Community intelligence is the third bottleneck.

Across the Ohio, Wisconsin, Indiana, Virginia, and West Virginia community reporting we read in the journal, the same sequence repeats. A developer approaches a small-town planning commission under a shell company name. The initial approach is bound by a nondisclosure agreement that limits what local officials can share publicly. The project proceeds into a municipal water hookup negotiation — municipal water is the path of least resistance because it bypasses state-level water withdrawal permitting — and into a local tax abatement discussion.

Somewhere in the middle of the process, the community finds out what is being planned. A local journalist, an elected official, a neighbor, or an environmental group discovers the hookup request or the zoning change or the incentive application. The NDA means the community's first information about a multi-billion-dollar project is usually incomplete and often adversarial in tone. Opposition forms quickly. A moratorium motion gets introduced at a municipal council meeting. Neighboring towns get calls from journalists who read the first story. The moratorium spreads.

The West Virginia Tucker County case is the canonical bad-case study. The state's Department of Environmental Protection permitted a 1,600 MW on-site gas plant with approximately 30 million gallons of diesel storage near an elementary school without disclosing that the load was a data center. By the time the community understood the scale of what had been approved, the permits were already issued. The local backlash fed directly into the passage of HB 2014, a state bill that defines a 'high-intensity data center' at over 90 MW, preempts local zoning authority for HIDCs, originally directed 100 percent of real estate tax revenue to the state, and builds the use of confidentiality into the permitting framework as a matter of state law. Live audience polls at community forums in the state showed roughly 80 percent support for data center development with guardrails, approximately 5 to 6 hard opponents, and 2 strong boosters — but the process itself was so opaque that the positive majority never had a voice until after the permits had been issued.

Maine was the first state to formalize this pattern at the statewide level with a moratorium rather than a preemption. LD 307 passed in 2025, banning new data centers over 20 MW through November 2027, pending a legislative study commission. Oklahoma followed with SB 1488. Wisconsin and Georgia have draft moratoriums. xAI's 41-gas-turbine Memphis Mississippi project was permitted over public opposition, generating a secondary political backlash that has intensified the legislative calendar in adjacent states. The Data Center Coalition's own tracking shows 300-plus state-level data center bills pending across 30-plus states as of early 2026 — a mix of moratoriums, tax-abatement rollbacks, community benefit mandates, water-reporting requirements, and NDA bans. None of them will pass in every state, but enough of them will pass that the operating environment for data center developers is materially different in 2027 than it was in 2024.

Data Center Watch — a privately-run project tracker — logged approximately $98 billion in announced data center projects that were blocked, delayed, or canceled in Q2 2025 alone. The proximate causes were overwhelmingly community opposition, local moratoriums, and permitting failures rather than queue or equipment bottlenecks. This is the bottleneck our industry talks about least and loses the most money to.

The most striking finding in our research on community benefit agreements came from a March 2026 Data Center Coalition CBA webinar: there is no public database of data center community benefit agreements. None. The executed CBAs live in municipal filing cabinets, in private counsel's DMS systems, and in local government meeting minutes that are not indexed anywhere searchable. When a developer asks 'what is a market-standard CBA for a 500 MW campus in Ohio,' there is no authoritative answer.

The absence has specific consequences. First, developers have to start every CBA negotiation from scratch. Each municipality reinvents the wheel because nobody has a reference library. Second, community coalitions have asymmetric information — some of them are well-networked and share template language, others are starting from zero. Third, the CBA language that ends up in the public filings is often the most-negotiated version that survived opposition, not the version the developer initially proposed. The actual bargaining zone is invisible.

The counter-signal worth noting: CyrusOne has published European survey data showing 51 percent positive, 42 percent neutral, and only 7 percent negative community sentiment toward data center facilities. That is not a hostile baseline — it is a vacuum of understanding that fills with whichever narrative arrives first. Only 35 percent of 16-to-24-year-olds surveyed made the connection between WhatsApp, TikTok, Spotify and the data centers that host them. The real community-intelligence risk is not overwhelming opposition; it is letting the first narrative be the adversarial one. CyrusOne has piloted 30-year community management agreements — a significantly longer horizon than most CBAs — as a way to address the trust gap directly.

We are building the CBA database as a core capability of the Community Intelligence module. Every executed data center CBA we can access through public municipal records, every template language we can extract from executed agreements, every benchmark on investment-per-megawatt and jobs-per-megawatt, every environmental and water commitment, every revenue-sharing clause — indexed and searchable. The public version includes the benchmark statistics; the customer version includes the full text library with redlining workflow.

The reference library is not going to be perfect at launch. We are aware that a single source-of-truth CBA database creates adversarial incentives on both sides — developers who fear their language becomes market-standard, community coalitions who fear their leverage is documented. The response is to publish transparently and to solicit corrections from both sides. The same research desk that publishes the queue brief also publishes the CBA brief; we take corrections from both sides of the table.

The 300-plus data-center-specific bills pending across 30-plus states is not a homogeneous category. We sort the bills into six sub-types, each with a different risk profile for operators.

Moratoriums are the loudest category but not the most common. Maine LD 307 is the marquee example. Oklahoma SB 1488 is similar. Wisconsin and Georgia have draft bills. A full moratorium forces a project to pause entirely until the sunset date, which is usually one to three years. Moratoriums are the hardest to negotiate around because they are statewide — local concessions cannot override them.

State-level preemption is the opposite category — also the loudest, but working in the other direction. West Virginia's HB 2014 is the cleanest template. It defines a 'high-intensity data center' threshold, overrides local zoning and permitting for projects above that threshold, and directs the revenue split between state and local jurisdictions. Operators who want the preemption framework see it as a fast-track around hostile local politics. Local jurisdictions see it as a taking of their authority. Both sides are actively lobbying in states where similar bills are in draft (Indiana, Ohio, Tennessee, Kentucky), and the legislative dynamics are different from moratorium bills.

Tax abatement rollbacks are the third category. Virginia's JLARC audit showed $732 million in foregone sales tax revenue from the DCRSUT program in 2023 alone, and legislators in Virginia, Ohio, Georgia, and Texas have introduced bills to roll back or cap existing abatements. Kate Gordon has flagged the state sales-and-use tax exemption regime as the obvious revenue-raiser for states that want hyperscalers to fund grid resilience and community impact, noting that GPU prices at $30,000 to $40,000 each have blown through the revenue projections that were the basis for the original exemptions. The abatement rollback bills rarely pass in their strongest form but often pass in a diluted form — a cap, a sunset, a community benefit contingency. Operators who modeled projects against the pre-2025 abatement regime face material capital stack revisions.

Community benefit mandates are the fourth category. Ohio HB 1210 is the template — a mandatory community benefit contribution for any data center above a specific megawatt threshold. Similar bills are in committee in Indiana, Wisconsin, and Illinois. The mandate language varies, but the common thread is converting informal CBA negotiations into a statutory minimum. For developers, the statutory floor can be higher than the negotiated-CBA norm, which shifts deal economics meaningfully.

Water disclosure mandates are the fifth category. Virginia HB 496 and SB 553 both require large data centers to report water withdrawal and consumptive use. Ohio HB 646 creates a study commission. Illinois is considering similar language under its Power Act framework. The disclosure mandates are not permit denials but they are public-scrutiny triggers — an operator who reports 5 MGD of water use in a drought region will face a different public reaction than an operator who was not required to report.

NDA-ban legislation is the sixth category and the most strategically consequential. Minnesota's 2025 trade-secret legislation includes language that constrains the use of non-disclosure agreements between data center developers and local governments. The bill frames NDAs as contrary to open-records principles and limits their enforceability in the municipal context. Similar language is being drafted in Oregon, Washington, and Colorado. But the more consequential signal is voluntary: Microsoft announced in late March 2026 that it would stop requiring NDAs on its data center projects. This is the first public commitment from a top-tier hyperscaler to abandon the NDA-first siting playbook. Once one hyperscaler drops NDAs, the competitive dynamics shift — community groups now have a public example of a developer operating transparently, and every other hyperscaler's NDA is implicitly compared against Microsoft's open posture.

Energy-cost allocation bills are the seventh category. Ohio's 85 percent take-or-pay tariff for AEP is the most aggressive version. Virginia's GS-5 rate class (effective January 2027) is a more negotiated version. Georgia's stipulated agreement with Georgia Power is a third variation. Oregon's POWER Act adds a growth multiplier — a 25 percent uplift on the data center rate class with additional uplift as load grows, framed as 'growth pays for growth.' Indiana I&M's 80 percent take-or-pay large load tariff with exit fees and collateral posting is the cleanest template. DSIRE now tracks 34 states and 60 distinct new large-load tariffs on these patterns. These bills are effectively already law in several states and are being copied in others.

Of everything we captured in the second pass on the journal, the single most consequential specific event for the community-intelligence landscape is Microsoft's late-March 2026 announcement that it would stop requiring NDAs on its data center projects.

Microsoft is not a marginal player. It is one of the top three hyperscalers by new capacity. It is the single largest counterparty in the nuclear-coupling BTM wave (Constellation, Three Mile Island). It has the highest-profile community controversies of any hyperscaler (Mount Pleasant, Quincy). The fact that Microsoft, specifically, is walking away from the NDA-first siting model tells us three things.

First, the public-disclosure costs have become lower than the opposition costs. Microsoft's internal math clearly concluded that negotiating in public was cheaper than negotiating in private and then dealing with the backlash. That math only holds in an environment where the backlash to private negotiations is already significant enough to outweigh the information-asymmetry advantage of private negotiation.

Second, the other top-tier hyperscalers now face a comparison problem. When a community group in Ohio asks why a proposed Microsoft project can be discussed publicly but a proposed Google or Meta project requires an NDA, the answer has to be something more persuasive than 'that is just how we do siting.' Every hyperscaler NDA is now visible against the Microsoft baseline.

Third, the NDA-ban legislation that is pending in Oregon, Washington, Colorado, and Minnesota becomes easier to pass. The counter-argument that hyperscalers will not participate in public processes is harder to make when one of them is already doing exactly that.

The prudent assumption for operators is that the NDA-first siting model is ending in 2026 and 2027 whether or not specific legislation passes, and the operators who figure out how to run successful public siting processes will have a durable advantage. Our Community Intelligence module is being built on that assumption.

Cliffcenter's existing Site Intelligence product already includes a Moratorium Watch feature and a basic opposition-and-community-intelligence overlay. The full Community Intelligence module adds five capabilities we do not ship today.

First, the CBA benchmark database. Every executed data center CBA we can find, indexed by megawatts, state, community, and negotiated terms. Benchmark statistics on investment-per-megawatt, jobs-per-megawatt, water commitments, noise commitments, decommissioning commitments. Customers get the full text library; the public version ships the benchmark chart.

Second, the bill tracker. The 300-plus pending bills indexed by state, category, current status, committee, and realistic probability of passage. Integrated with Queue Intelligence and Site Intelligence so that a candidate site flags the pending legislation that could affect its timeline. Weekly refresh from state legislative APIs where available, and manual updates for states without machine-readable feeds. The preemption category (WV HB 2014 template) is tracked separately from the moratorium category because the two have opposite operator implications.

Third, the NDA-ban watchlist plus the voluntary-disclosure tracker. Every state actively considering NDA-related legislation, plus every hyperscaler and large operator who has made a voluntary public-disclosure commitment. Microsoft's March 2026 announcement is the first entry in the voluntary tracker. Status tracking, bill language diffs, and scenario analysis for each state. This is the highest-salience category because the downside is the most structural.

Fourth, the community engagement workflow. A project management layer that turns community engagement from an ad-hoc, relationship-driven process into a structured workflow with standard deliverables. Stakeholder maps, outreach schedules, public meeting preparation, CBA negotiation tracking, benefit disbursement tracking, and post-commissioning community impact reporting. This is where the module shifts from intelligence to operations.

Fifth, the large-load tariff atlas — DSIRE's 60-tariff corpus cross-referenced with Cliffcenter's own tariff schema, so a candidate site can see exactly what it will be charged under each serving utility's large-load rate class, with side-by-side comparisons across Indiana I&M, Georgia Power, Oregon POWER Act, Ohio AEP, Virginia GS-5, and every other published tariff.

We have been debating internally whether this is the sixth Cliffcenter solution or a module within Site Intelligence. The case for a sixth solution is that the customer for community intelligence includes public-affairs teams and community-engagement consultants who are not the primary buyers of Site Intelligence. The case against is that the data overlaps heavily with Site Intelligence and a sixth solution increases our sales surface. The current plan is to ship it as a Site Intelligence module first, and separate it into its own solution if the customer feedback supports the split.

The Community Intelligence module is shipping in three phases across 2026.

Phase one, launching in May 2026, is the bill tracker, the preemption watchlist, and the moratorium map. Every pending state bill indexed, every active moratorium with a sunset date, every draft NDA-ban bill with status, and every state with an active preemption discussion. This phase is mostly data ingestion and UI — it does not require new legal or engineering capability, just consistent tracking.

Phase two, launching in Q3 2026, is the CBA benchmark database plus the large-load tariff atlas. This requires municipal records research across dozens of states and the construction of a canonical data model for CBA terms and tariff structures. We expect the initial release to cover roughly 150 executed agreements and 60-plus tariff structures, with monthly updates as new filings are published.

Phase three, launching in Q4 2026, is the community engagement workflow. This is the operational layer — stakeholder mapping, outreach scheduling, CBA negotiation tracking, benefit disbursement tracking. We are building this with input from three early customers whose teams already run community engagement at scale.

The research desk will publish quarterly updates on the bill landscape, the CBA benchmark statistics, the voluntary-disclosure tracker (starting with Microsoft), and the moratorium status as each phase ships. Customers get the full module access; the public version ships the benchmark charts and the moratorium map.

If you are reading this brief because you are running into community-intelligence problems today — a stalled zoning vote, an unexpected moratorium, a CBA negotiation that has stalled, a preemption bill that surprised your counsel — email the research desk. Phase one is early and we are taking customer input on what to prioritize.