Why are VNET and CATL working together—and what could it mean for AI infrastructure?
The VNET–CATL relationship is a useful test of whether AI data centres and energy infrastructure are beginning to converge. This research map separates confirmed ownership and commercial evidence from inference, then asks what batteries, power systems, financing and customer demand would need to contribute for the relationship to create real economic value.
Why this may be much more than batteries plus data centres: a research map of the physical system, customer ecosystem and capital architecture — with confirmed evidence separated from inference and thesis.
Compute–energy coordination raises the stakes, not the level of proof.
China’s renewed focus on computing–electricity coordination and VNET’s identified Dongying grid dependency sharpen the strategic relevance of CATL-linked energy expertise. The completed CATL-affiliated share acquisition remains an ownership fact, not proof that CATL finances VNET’s new campuses or supplies a particular energy system.
CATL-affiliated VNET acquisition fully completed
The remaining 70% tranche closed on 21 September, following the first 30% on 24 August. Acquisition-completion risk is resolved; the commercial reason for accelerating the final closing remains undisclosed.
Thesis implication: stronger evidence of strategic follow-through. This secondary share transfer adds no new cash to VNET and does not confirm project funding or CATL-ecosystem customer contracts.
Timeline, sources and what changes → · HKEX completion filing
What completion changes in the VNET–CATL thesis
22 September context: Alibaba’s 20 GW ambition puts compute and energy together at the centre of the delivery question. This strengthens the relevance of investigating VNET–CATL’s energy economics, without confirming a new commercial outcome. Read the compute–energy interpretation →
The ownership transaction is now established. The next question is how the relationship translates into funded infrastructure, energy economics and customer demand.
Established
The full agreed share sale has completed. This is a secondary transfer, so the proceeds do not become new cash on VNET’s balance sheet.
Our interpretation
Completion strengthens the evidence of strategic follow-through. It does not identify why the final tranche closed early or establish a commercial catalyst.
The next evidence
Disclosed project capital, asset ownership and return allocation; contracted power or storage arrangements; and signed compute demand from CATL or its wider ecosystem.
Customer distinction: CATL’s own compute needs and demand from companies in its wider industrial ecosystem are separate potential channels. Neither is confirmed by the share transfer.
Model treatment: completion is an observed fact, not a probability. The earlier funding and customer scenarios below remain hypotheses; this update does not mechanically increase their probabilities or valuation outputs.
Read the filing timeline, timing investigation and sources and context →On this page
← VNET THESISSYSTEMECOSYSTEMMARKET GAPFINANCINGEVIDENCEWHAT TO PROVEVNET has the AI load. CATL has the energy system.
The simplest way to understand the partnership is that each company controls something the other increasingly needs. AI infrastructure is becoming an electricity problem as much as a real-estate or server problem.
GW-scale AIDC sites, operating expertise, network reach, customer relationships and an expanding domestic/overseas resource pipeline.
Storage, zero-carbon energy technology, industrial scale, supply-chain leverage and an expanding position in the electrical layer around AI infrastructure.
Together: electricity → compute → tokens
The companies explicitly describe a three-layer ambition: GW-scale compute-energy facilities → distributed compute-energy networks → a zero-carbon token ecosystem. Our thesis asks what that architecture could mean economically if execution begins to match the language.
The relationship has several possible layers.
ENERGY
Green generation / direct green power, storage and grid interaction.
POWER + FACILITY
HVDC / 800V direction, electrical conversion, cooling and high-density AIDC.
COMPUTE + NETWORK
GW campuses connected into increasingly distributed compute infrastructure.
CATL ITSELF
Could consume compute, but we do not currently have evidence that CATL will become a material VNET wholesale customer.
CATL ECOSYSTEM
The larger hypothesis: industrial, energy, manufacturing and technology companies around CATL may create digital-infrastructure opportunities for VNET.
VNET CUSTOMER BASE
VNET provides a route for CATL's integrated energy solutions into a large existing data-centre and enterprise footprint.
STRATEGIC ALIGNMENT
CATL-affiliated investors create long-duration strategic alignment, but the share purchase is not equivalent to fresh VNET capex funding.
PROJECT BANKABILITY
VNET says it acquires land initially and begins M&E fit-out only after firm customer orders — important evidence for how expansion risk is controlled.
CAPITAL RECYCLING
REIT monetisation creates a potential develop → stabilise → recycle → redeploy loop.
The opportunity may be the ecosystem around CATL — not simply CATL.
A narrow thesis asks whether CATL buys batteries for VNET or becomes a compute customer. A broader thesis asks whether the partnership creates reciprocal distribution: VNET exposes CATL's energy architecture to AI infrastructure, while CATL's industrial ecosystem creates routes for VNET's compute and digital infrastructure.
What the market sees — and what may still be underappreciated.
VISIBLE
CATL-linked strategic investors; AI demand; VNET MW growth; batteries/storage as a natural data-centre adjacency.
UNDERAPPRECIATED?
The announced relationship is explicitly about compute-energy integration, direct green power and an integrated three-layer architecture — not simply equipment procurement.
OUR RESEARCH
CATL is moving from energy toward the electrical layer around AI, while VNET is moving from IDC toward power-aware, distributed AIDC. The trajectories appear to be converging from opposite directions.
THESIS
The strategic position may ultimately be better understood as participation in an emerging AI compute-energy production system than as a conventional battery-company investment in a data-centre operator.
22 SEPTEMBER 2026 · THESIS UPDATE
The CATL-linked ownership block has explicit governance terms
VNET’s 21 September release confirms that PJ Millennium I and II acquired 650,424,192 Class A ordinary shares, approximately 38.1% of outstanding ordinary shares on the disclosed June share count. The buyer structure runs through PJ Millennium Partnership and Lochpine; VNET describes Lochpine Capital as a non-controlled, non-consolidated CATL affiliate. The figure is an economic ownership measure, not a statement of CATL’s direct ownership or voting control.
The investor-rights and founder voting/consortium agreements became effective at closing. The disclosed terms provide for founder-directed voting of certain acquired shares, transfer restrictions for a specified period and support for stability of control. These are substantive governance arrangements, not merely a passive portfolio holding.
Thesis implication: explicit governance alignment strengthens the relationship evidence. The roughly US$942 million secondary purchase consideration went to existing sellers; it is not a VNET capital raise or a CATL project-finance commitment.
VNET closing release · Original investor-rights and voting terms
Research change record →The 38.1% block combines CATL corporate capital with CATL founder-family capital.
The legal distinction still matters: PJ Millennium is the buyer and CATL does not directly own the entire VNET block. But the economic ownership chain is more strategically aligned than a generic consortium label suggests.
71% · Green Alchemist
Green Alchemist holds 71% of PJ Millennium’s economic interest. Its disclosed ultimate beneficial owner is Zeng Yuntian (曾云天); Chinese prospectus evidence identifies Zeng Yuntian as the son of CATL founder and chairman Zeng Yuqun (Robin Zeng).
29% · CATL Hong Kong
CATL’s Hong Kong entity holds the remaining 29% economic interest. On the governance side, PJ Millennium’s GP runs through Lochpine Capital, in which CATL Investment has a disclosed 45% ownership interest.
Green Alchemist 71%; CATL Hong Kong 29%; Zeng Yuntian is Green Alchemist’s disclosed UBO; Lochpine ownership includes CATL Investment 45%, Wang Hongbo 35% and Yee Chun Keung 20%.
Chinese prospectus evidence identifies Zeng Yuntian as Robin Zeng’s son. We therefore describe the 71% interest as CATL founder-family capital while preserving Green Alchemist as the legal/economic vehicle.
This does not make Green Alchemist legally CATL, does not establish CATL control of the full 38.1% VNET block, and does not prove a CATL or Zeng-family commitment to fund VNET’s future data-centre projects.
Why it changes the thesis: the other side of the consortium is not best understood as an unrelated financial investor. The disclosed economics place most of PJ Millennium alongside the CATL founder’s family, with CATL corporate capital alongside it. That makes the strategic relationship more significant, while leaving future project financing as a hypothesis to be proved.
The capital architecture may matter almost as much as the physical architecture.
Multiple sources
VNET debt/liquidity
REIT recycling
strategic alignment
project / energy capital?
customer commitments
De-risked expansion
Land/resource reserve → firm order → M&E fit-out → delivery → utilisation.
Cash-flow loop
Operating cash flow → mature asset monetisation / REIT → capital redeployed into the next AIDC build.
VNET reported RMB8.14bn of new debt, refinancing, equity and other financing in Q1 2026 and explicitly called its two REIT listings a scalable capital-recycling model. In Q2 management also said it initially uses its own funds for land and starts mechanical/electrical fit-out only after firm customer orders.
Evidence-weighted research assessments.
These are analyst-style judgement ranges, not statistical forecasts. They force us to state what the evidence currently supports and should move as new contracts, financing structures and customers are disclosed.
21 September update: ownership acquisition is fully completed. The ranges below remain earlier subjective estimates for future commercial outcomes, not completion probabilities, disclosed commitments or mechanically revised forecasts. Evidence update →
Relationship materially extends beyond equipment procurement
90–95%Direct support: strategic agreement, three-layer architecture, technology/infrastructure/supply-chain cooperation.
CATL technology / storage becomes material inside VNET ecosystem
75–90%Strong industrial logic and explicit compute-energy partnership; project-level disclosure remains limited.
Joint compute-energy infrastructure projects emerge
70–85%The announced objective points directly here, subject to definitive project agreements and execution.
REIT / asset recycling becomes a recurring material funding mechanism
65–80%Already operating; key unknown is scale and repeatability across future AIDC assets.
CATL relationship lowers VNET's effective funding constraints
50–70%Strategic alignment is clear; direct CATL project/vendor financing has not been publicly established.
CATL ecosystem generates identifiable VNET customer/projects
40–60%Strategically plausible, especially industrial/overseas; currently lacks disclosed customer-level evidence.
Dedicated JV / SPV / project-finance structures emerge
35–55%Economically logical at GW scale, but no disclosed structure yet.
CATL itself becomes a material VNET compute customer
20–40%Possible, but weaker than the broader CATL-ecosystem thesis and not required for the partnership to succeed.
The economic model changes if more of the thesis proves true.
DEMAND QUALITY
Strategic ecosystems could widen customer channels and improve visibility beyond conventional colocation demand.
CAPITAL INTENSITY
Customer-backed build, asset recycling and energy/project capital could reduce the amount of growth VNET must permanently carry on balance sheet.
ASSET VALUE
A node combining power access, storage, cooling and distributed networking may become scarcer than a conventional powered building.
POWER ECONOMICS
Energy moves from a simple operating cost toward an optimisable input across storage, grid interaction and workload placement.
NETWORK EFFECT
Distributed compute may make separated campuses more useful as a common infrastructure resource.
VALUATION QUESTION
If economics migrate beyond renting MW, the long-run question becomes whether a conventional IDC framework captures the whole business.
What is confirmed versus what we are testing.
Three-layer compute-energy ecosystem; green DC/direct green power; technology, infrastructure and supply-chain cooperation; CATL-linked strategic investors; VNET REIT recycling; order-gated M&E fit-out.
CATL is expanding into the AIDC electrical stack, including its 49% investment in Zhongheng's controlling shareholder; VNET's strategy is increasingly power-aware and GW-scale.
The two trajectories are converging into a broader compute-energy platform whose economic value could exceed a supplier/customer relationship.
CATL ecosystem customer referrals; direct CATL project/vendor financing; dedicated JVs/SPVs; separate energy-asset ownership; CATL as a material compute customer.
The research agenda.
Watch for named joint projects, ownership of storage and electrical assets, financing counterparties, CATL ecosystem customers, customer prepayments or guarantees, overseas co-development, REIT eligibility of new AIDC assets, and evidence that distributed compute-energy networking changes utilisation or margins.