China's Data-Centre Map: 210 New PowerlandMap Records in a Day, and the Limits That Matter

A data-centre database becomes useful when it can move from a country-level headline to a defensible list of assets, locations, sources and open questions. China is an especially demanding test. Its computing infrastructure is enormous, its growth is uneven across regions, and a facility, campus, building, phase and computing hub are not interchangeable units.

On 23 October 2026, 210 records were created in PowerlandMap's China data-centre inventory. A direct read of the production database on 24 October shows 467 China records, versus 257 records created before the 23 October enrichment. That is an 81.7% increase in the number of records in this inventory, not an 81.7% increase in China's number of data centres, operating capacity, electricity use or investment.

This distinction is the starting point for a useful investment decision. A broader candidate set improves discovery, but the decision to include a location in a development, power-origination or acquisition shortlist depends on the quality and comparability of evidence attached to each record.

What the PowerlandMap database actually shows

The following figures come from the PowerlandMap production supply inventory, filtered to `country_code = 'CN'` and `demo = false`. The unit is an inventory record. A record may refer to a facility, building, phase or campus and must not automatically be interpreted as a unique, independent physical data-centre site.

Method and limitations. The figures above are a direct database snapshot, not an external market survey. Creation date tracks when a record entered PowerlandMap; it is not the construction or announcement date of a facility. The population excludes demo records. We counted rows, not unique campuses, and did not sum MW from incomparable fields. Blank values remain unknown. Published source-link fields may repeat across many records and do not prove site-level correspondence without checking the document.

Where the inventory is concentrated

The current distribution has 78 records in Shanghai, 56 in Guangdong, 41 in Inner Mongolia and 38 in Jiangsu. Those four region labels account for 213 of 467 records, or 45.6% of the inventory.

The additions on 23 October were especially concentrated in Shanghai (50), Guangdong (24), Beijing (16) and Jiangsu (14). Together those four regions supplied 104 of the 210 new records, or 49.5%.

These are descriptions of PowerlandMap's documented coverage, not rankings of China's physical capacity. A dense cluster of records could reflect a mature colocation market, a concentrated operator footprint, better public disclosure, more accessible local-language sources, a particular import batch—or several of these factors at once. An investor treating record counts as market share would get a misleading result.

The contrast between coastal digital demand and inland energy-led computing hubs is real as a policy and infrastructure issue, but must be validated with separate evidence. In its 2026 reporting, China's Ministry of Industry and Information Technology reported national intelligent computing power of 2,185 EFLOPS as of the end of June 2026, up 177% year on year. This is a national computing-power measure, not an estimate of IT megawatts or utilisation for the 467 PowerlandMap records.

Source: MIIT, first-half 2026 computing infrastructure briefing.

The major gap is capacity evidence, not the number of names

A long list of facilities is an excellent starting point for market intelligence, but it is not yet a power model.

Only 14 of 467 China records (3.0%) currently contain an explicit IT MW field. In the 210 records created on 23 October, none contains a populated IT MW field. It would be wrong to sum these records into a claim of newly identified hundreds or thousands of MW IT.

The stage taxonomy also needs reconciliation. 153 records have the exact stage value `Operational`, while 134 say `Unknown` and 69 say `Unverified`. Other labels include development, programme, construction, announcement and more specific descriptions. We do not convert the 153 labelled operational records into a verified national operational count without checking source dates, site identity and the meaning of each stage.

This is why the platform needs to keep several distinct dimensions: asset identity, geographic precision, source evidence, development/operating stage, electrical connection or generation capacity, and confirmed IT capacity. Combining them into one generic `MW` field would make a large database less useful rather than more useful.

A real example illustrates the difference. A China Mobile facility in Harbin was described by a government-hosted operator report dated 27 April 2026 as deploying 209 cabinets with a total power of 2,508 kW for a specific network-cloud room. That is a documented figure for the scope described in the source. It is not an independently verified 2.508 MW IT data-centre campus, nor is it a standard for estimating every data centre in China.

Source: Heilongjiang MIIT administration, Harbin facility report, 27 April 2026.

What a developer, equipment supplier or energy investor can decide

There are three practical uses for this enrichment.

For a data-centre developer, the 210 new entries expand the universe for competitor-site and land-screening work. Their immediate value is as candidate records to check against power access, fiber diversity, industrial zoning, ownership, permitting and actual development stage. The next action is not to assume that every row corresponds to a buildable or operational campus.

For an energy company or battery, generator and UPS supplier, those same entries broaden the geography of potential accounts. But a facility without public MW, confirmed activity or a credible owner should receive a different qualification grade from one with a live site, named operator and substantiated technical demand. An equipment vendor can use the data to prioritize research, not to assume a purchase order.

For an investor, the distinction between an existing building, a campus phase and a new project matters to both transaction perimeter and comparables. Counting several buildings in one campus as unrelated investments inflates the addressable universe and introduces duplicate diligence effort.

A simple illustrative effort sensitivity shows why screening depth should follow evidence quality:

This is not a market price survey, PowerlandMap customer ROI claim or savings forecast. It is a transparent workload model: 210 × minutes per record ÷ 60, multiplied by an assumed €100/hour. The real economic benefit of a documented inventory depends on the quality of filtering, the initial material already available and how many facilities ultimately justify deeper diligence. A different rate, review time or shortlist threshold changes the result linearly.

A broader database is valuable only if its uncertainty is visible

The strongest counterargument to celebrating a large single-day enrichment is that more records can also increase noise. That objection is correct. A list of 467 names has limited value if its user cannot distinguish a verified facility from a campus placeholder or an archived plan.

The enrichment should therefore be treated as two separate achievements: first, the discovery and consolidation of candidate records; second, the more demanding process of enriching each with enough publicly verifiable, dated evidence to support commercial or technical decisions. On 24 October, the first step has moved materially. The second remains work in progress, as the coordinates, capacity and stage coverage figures show.

Nor should our 467-record sample be compared directly with national statistics expressed in standardized cabinets or computational throughput. China's Ministry of Industry and Information Technology reported over 13.73 million standard racks in use at end-2025, a national infrastructure measure with a completely different denominator. These figures provide evidence of the scale of the national ecosystem, not a way to estimate PowerlandMap's percentage coverage.

Source: MIIT commentary on green computing infrastructure, September 2026.

The next stage of PowerlandMap's China work

The immediate priorities are concrete: reconcile campus, building and phase identities; strengthen operator and official-document evidence; resolve ambiguous development stages; geolocate records only to the precision supported by evidence; and connect facilities with sourced electricity and fiber network infrastructure where records permit. For capacity benchmarking, the priority is to identify confirmed MW IT or clearly scoped electrical values rather than infer a headline GW number from incomplete fields.

A credible market map is not a contest to display the largest number. It should allow a user to move quickly from a country to a region, from a region to a candidate, and from a candidate to an evidence-backed question.

PowerlandMap's China inventory now contains 467 records after 210 additions in one day. The more important measure of progress will be how many of those records become sufficiently verified to support a real commercial, financial or engineering decision.

Explore PowerlandMap for infrastructure coverage and market intelligence, and contact us for a scoped view of China data-centre and power-market qualification.

*Method note: Production database snapshot checked 24 October 2026. Country CN, non-demo rows. Counts reflect available records and their current fields. No confidential client data, non-public project names or unverified national MW totals are used. Illustrative costs are assumptions, not observed prices.*