Version 2

New York Governor Kathy Hochul signed Executive Order No. 62 on July 14, 2026, imposing a one-year moratorium on data centers consuming 50 megawatts or more — the first statewide construction ban of its kind in the US [1][2]. The New York legislation was sponsored by State Senator Kristen Gonzalez [3]. At the federal level, Senator Bernie Sanders introduced S.4214 and Representative Alexandria Ocasio-Cortez introduced a companion House bill, H.R.9442 — together the Artificial Intelligence Data Center Moratorium Act — though a separate data center energy bill in Congress hit a procedural obstacle in late June [5][6][9]. NVIDIA argues hardware efficiency gains make regulatory action unnecessary, while Republicans and the Trump administration frame moratoriums as threats to US AI competitiveness [10][4].

New York's moratorium is now a signed executive order with formal legal standing, not merely announced policy, and the federal companion bills give the restrictive approach a two-chamber foothold in Congress. Whether efficiency gains at the chip level can actually reduce total grid draw — rather than simply enabling more compute at the same power budget — remains the central technical question on which the policy debate turns.

On July 14, 2026, Governor Kathy Hochul signed Executive Order No. 62, making New York the first US state to formally prohibit new hyperscale data center construction [1][2]. The order imposes a one-year moratorium on facilities consuming 50 megawatts or more of power while the state develops consistent regulatory standards. New York State Senator Kristen Gonzalez sponsored the underlying Senate Bill S10642 through the legislature [3]. Hochul's stated rationale centers on establishing standards before approving further construction; community concerns include local pollution, rising energy costs, and water consumption [4].

At the federal level, the Artificial Intelligence Data Center Moratorium Act has now taken formal legislative shape. Senator Bernie Sanders introduced S.4214 in the Senate, and Representative Alexandria Ocasio-Cortez introduced companion legislation H.R.9442 in the House [5][6][7]. Roll Call coverage from March 2026 shows the federal push predates New York's state action, framing the moratorium as a response to AI infrastructure's broader environmental and grid impacts [8]. The political path for both bills is narrow: the Trump administration and congressional Republicans frame such restrictions as threats to US AI leadership and are unlikely to support them [4]. A separate data center energy bill in Congress encountered a procedural obstacle in late June 2026, adding uncertainty to the legislative picture [9].

NVIDIA has positioned itself as the primary industry voice against moratorium logic, arguing that performance per watt — not raw capacity — is the correct metric for evaluating data center impact. The company claims its GB300 NVL72 system delivers up to 25x better performance per watt than its prior Hopper generation on DeepSeek V4 Pro benchmarks, and that software optimizations alone improved DeepSeek V4 performance per watt by 5x within a month [10]. NVIDIA also notes that current facilities convert only about 60% of grid electricity into useful AI work, pointing to infrastructure efficiency gains as an alternative to construction bans [10]. The efficiency argument, however, sits alongside the fact that the Blackwell B200 GPU draws up to 1200W per chip — nearly double the 700W of the H100 it replaces [11][12] — leaving the question of total grid impact genuinely open.

The American Action Forum, a center-right policy organization, has published analysis framing AI infrastructure constraints as a policy risk in the 2025–2026 period [13], consistent with the broader conservative and industry opposition to moratorium-style regulation. The regulatory picture is now split: state-level action in New York has legal force, federal moratorium legislation exists in both chambers but faces structural opposition, and the efficiency-versus-capacity debate remains unresolved.