
The International Energy Agency's 2026 "Electricity" report highlighted "move fast and connect things" measures, noting that more than 2,500 GW of renewable, large-load, and storage projects worldwide faced interconnection delays. Lawrence Berkeley National Laboratory cataloged more than 40 solutions spanning forecasting, interconnection process reforms, planning and procurement changes, market operations adjustments, and ratemaking in a June report titled "Speed to Power: Solutions for Accelerating Large Load Connections." The lab emphasized automation and data harmonization as essential to cutting timelines.
Duke Energy's John Pressley, managing director for digital strategy and engineering, said in September that the utility's engineers were "using specialized AWS AI Agents to study grid upgrade decisions around reliability, costs, and the impact on customers and the Duke Energy grid, all through auditable and codified Agentic workflows."
Whether Applied Kinetics can scale beyond a two-person operation to handle procurement coordination for the kind of gigawatt-scale projects piling up in interconnection queues remains an open question. The startup enters a field where established enterprise software vendors and utilities have spent decades building procurement systems, though not necessarily ones designed for the pace or complexity of the current buildout. For now, the company is betting that the gap between placing an order and receiving equipment has widened enough to justify a new category of automation.
