Alphabet to Acquire Intersect Power in End Run Around the Grid

Alphabet is making sure to secure its own supply of power for AI-era data centers, striking a deal to acquire Intersect Power in a deal worth $4.75 billion: all cash and some debt assumed. The idea is simple and audacious: marry hyperscale compute campuses directly with utility-scale renewables deployment and batteries, avoiding congested utility connections and speeding time to power.

A deal based on speed to power for AI data centers

The deal carves out Intersect Power’s pipeline of upcoming development projects and the team that will deliver them for Alphabet, while keeping existing operating assets behind, to be sold off to other investors and operate independently. The deal is expected to close in the first half of next year, pending customary approvals. Alphabet will be an anchor customer of the new campuses, but Intersect’s “data parks” are intended to be multi-tenant venues, containing other companies’ AI accelerators as well as Google’s.

Alphabet already had a presence in Intersect after Google and TPG Rise Climate teamed up to co-lead an $800 million strategic round, as part of a move that the company said could see it mobilize up to $20 billion for clean energy and related infrastructure by 2030. Intersect has said its first dedicated data-park sites near wind, solar and battery systems should begin to come online late next year with a full buildout by 2027.

Why bypass the grid now to meet soaring AI power demands

AI now seems to have a power problem. The International Energy Agency projects that worldwide electricity usage from data centers, AI and crypto may exceed 1,000 terawatt-hours by 2026 — about twice as much as in 2022. In the United States, interconnection queues have grown to record sizes; Lawrence Berkeley National Laboratory finds more than 2,600 gigawatts of proposed generation and storage currently in queues, with median wait times extending to around five years and a fraction of all projects making it into operation.

And those delays mean postponed AI capacity. By locating energy consumption on-site — frequently through private wires and “behind-the-meter” designations — Alphabet minimizes dependence on utility upgrade timeframes and capacity limitations. Co-located batteries can shave peak demand, assist with grid-forming functions and smooth renewable output — all critical tools for achieving the round-the-clock carbon-free generation goals that Google has publicly set for this decade.

What Intersect Power brings to Alphabet’s energy plan

Intersect Power is a U.S. developer specializing in large solar and battery storage projects and creative financing structures to move projects from concept through commissioning. Its development capabilities — site control, permitting, grid studies, and offtake — paired with Alphabet’s balance sheet are meant to contract the length of time it takes for multi-gigawatt campuses situated near high-quality renewable resources and accessible transmission lines to get projects up and running.

The “data park” concept is powerfully simple: marry industrial-scale compute workloads with dedicated clean generation and multi-hour storage on a single campus. That can reduce the delivered cost of electricity, lower curtailments by bringing load and generation closer together, and provide increased reliability when compared to remote PPAs using congested transmission corridors.

Big Tech’s new direction in energy and data center power

Technology giants are scrambling to get their hands on electrons at scale. Microsoft recently announced a partnership with investment firm Brookfield to bring 10.5 gigawatts of new projects online by 2030, and Amazon is now one of the largest corporate purchasers of renewable power in the world through long-term contracts. Alphabet’s decision also continues a retreat from the industry trend of negotiating for electricity, in favor of more traditional vertical integration: rather than contracting for power, buying land and figuring out how best to create the digital estate where the compute exists.

The strategy doesn’t do away with the grid — interconnection still matters for grid resiliency, oversupply sales and system balancing — but it alters the sums. Rather than standing in queue for increases to system capacity, the campus delivers power at the load, leveraging batteries and flexible operations to reduce reliance on strained substations and long utility construction cycles.

Policy and market context for data power and interconnection

Interconnection reforms are underway. FERC’s Order 2023 is trying to expedite queue processing and studies as regional operators such as PJM and CAISO redo models to clear up backlogs. Texas’s ERCOT has been speedier in the past, but congestion and curtailment are growing in its high-renewable regions. Through developing private generation and storage plants closely located to load, Intersect’s campuses can help quash these regional frictions without having to wait for multi-year-long transmission builds.

There are regulatory nuances. “State siting rules, local land use approvals and intertie arrangements still matter” as do participation in capacity markets or the provision of ancillary services, which may differ depending on geography. But the underlying driver is the same: compute demand scales faster than a grid can grow, and developers able to align siting, power and permitting have an overwhelming edge.

What to watch next as Alphabet integrates Intersect Power

Key markers will be final deal approval, disclosures on the initial campus locations and nameplate capacity, and procurement roadmaps for wind, solar, batteries and potentially next-generation resources such as long-duration storage. Look for metrics related to 24/7 carbon-free energy progress, and the share of on-site generation serving AI workloads.

If all that goes according to plan, Alphabet’s spending spree could “reset the way hyperscalers think about their energy strategy” — as less of a commodity they purchase from utilities and more like an asset built into the data center itself. In a world where seconds of latency count and megawatts are in short supply, controlling not just the compute but also the electrons that power it may prove to be the ultimate competitive moat.