What's shipping into DC Hub — new capabilities for AI agents, plus the infrastructure data our crawlers add every day.
Updated from /api/v1/canon/phrases · machine-readable changelog: /api/v1/whats-new
Approved platform updates, rendered from the live feed — queryable by any MCP agent, open under CC-BY-4.0.
Copy of the 26 approved updates in /api/v1/whats-new, taken (UTC). The live list replaces it when the page's script runs.
The homepage is back to its July look with every figure read live, the nav carries a live ticker, and the /ai page shows live request counts again. A free key bound to an email now gets full answers, capped per day, from the grid, interconnection queue and market intelligence tools, not only previews. Facility discovery adds data centers found in EPA ECHO air-permit records (US federal, public domain), and a daily check at /api/v1/ops/data-onboarding shows whether each core data layer is still growing and whether its main source is current. The OpenStreetMap refresh retries the states that failed instead of skipping them. MCP clients can connect directly at https://dchub.cloud/mcp, which keeps working after Smithery retires its hosted proxy URLs on 2026-10-31.
Daily data-onboarding check (JSON) →Operators and brokers can now list available data center capacity on DC Hub for free at dchub.cloud/listings: powered land, powered shells, turnkey capacity and colocation to buy or lease. Buyers and AI agents search it by size, in kilowatts or megawatts, and by location with the source_capacity MCP tool or on the listings page, then request an introduction with request_capacity_intro. DC Hub sends the provider only the buyer's company and requirement; contacts are shared only if the provider accepts, and nothing is shared on a decline.
List capacity or search listings →A new research page lists generating units with filed retirement dates across the US ISOs, the interconnection capacity they free up, and how far each sits from a data center market, next to recent interconnection queue exits for each ISO. Every figure is built live from EIA Form 860M filings and ISO queue feeds, with its as-of date. The page is explicit about its limits: a queue exit is not available capacity, retirement dates can be extended, and renewable curtailment is modeled, so it is not reported as a measurement.
Open the Stranded Power Map →Each market score in the Data Center Power Index now comes with a confidence rating and a plain-language list of its inputs: which come from live sources such as the interconnection queue, planned generators and grid telemetry, and which are modeled estimates, such as renewable curtailment and demand growth. Confidence reflects how many inputs are live and how fresh the score is; when the record behind a score is incomplete, the rating is left empty with a reason rather than guessed. The breakdown is free on every tier, in the API and through the MCP tools get_market_dcpi_rank and rank_markets.
See the Power Index →A short comparison brief sets DC Hub beside datacenterHawk, DatacenterDynamics, Data Center Frontier, Baxtel and DC Byte, and beside research from CBRE, JLL, Structure Research and 451 Research. It sorts the field into publishers, data businesses, and brokerage and analyst research, and credits what each does well. It places DC Hub as the live layer for checking what you read elsewhere: live grid telemetry, a daily power index for every market we score, and an MCP server that AI agents query directly. It is written for customers, analysts and investors, every competitor detail cites a public source on its sources page, and it downloads as a PDF. DC Hub's own figures are self-reported, and corrections are welcome.
Read the brief →Agents can now ask which gas pipeline and transmission line projects are planned, approved or under construction, not only what is already built. get_infra_projects returns the EIA natural gas pipeline projects list for the whole United States: operator, status, states crossed, capacity, miles, cost and docket. It also returns ERCOT's transmission project list, which covers Texas only so far: owner, endpoints, voltage, miles and in-service dates. Filter by state, status, capacity, voltage or in-service window, or ask only for projects first seen since a date; the initial load is never counted as new. Every row cites its source and license. The tool is free with or without a key, and it is in the grid and gas tool packs. New power generation stays in get_power_pipeline, and existing pipelines and lines stay in get_infrastructure.
Connect an MCP client →Pocket listings are now Capacity Source, built for enterprise buyers and the AI agents that procure for them: powered land, powered shells and turnkey capacity to buy or lease, with every listing showing when it was last updated. It is SEARCHABLE BY SIZE — a floor in kilowatts or megawatts — AND BY LOCATION: region such as North America or Europe, country, US state, or metro. Anyone can browse the cards; a free account opens a listing, and an agent opens one with source_capacity using a key with its human's email bound, or an OAuth connection. Register with request_capacity_intro and DC Hub sends the provider only your company and requirement; the provider then accepts or declines, and identity and contacts are exchanged only on acceptance. Every registered lead has a public verification record. Agents that learned the old tool names keep working. While the first listings are onboarded, register a requirement to get first access.
Open Capacity Source →The menu now opens on the question DC Hub is built to answer: how much power is available at a specific site, and when. Land & Power brings the map, the Power Index, grid intelligence, the interconnection queue and the site-selection tools together. AI and What's New now sit in the main menu, and the agent tools are grouped under MCP. Every page that was in the menu is still in it.
Open the Land & Power map →/mcp/siting is an MCP endpoint on dchub.cloud that lists the siting tools, including execute_plan, find_sites, site analysis and comparison, site and market ranking, composite site scores, fiber readiness, water, disaster and climate risk, and tax incentives. Point an agent at it when the job is choosing where to build.
/mcp/grid lists the power tools, including grid intelligence, interconnection queues, hosting capacity, power-availability timelines, the power pipeline, retirement headroom, energy prices, ISO comparisons and live grid telemetry. Built for agents answering how much power a site can get, and when.
/mcp/fiber lists the connectivity tools, including long-haul and metro fiber, fiber readiness at a site, lead-in route planning, peering and internet exchanges, subsea cables, and latency clustering of candidate sites.
/mcp/deals lists the transaction tools, including M&A and capital transactions, hyperscaler AI deals, deal autopsies, the construction pipeline, news and intelligence search, market intelligence and market trajectory forecasts.
/mcp/gas lists the gas tools, including gas access and pipeline intelligence, delivered gas economics, gas pricing, energy prices and the power pipeline, for agents weighing gas-fired or behind-the-meter power against the grid.
/mcp/site lists the lookup tools for a single facility or market, including search and fetch, facility records, a market's power-index rank and its market context.
/mcp/deepresearch exposes only search and fetch, in the connector format OpenAI Deep Research and ChatGPT expect: find data-center facility records, then fetch a citable public summary of each.
Every MCP capability on one page, read from the running server on every load — newest first, with the authoritative upstream sources named, provenance on every response and explicit unknowns instead of confident blanks. The catalog cannot describe a capability the server does not serve.
Browse the capabilities →A playground for the live layers: search the facility inventory, rank power markets by the index, pull the biggest movers and the latest industry news — straight from the browser, no account. When the answers earn it, connect your own AI to the same endpoint you just tried.
Open the playground →We asked the leading models to rank the data-center intelligence field against the incumbents they already know, and published their verdicts word for word — dated, with the prompts. Read what the agents' own models concluded before an agent ever calls a tool here.
Read the verdicts →Every market the index scores now carries its own daily brief — power verdict, grid context, pipeline, deals, tenants and news for that metro, rebuilt from the live layers on every read. The sections an analyst would assemble, generated from the data instead of about it.
Open a market brief →One brief per hyperscaler: announced megawatts, capex, power agreements, acquisitions, ISO exposure and water context, assembled from the same deal and infrastructure layers the MCP tools query — so an agent and its human read the same pipeline.
Read a hyperscaler brief →Live, cited analysis of the infrastructure behind AI: power-market rankings, grid headroom and time-to-power, deal flow and tenant briefs — written from the live layers, refreshed on cadence, every figure carrying its source.
Read the research →Site valuation compares grid power, behind-the-meter gas and gas-to-grid into a net-present-value view, layered with the power verdict and site-readiness premiums, beside comparable sales. Subscriber tooling, and the paywall states what it withholds.
Value a site →One endpoint, every agent: a walkthrough for adding DC Hub as a custom MCP connector on ChatGPT, Microsoft Copilot Studio, Perplexity, Gemini Enterprise, Le Chat and Grok. The setup page an agent can hand its human.
Add the connector →The standing page tracks the registries that list the MCP server, the platforms actively connected, and the search rankings behind the claims — each figure bound at render from the live sources, with the measurement dated. The receipts behind the data layer agents already use.
See the standing →Every shipped capability on one page, rendered from the same PR-approved feed that powers these cards — each card names its basis and binds its figures at render time, so the product page cannot claim what the platform does not serve.
Browse the product surface →A public intake for bugs, feature requests, wrong figures and broken links. The brain triages every submission: trivial fixes ship same-day, judgment calls queue for review — and the page shows recent feedback beside what already got fixed because of it.
Drop feedback →platform array,
one object per update, each carrying its title, body, date and the tool it relates to.
A card carries a metric token, never a figure: values are bound at render time
from /api/v1/canon/phrases, so no number on this page
can be older than the request that drew it. Read the JSON rather than this page's text.
Data centers, deals, transmission, gas pipelines, fiber routes, middle-mile and midstream. Each layer shows the status behind its number — whether it grew, was rewritten in place, is quiet on schedule, or is stuck — and says why. Counts are re-baselined after every ingest; growth is measured day over day.
items array, one
object per layer. Each entry carries its total, what it added over
window_days, a status, the status_reason that
produced that status, a known_issue where an audit finding is open, and
count_captured_at — the moment the count was taken, which is not the same as
the day it belongs to. A layer that could not be measured returns the reason it could
not, never a zero. The status VOCABULARY IS THE SERVER'S, not this page's:
the values rendered with their own colour today are growing,
refreshed, on_cadence, measuring,
unmeasurable, unjudged, overdue and
flat, and any value outside that set still renders, in a neutral tone —
so treat this list as what the page recognises, never as what the field can contain,
and read status_reason for the meaning. For what each population is and
the schema behind every figure, see /bind.
Subscribe on Substack to get DC Hub updates by email.
Source: DC Hub (dchub.cloud), CC-BY-4.0 · counts are live DB snapshots —
a global data-center facility map,
published as facilities_distinct / facilities_records,
plus unified grid, gas and fiber layers (HIFLD / FCC / EIA).
See the daily brief at /daily, explore the live map, or query it via MCP.