The First 100 Days of an Andy Ellis Presidency: What the Record Suggests

Photo: m01229, Wikimedia Commons

01 The record and the scenario

Andy Ellis is the Green Party nominee for governor of Maryland in the November 2026 election, running with Owen Silverman Andrews, the party's candidate for lieutenant governor, on a ticket branded Andy Ellis For Governor 2026. The research record assembled for this series contains no evidence of prior public office; what it contains instead is something rarer — a fully published, day-one-engineered policy plan on a single signature issue, released with a formal campaign press announcement on August 25, 2026 and posted at the campaign's own data-centers page.

That signature issue is the construction of data centers. “The data on data centers is clear,” Ellis's campaign argues: “nobody wants them except the tech/surveillance firms, and the billionaire class.” The harms his plan alleges are concrete — “skyrocketing electricity bills,” “undue pressure on public infrastructure,” and facilities that “poison natural resources and water.” The plan's targets are equally named: “bipartisan legislation in Annapolis [that] provided massive tax breaks to data center companies and removed state level regulation of their power generation,” secret “dealmaking between government and industry ... with iron-clad NDAs,” and Governor Wes Moore's May 2024 proclamation that his administration's goal is to “Supercharge the data center industry” in Maryland. The answer, in the campaign's words, is “STOP THE DATA CENTERS.”

The plan itself — unusually operational for a third-party gubernatorial candidacy — leads with “a day one executive order to: Stop all data center permit approvals from state agencies. Prohibit any state agency from engaging in NDAs with data center developers, or AI, or surveillance companies. Conduct immediate audits of MDE” — the Maryland Department of the Environment — with the stated goal of halting new data centers on day one and imposing “the strictest rules in the nation on the ones already built,” followed by 2027 legislation to ban new construction outright. The August 25, 2026 press release describes the policy's origin in equally distinctive terms: “crafted by spending months traveling around the state and talking to people who are fighting against the data centers ... from the people and the movements on the frontlines, not the Annapolis advocacy organizations, the industry, or the establishment politicians.” His framing of what frontline communities want is blunt: “They want to stop the data centers from coming.”

The rest of the record is thinner. The anchor video for this article is Ellis answering questions at the Maryland Green Party's 25th Annual Assembly on June 14, 2025. The campaign site's structure — Issues, Priorities, Platform — evidences a broader agenda, but the retrievable documents center on the data-center plan and the governance critique embedded in it: secret NDAs, captured regulators, tax breaks negotiated behind closed doors. The Green Party's national platform is not treated as evidence of Ellis's personal positions anywhere in this analysis; only his own published words are.

This article does not predict that Andy Ellis will become president. It examines how the individual's documented positions and record could translate into executive priorities if placed in presidential office.

Analysis — not prediction. N43 and Hermes AI grounds every scenario in the documented record as of September 21, 2026; where evidence is incomplete we say so.

02 Day One: what a president could do alone

Of the candidacies in this series, Ellis's is the one whose signature plan was engineered from the start as day-one executive action — so the translation into presidential power is unusually direct. His state-level order names three instruments — a permit-approval halt, an NDA prohibition, and immediate environmental audits — and all three have federal analogues that a president can reach without Congress. Each item below is labeled by evidence strength, and each federal translation beyond his published words is an inference this analysis flags as such.

Likely Day-One Actions

• A halt on new federal permit approvals for data-center and AI-infrastructure construction. Ellis's plan stops “all data center permit approvals from state agencies” on day one; the federal version directs the Army Corps of Engineers, the EPA and the agencies that control federal land, leases and grid interconnections to suspend new approvals pending review. The permit-halting instrument is the very one he promises to use — only the jurisdiction changes.
• A prohibition on NDAs between federal agencies and data-center, AI or surveillance companies. His plan bans state agencies from “engaging in NDAs with data center developers, or AI, or surveillance companies”; the identical prohibition binds federal agencies by executive order, and it attacks the secrecy he identifies as the deals' enabling condition.
• Immediate environmental-enforcement audits of existing AI and data-center facilities. His order directs “immediate audits of MDE” enforcement; a president would order EPA and Justice Department audits of water-use permitting, discharges and emissions compliance at operating hyperscale facilities. The audit is his instrument, translated to the agencies a president commands.
• A review of federal electricity and water policy for AI infrastructure. The plan's stated harms — “skyrocketing electricity bills,” pressure on “public infrastructure,” facilities that “poison natural resources and water” — map onto DOE electricity-policy attention, FERC proceedings and EPA water permitting review. Ellis has not written a federal energy or water policy; the translation is N43's inference from his documented rationale.
• A procurement-leverage directive conditioning federal AI and cloud contracts on the transparency terms his plan demands — no NDAs, full environmental disclosure, no state tax-break dependence. Federal purchasing is an executive power requiring no new law.
• A lobbying-ethics directive targeting the “corporations and their lobbyists” he says both parties listened to “and not the people of Maryland” — plausibly a revolving-door restriction on agency dealings with data-center and AI-industry representatives. The record supports the direction; the instrument is inferred.

Actions the Candidate Could Not Do Alone

• A nationwide ban on new data-center construction. His own plan concedes the shape of the constraint: the Maryland ban runs through 2027 legislation, and a federal ban would require Congress to write it — a far larger ask than any state law.
• “The strictest rules in the nation” for existing facilities, as binding national standards. Durable environmental rules for an industry come from statute and APA rulemaking, not executive order alone.
• Repealing the Maryland tax breaks and the state deregulation of data-center power generation he targets — state law a president cannot touch.
• Ending federal tax incentives for data centers and AI infrastructure — Congress writes those provisions and Congress would have to unwrite them.
• Restructuring federal electricity ratemaking or grid-interconnection law to protect residential ratepayers from industrial AI demand growth — statute, FERC proceedings and the states.

03 Days 2-30: reorienting the executive branch

The Ellis record maps onto a specific set of federal agencies, and — unusually for this series — the mapping is his own, because his plan names the instruments. The Environmental Protection Agency is first among them: his day-one order's “immediate audits of MDE” presages exactly this kind of agency — enforcement review, water permitting attention, and compliance auditing — and his charge that existing facilities poison “natural resources and water” places EPA enforcement at the center of a reorientation. The Army Corps of Engineers and the Interior Department follow directly, as the agencies whose permits and leases control where data centers and their transmission lines get built on and across federal jurisdiction.

The Department of Energy is implicated by the plan's electricity argument — “skyrocketing electricity bills” and “undue pressure on public infrastructure” — through its electricity-policy, grid-research and loan-program machinery, and FERC's interconnection proceedings shape how fast industrial compute demand reaches the grid. The General Services Administration and the Office of Management and Budget carry the procurement-leverage lever: federal cloud and AI contracts are executive instruments, and conditioning them is within presidential direction.

The Department of Justice appears twice in this record: once for environmental-enforcement litigation, and once for the secrecy problem — “iron-clad NDAs” between government and industry are, in his telling, how the deals happened, and a president holding that view could direct agency counsel to refuse and unwind them where legally possible. Beyond these, the record thins. No documented position exists on the Defense Department, the intelligence community, the State Department, HHS, Education or Labor — and no personnel philosophy beyond the implicit one in his campaign's insistence that policy come “from the people and the movements on the frontlines, not the Annapolis advocacy organizations, the industry, or the establishment politicians,” which reads as an appointments-and-consultation doctrine: staff and advice drawn from affected communities rather than established interests. The record supports no more characterization than that.

04 Days 30-60: the first legislative package

A first Ellis legislative package would be built around the data-center plan, whose own text concedes that its durable parts are legislative — the 2027 Maryland ban he proposes is, structurally, a bill. The labels below describe the strength of the candidate's documented commitment, not any bill's prospects.

• High evidence — a federal data-center and AI-infrastructure siting and construction statute. His plan's centerpiece is a ban on new data centers, written as state legislation for 2027. A federal version — a moratorium or a strict siting-and-environmental-review regime for hyperscale facilities — is the direct translation of the most developed policy in his record.
• High evidence — the strictest-in-the-nation operating rules for existing facilities. His plan promises exactly this for Maryland; its federal form is a statute (and APA rulemaking under it) imposing the water, electricity, disclosure and community-notice conditions his plan demands. He has not specified the federal rule text, so the mechanism is an inference, but the commitment itself is his stated goal.
• High evidence — ending federal tax breaks for data-center and AI companies. The plan attacks “massive tax breaks to data center companies” as the deals' engine; ending their federal analogues is the same position at a different level of government, and it runs through the tax-writing committees.
• Moderate evidence — transparency legislation banning government NDAs with data-center, AI and surveillance companies. Documented in the plan as a day-one executive prohibition; making it bind state and local governments as well would require statute.
• Moderate evidence — electricity-ratepayer protection for residential customers facing industrial AI demand growth. His stated harms — “skyrocketing electricity bills,” “undue pressure on public infrastructure” — establish the position; the federal ratemaking instruments are inference.
• Low evidence — water-resources protection for AI facilities. “Poison natural resources and water” is documented as a harm he alleges; a federal water statute is not specified. Included because the concern is stated, not because a bill's shape is known.

No items are padded in from the Green Party's national platform. On healthcare, housing, criminal justice and foreign policy, the retrievable record does not document Ellis's personal positions, and this analysis declines to infer them.

05 Days 60-100: where governing meets reality

The Ellis scenario has an unusual shape: the unilateral slice of his agenda is larger than most in this series — his plan was engineered for day-one execution — while the durable slice collides with some of the hardest constraints in federal policy. A nationwide construction ban or strictest-in-the-nation operating regime requires Congress in an environment where AI infrastructure has powerful bipartisan constituencies: both parties have passed semiconductor and AI-competitiveness spending, both have courted data-center jobs, and the states compete for the facilities with the very tax breaks Ellis wants gone. A president proposing the ban would be proposing to reverse a bipartisan industrial consensus.

The unilateral actions face their own realities. A permit-approval halt across multiple agencies invites Administrative Procedure Act litigation, takings and estoppel claims from project sponsors, and the precedent of contested federal permitting freezes — courts have been receptive to arguments that indefinite general suspensions exceed delegated authority. The audit and NDA directives are on firmer ground but narrower in effect: audits find violations slowly, and NDAs already signed may be enforceable contracts the government cannot simply abrogate by memorandum. Procurement leverage works, but the federal government is a large AI customer — the leverage points in both directions.

Federalism cuts differently here than in most candidacies in this series. Ellis's plan is state policy aimed at state agencies — MDE audits, state permits, Annapolis tax law — and much of what he wants a governor to do has no federal command analogue. But the subject matter itself is interstate: AI facilities cluster around cheap power and water across state lines, transmission crosses FERC jurisdiction, and the electricity-bill and water-availability harms he documents do not respect borders. His federal presidency could reach infrastructure politics a state governor cannot — and would immediately meet the federalism objection that siting is traditionally state business.

The appropriations and personnel frictions are conventional: EPA and Corps permitting staffs process directives at their own pace; Senate-confirmed leadership arrives on the Senate's calendar; and the “frontline movements” his campaign credits as the policy's authors are not the interest groups that populate federal regulatory dockets. The honest 100-day picture is fast-moving executive memoranda, immediate litigation, and a legislative centerpiece with no visible path through either chamber.

06 The policy record in detail

Technology, AI & Surveillance

This is the deepest layer of the Ellis record, and it fuses infrastructure, economics and civil liberties into one argument. His plan groups “data center developers, or AI, or surveillance companies” into a single object of regulation — an analytical move worth noting in itself: the data center is treated as the physical plant of an AI-and-surveillance economy whose costs (electricity, water, infrastructure) are socialized while its gains are private. The documented positions: nobody wants the facilities “except the tech/surveillance firms, and the billionaire class”; they cause skyrocketing electricity bills, pressure public infrastructure, and “poison natural resources and water”; state leaders want the public to “foot the bill”; Annapolis “politicians from both parties” listened to “the corporations and their lobbyists and not the people”; the deals happened in secret “with iron-clad NDAs”; and the answer is to stop the data centers — day-one permit halt, 2027 construction ban, strictest-in-the-nation rules for what exists. On the specific connection his argument draws: the AI-infrastructure buildout is, in his documented framing, an electricity-and-water problem — the facilities' power demand lands on household bills and their water use threatens natural resources, so regulating AI infrastructure is for Ellis inseparable from protecting ratepayers and water. Beyond that framing, Ellis has not publicly detailed a federal policy on AI safety, model regulation or surveillance law; this analysis does not infer one.

Economic Policy

The documented economics are oppositional: an end to “massive tax breaks to data center companies,” an end to the deals “behind closed doors,” and the demand that the industry — not ratepayers or the public — bear its own infrastructure costs. “State leaders want YOU to foot the bill for this” is the campaign's cost-allocation claim. No position on taxes generally, deficits, trade or monetary policy is documented.

Energy & Environment

Electricity and water are the plan's stated environmental stakes: “skyrocketing electricity bills,” “undue pressure on public infrastructure,” water and natural resources “poisoned.” The direction is clear — industrial AI demand growth is an environmental and consumer harm to be halted, not accommodated — but no federal energy or climate mechanism is specified beyond the permit-halt and audit instruments.

Civil Liberties & Criminal Justice

The grouping of “surveillance companies” with data-center developers is the documented evidence — a stated hostility to the surveillance economy's physical and data infrastructure. No position on criminal justice, policing, clemency or surveillance law in detail is documented in the available record.

Government Transparency & Governance

The second-deepest layer of the record: the NDA prohibition is a named day-one instrument, and the secrecy critique — “dealmaking between government and industry has occurred in secret with iron-clad NDAs” — is the plan's diagnosis of how the industry captured both parties. The demand that policy come “from the people and the movements on the frontlines” is a governance doctrine as much as a campaign line.

There is not enough public evidence to determine Andy Ellis's position on healthcare, housing, education, immigration, foreign policy or defense. The retrievable record centers almost entirely on the data-center plan and its governance critique; his broader Issues and Priorities pages could not be retrieved for this analysis, and this article does not infer their content.

07 What could not happen alone — and what could

What President Ellis Could Not Do Alone

• Enact a nationwide ban on new data-center construction — his own plan writes the Maryland ban as 2027 legislation, and the federal version requires Congress.
• Impose “the strictest rules in the nation” on existing facilities as durable national standards — statute and APA rulemaking, years in the making.
• Repeal federal tax incentives for data centers and AI infrastructure — tax law belongs to Congress.
• Undo the Maryland tax breaks and state power-generation deregulation he targets — state law outside presidential power.
• Rewrite federal electricity ratemaking or interconnection law to shield residential customers — statute, FERC and the states.
• Ban state and local government NDAs with AI and data-center companies nationally — binding the states requires federal statute.
• Guarantee that AI demand growth never raises electricity or water costs for any household — a physical-economy outcome no order can deliver; only its regulation can be attempted.

What Could Change Within 100 Days

• A federal permit-approval halt for new data-center and AI-infrastructure projects, via agency directive — the direct translation of his day-one instrument.
• An executive prohibition on agency NDAs with data-center, AI and surveillance companies.
• EPA and DOJ environmental-enforcement audits of operating hyperscale facilities — the federal analogue of his “immediate audits of MDE.”
• Procurement-leverage conditions on federal AI and cloud contracts — transparency and environmental-disclosure terms.
• DOE and FERC review proceedings on AI-infrastructure electricity demand and grid interconnection — opening the ratepayer question his plan raises.
• A revolving-door and lobbying-contact directive for agencies dealing with the AI-infrastructure industry.

Potential Coalition Partners

The documented positions cut across the usual partisan grain. Communities fighting local data-center projects — the “people and the movements on the frontlines” his campaign credits — exist in both parties' counties, and their organizing has produced legislative allies in both parties. Ratepayer-protection politics aligns with state utility-reform coalitions. The anti-subsidy thread overlaps with fiscal hawks who oppose corporate tax breaks on principle. And the surveillance-skepticism embedded in his grouping of “AI, or surveillance companies” shares ground with civil-liberties progressives in both parties who have fought government data purchases.

Areas of Likely Conflict

The conflicts are predictable from the plan's own targets. The AI-infrastructure buildout has bipartisan support — the industrial-policy consensus behind semiconductor and AI-competitiveness spending is precisely what a construction ban would attack. Both parties' state-level operations court data-center investment with the tax incentives he wants repealed. Utility, technology and real-estate lobbies are among the best-resourced in Washington. And the permitting halt would meet immediate legal challenge from project sponsors, state economic-development agencies and the bar. The record gives no evidence about how an Ellis administration would trade any of these conflicts away.

08 The 100-day scorecard and what the record tells us

Read as a whole, the Ellis record is the profile of a single-issue insurgency with that issue unusually well worked out. The data-center plan is the one deeply documented policy in the record, and it is a plan for executive action — permit halt, NDA prohibition, audits — followed by legislation for the durable ban. That shape makes the 100-days translation cleaner than for most candidacies in this series: the federal analogues of his day-one instruments exist, and they are the presidency's to direct.

What moves fast in this scenario is the enforcement-and-transparency layer: the permit halt, the NDA ban and the facility audits could plausibly exist on paper within days of an inauguration, because all three fit inside ordinary executive direction of the agencies that permit, contract with, and regulate the industry. What moves slowly or never is the ban itself — the centerpiece requires a Congress whose bipartisan consensus is that AI infrastructure should be subsidized, not stopped.

Where the record is thin is everywhere beyond the plan: healthcare, housing, criminal justice, immigration, foreign policy, defense — all undetermined in the retrievable record. A fair reading notes that this is a candidacy that chose one fight and documented it, and that its federal program in office would be, on the available evidence, that fight continued at a larger scale.

Philosophically, the difference from both major parties is the cost-allocation frame. Where both parties' AI policies argue about how fast to build and how much to subsidize, Ellis's record argues about who pays and who decides — ratepayers or shareholders, frontline communities or “the corporations and their lobbyists.” That frame makes his record unusually specific on one industry and unusually silent on everything else.

Whether a president could hold a permit halt against the litigation, lobbying and bipartisan industrial consensus it would attract is a question about politics, not evidence. What the evidence supports is narrower and more useful: if the office ever reached him, the first 100 days would be about halting, auditing and opening the books on AI infrastructure — and the record says so in his own words.