The federal government has identified AI infrastructure and domestic energy production as national priorities, but translating policy direction into energized projects depends on permitting reform, transmission investment, and regulatory execution. A senior White House energy official opens DPX with a policy-level read on large-load development: which federal actions are underway, where coordination is affecting project timelines, and what utilities, developers, and data center customers should watch over the next 12 months.

Federal policy can set priorities, but regulatory sequencing will shape which large-load projects advance, where costs are assigned, and how quickly infrastructure can be built. Co-location reform, large-load tariff design, cost allocation, and regional transmission planning are moving through active FERC proceedings as utilities, developers, and hyperscalers make capital decisions. A senior regulator will address which decisions are on the commission's docket, what tradeoffs regulators are weighing, and how decision timing affects projects already in development.

AI load growth is forcing utilities and developers to translate semiconductor and data center trends into power demand forecasts. Higher rack densities, larger accelerator deployments, and the shift from training-heavy growth to broader inference workloads are changing where load appears, how concentrated it becomes, and how quickly sites move through planning. This briefing will connect chip and rack-level trends to delivered megawatts, explain why demand forecasts still vary widely, and identify the assumptions utilities and developers should test before committing capital.

Across the country, communities are taking a harder look at proposed data centers. Dozens of cities and counties have delayed, paused, or rejected projects over concerns about power demand, water use, noise, land use, and quality of life. What began in Northern Virginia has become a national trend, with high-profile debates unfolding in Texas, Georgia, Tennessee, Arizona, Ohio, and other fast-growing markets. Local officials — regardless of political affiliation — are imposing new requirements, considering moratoriums, and demanding greater community engagement and benefits before approving projects. This panel explores why community engagement is critical to data center development, separates fact from fiction on the drivers of opposition, examines the disputes shaping future policy, and discusses what actually changes stakeholder opinions and earns public trust.

Across the country, communities are taking a harder look at proposed data centers. Dozens of cities and counties have delayed, paused, or rejected projects over concerns about power demand, water use, noise, land use, and many other factors. What began in Northern Virginia has become a national trend, with high-profile debates unfolding in Texas, Georgia, Tennessee, Arizona, Ohio, and other fast-growing markets. Local officials—regardless of political affiliation—are imposing new requirements, considering moratoriums, and demanding greater community engagement and benefits before approving projects. This panel explores why community engagement is critical to data center development, separates fact from fiction regarding opposition drivers, examines the disputes shaping future policy, and discusses what actually changes stakeholder opinions and earns public trust.

Hyperscaler offtake is now the most bankable signal in the power sector — and the voluntary Ratepayer Protection Pledge has reset the baseline for what lenders expect. Traditional PPA-anchored project finance is still hitting walls. Build-and-transfer, multi-party ownership, and hybrid risk-allocation structures are emerging because credit committees want covenant packages, tenors, and contract terms that traditional PPA-based project finance was not built to handle at this scale. An economic consultant explains what lenders and equity actually require in 2026, while a developer and a utility examine the deal structures and take-or-pay terms now in market.

Federal interconnection reform promised to cut timelines from seven years to closer to two. PJM has processed more than 300 GW of proposed generation since 2020 and signed interconnection agreements for about 103 GW — yet only 23 GW is in service. ERCOT, meanwhile, is managing hundreds of gigawatts of large-load requests with realization rates far below the volume of requests. For hyperscaler procurement and developers, the same question decides whether a 2027 or 2028 energization date is real, and how long they can wait before behind-the-meter options become the only credible backup. This session focuses on live cycle data: study throughput, withdrawals, network upgrade cost trends, and large-load prioritization.

AI-driven data center growth is straining the availability of transmission, gas turbines, reciprocating engines, and large power transformers, but those constraints are also changing how projects are planned. Utilities, developers, suppliers, and large-load customers are reassessing grid interconnection, on-site generation, modular deployment, fuel strategy, equipment reservations, and long-lead procurement. This panel will compare where bottlenecks are most acute, where alternative technologies and procurement models can create near-term options, and what project teams should evaluate before committing capital.

Capital plans at several large U.S. utilities have stepped up sharply to serve data center corridors where contracted and requested load have surged over the last two years. The open question is how those utilities phase new generation, transmission, and distribution so they can serve committed projects without stranding assets if hyperscaler timelines slip or demand forecasts prove overstated — recent analysis suggests roughly 25 GW of potential double-counting in aggregated data center forecasts. In this keynote address, a senior utility executive lays out the numbers behind one utility's plan: how much capacity is tied to firm contracts, how much depends on requests that may not materialize, and what assumptions must hold for the build to make economic and regulatory sense.

Hyperscalers are committing capital on timelines that do not always match utility planning, regulatory review, or grid construction. This keynote gives the data center customer's position in direct terms as part of Day 2's three-part opening sequence. A senior hyperscaler energy leader explains what siting requirements, reliability standards, flexibility expectations, and contract structures have to look like before a large power commitment can be signed.

This fireside closes Day 2's opening sequence by turning the two keynote perspectives into one concrete question: what would it actually take to move large-load demand from request to executed agreement? The discussion will examine how minimum payments, capacity commitments, cost recovery, flexibility expectations, and risk allocation show up in real contracts, and how each side thinks about moving from interest to a signable deal.

AI load has generated a wave of federal policy, utility pledges, and state cost-allocation agreements. The demand forecasts driving all of it are still in dispute. Speed-to-power has become a governing pressure, but utilities, regulators, developers, and communities still have to decide how much AI-driven load is real, how quickly it will arrive, what infrastructure it requires, and who pays if the forecast is wrong. This panel puts demand modeling, grid planning, public-power cost allocation, AI-first infrastructure, and colocation development in one conversation: what does the next decade of build look like under current commitments—and where do the assumptions begin to fail?

Gas, nuclear, and geothermal face different realities on permitting, equipment, financing, and time to commercial operation. Gas is already being built for data center load but runs into turbine slot backlogs, siting fights, and air permits. Advanced reactor projects are moving through new licensing paths and have strong bipartisan policy support, yet first-of-a-kind execution and financing still push most commercial-operation dates into the 2030s. Enhanced geothermal has a handful of scaled projects and project-finance precedent. This session asks what each of these technologies can credibly do for data-center-driven firm load over the next decade and under what conditions.

Bring-your-own-power strategies have moved from contingency plan to active option for large data center loads facing interconnection delays, tariff uncertainty, and equipment bottlenecks. Multi-gigawatt programs pairing data centers with dedicated gas and fuel-cell capacity are already in deployment, and recent federal co-location and behind-the-meter reforms are poised to reshape the economics for some projects. This session examines the full BYOP/BTM decision: under current queue timelines, cost-allocation and tariff structures, co-location rules, fuel-supply and pipeline firmness, emissions and permitting obligations, and day-to-day operating duties — when does it make sense to secure your own on-site power, and when is a grid-served arrangement still the better choice?

AI-scale loads are changing grid-stability requirements. Fast-switching, high-density data center loads can create voltage, frequency, and ride-through challenges that traditional commercial load planning did not anticipate. This session compares four approaches to stabilizing systems under large-load growth: synchronous condensers, advanced storage, supercapacitors, and algorithmic load control. A utility grid operations executive evaluates each option from the buyer's perspective. The format is structured to avoid product pitches and focus on use case, cost, deployment readiness, and operational fit.

Data centers are not typically flexible loads. Today, about a gigawatt of contracted data center flexibility is in place across multiple U.S. utilities.  This flexible capacity has the potential to expand in the future because of EPRI's DCFlex work with hyperscalers, utilities, technology providers, system operators and data center developers. Over the last two years, DCFlex developed a framework, called Flex MOSAICTM, that provides a common language for flexibility in the data center industry. The DCFlex initiative has also tested how far that flexibility can go. This briefing presents an overview of the Flex MOSAIC classification framework, potential applications of the framework and evidence from DCFlex's in-production demonstrations.

DPX closes with the organizations whose members must live with the outcomes: utilities, data center operators, generation developers, nuclear advocates, and clean-energy buyers. After two days of constraint analysis, this discussion asks where the industry has credible momentum and where risk is still being understated. Which permitting battles are winnable? Which queues are realistic? Which cost-allocation fights could kill deals that appear viable? Which technologies are moving from promise to execution?