Good news fully priced in, yet plunges over 17%: Why did VNET's earnings report miss estimates despite backing from CATL?

On August 18, ahead of US market open, VNET Group (VNET.US) released two key updates: Q2 earnings showed total revenue of RMB 2.78bn (+14.2% YoY), with adjusted EBITDA up 25.4% YoY, and wholesale IDC operational capacity surpassed 1GW for the first time. On the same day, the company announced a strategic partnership with CATL to build a three-tier system comprising gigawatt-scale computing-power plants, distributed computing-power networks, and a zero-carbon token ecosystem.

The market did not buy it. The stock opened lower, briefly spiked to $7.54 early in trading before oscillating downward, hitting an intraday low of $6.52 with a max drawdown of over 17%. As of 13:38 ET on August 18 (1:00 AM Beijing time on the 19th), shares were at $6.58, down ~17%, still trading. Q2 EPS was a loss of $0.08, missing the consensus estimate of a $0.06 loss; full-year guidance remained unchanged. Capital that had hoped for an upward revision exited, and many investors who bet on CATL's shareholding bringing incremental funds cashed out after the deal landed.

EBITDA rose, but why is gross margin falling?

Looking solely at adjusted EBITDA, this report is bright enough: profit margin expanded 3 percentage points YoY to 33.0%, growing far faster than revenue. However, breaking down the income statement reveals the growth structure is not solid.

Q2 gross profit was RMB 505mn, down 7.8% YoY and 18% QoQ from RMB 616mn in Q1; gross margin narrowed from 22.5% YoY to 18.2%. The biggest profit drainer was depreciation: total depreciation and amortization reached ~RMB 684mn (+24.3% YoY), with ~RMB 657mn (+~28% YoY) allocated to COGS, a growth rate nearly double that of revenue. In heavy-asset expansion cycles, depreciation rising before revenue hits is industry norm, but it has indeed obscured true profitability behind the numbers.

Even adding back depreciation to assess cash profitability, adjusted cash gross margin was only 41.8%, below the 43.6% YoY and a 3.2pp drop from Q1's 45.0%. While revenue scale is expanding, unit cash gross margin is contracting. Co-CEO Teng Teng explained on the earnings call that surging power consumption in Q2, settled via a pass-through mechanism, squeezed cash margins; additionally, a one-time gain in Q1 raised the comparison base. This means the rise in variable costs like power and O&M is real, and pressure comes from more than just depreciation.

The expansion of EBITDA margin relied heavily on expense compression. Q2 admin expenses fell from RMB 213mn YoY to RMB 165mn, driving overall opex down by nearly RMB 100mn. However, R&D expenses grew rigidly, with cost cuts concentrated in management and sales, leaving limited room for further compression.

Interest burden is also weighing heavier. Q2 net interest expense grew 33.6% YoY. As of end-June, interest-bearing debt stood at ~RMB 23.4bn (excluding lease liabilities), with cash and short-term investments of ~RMB 7.2bn. According to the debt maturity structure disclosed at end-2025, ~46.6% of debt matures between 2026 and 2028, posing significant refinancing rollover pressure over the next three years.

From a net profit attributable to shareholders perspective, H1 cumulative loss was ~RMB 667mn, but the loss structure varied greatly: Q1 losses were driven mainly by one-time income taxes related to capital transactions, with limited operating loss contribution; Q2 attributable loss narrowed to RMB 136mn, with adjusted metrics showing a marginal profit of RMB 7.4mn. Profitability is just showing signs of life, but cash flow hasn't caught up—Q2 net cash from operating activities was RMB 218mn, down 40.5% YoY, mainly due to increased working capital absorption.

In terms of business structure, wholesale and retail have diverged into two lines. Wholesale IDC revenue share rose to 39.8% of total revenue, exceeding retail for two consecutive quarters; retail IDC revenue grew 9.1% YoY, with average monthly recurring revenue per rack reaching RMB 9,799, but active racks dropped from 52.1k YoY to 50.1k, with utilization at 64.5%, indicating the business has entered a stock-operation phase. Non-IDC revenue grew a meager 1.1%, almost stagnating.

Behind 1GW capacity, utilization is fading

Wholesale IDC operational capacity surpassing 1GW was the most watched number in this report. Over the past year, wholesale capacity grew from 674MW to 1,007MW, a 49.4% YoY increase, ranking among the fastest-expanding third-party IDC vendors in China.

The cost of rapid expansion is loosening utilization. As of end-Q2, overall wholesale IDC utilization was 73.9%, down 2pp from 75.9% YoY; mature project utilization fell from 94.6% to 92.5%, while ramp-up project utilization dropped from 45.0% in Q1 to 36.6%.

Teng Teng explained on the call that the decline in ramp-up utilization was mainly due to denominator effects—Q2 saw 100MW of new deliveries, pushing many new projects into ramp-up cycles where customer move-in typically takes 3-6 months. He stated that move-in pace is expected to improve marginally in H2, aided by the release of domestic chip capacity driving customer deployment. This explanation holds merit, but at least from Q2 data, capacity delivery pace outstripped customer move-in pace.

Signing rates also retreated. Current signed rate for operational wholesale capacity is 96.3%, down from 100% YoY.

Order backlog remains ample. In Q2, the company secured a 345MW wholesale order from a top cloud vendor in the Beijing area, adding 862MW cumulatively YTD. Including reserved capacity, total orders and reservations exceed 1.2GW. Of the 585MW under construction, the overall pre-signing rate is 94.2%, with ~333MW planned for H2 delivery this year and 252MW for H1 next year, primarily from the Ulanqab park. The weighted average remaining lease term for signed capacity is ~7 years, with over 90% of wholesale revenue being recurring, providing a floor for future growth.

However, delivery peaks are concentrated in H2 this year and H1 next year, meaning concentrated capacity rollout may further drag down overall utilization. The slowdown in wholesale business is clear: revenue growth was 112.5% YoY last year, 58.1% in Q1 this year, and 29.3% in Q2. High base effect is the core reason, alongside fluctuating capex rhythms of cloud vendors and the shift of AI compute demand from explosive growth to steady implementation, which is a common industry environment. Teng Teng noted that multiple top firms are expected to release GW-level tenders in 2026, but power quotas and chip supply remain bottlenecks to effective supply, with supply-demand mismatch likely persisting until around 2028.

Comparing with peer leader GDS Holdings, its Q2 revenue grew only 6.5%, adjusted EBITDA grew 2.5%, and EBITDA margin was 45.5%, significantly higher than VNET. However, GDS's current book net profit of RMB 838mn included large equity method investment gains from overseas financing dilution; roughly excluding non-recurring items, the core business is actually in a loss state. Income statements of top IDC firms are generally mixed with non-operating items, making single-metric analysis prone to distortion.

Capacity expansion pace is accelerating. In Q2, the company added ~1.4GW of land reserves, including 908MW domestically (mainly Inner Mongolia and East China) and 478MW overseas, pushing total operational plus reserve capacity past 4GW. The first overseas project is expected to land in Southeast Asia, with evaluations ongoing for Middle East and Europe opportunities. The company stated it will remain cautious, using own funds to secure land first, and only starting MEP construction after securing definite customer orders. Under the Hyperscale 2.0 strategy, green power direct connection and computing-power integration have become new competitive directions, and the core logic for binding with CATL.

CATL's RMB 6.4bn didn't enter the listed company

To expand capacity, one must keep throwing money. The company maintained its full-year capex guidance of RMB 10-12bn unchanged. H1 cumulative investment was ~RMB 3.55bn, completing only about 30% of the annual target. This means H2 needs to inject RMB 6.45-8.45bn, significantly accelerating spending rhythm. Full-year delivery target is 450-500MW, with only ~118MW completed in H1, meaning over 70% of capacity is concentrated in H2, bringing immediate pressure on both construction and fundraising.

Current financing paths for the company are mainly three: bank debt, asset securitization, and equity financing. Q2 saw RMB 3.77bn in new financing across categories, with debt remaining the primary source. From Nov 2025 to Mar 2026, the company issued three tranches of data center inter-institutional REITs, raising over RMB 7.2bn in total, accounting for over 80% of the national market for similar products. However, issuance amounts included assumption of existing project-level debt, so actual cash flowing back to the parent company was less than the total issuance. The company stated on the Q1 call that the goal to recoup cash via REITs in 2026 is no less than RMB 2bn.

The most market-watched CATL shareholding did not bring direct capital supplementation to the listed entity. In May, CATL affiliates acquired 650 million shares held by Shandong Hi-Speed Holding for ~$942mn (~RMB 6.4bn), resulting in a ~38.1% stake, expected to settle in Q4. This transaction was an old-share transfer, with all funds flowing to the seller Shandong Hi-Speed; the listed company received no incremental capital. Simultaneously, both parties signed a voting rights entrustment agreement, with CATL affiliates entrusting all voting rights to founder Chen Sheng, keeping actual control unchanged. CATL enters as a strategic shareholder, with core 诉求 pointing to industrial synergy in computing-power integration rather than controlling operations.

The strategic cooperation agreement signed on Aug 18 marks the formal start of synergy. Within the three-tier system, the GW-scale computing-power plant and distributed network focus on energy storage tech and green power direct supply, aligning with the core need for data center cost reduction and efficiency improvement; the zero-carbon token ecosystem lacks specific definition in the announcement, and Teng Teng provided no timeline or financial targets on the call, leaning more towards long-term narrative.

Computing-power synergy is CATL's recent core layout direction. On April 8, Zhongheng Electric announced its controlling shareholder signed a framework agreement with CATL, which intends to subscribe to RMB 4.1bn of newly increased registered capital; on Aug 14, both parties formally signed the 'Capital Increase Agreement', involving RMB 3.512bn in cash and 49% equity of Times Tianyuan valued at RMB 588mn, after which CATL holds 49% of the controlling shareholder. Additionally, CATL participated in financing for large model enterprises, placing bets from upstream power equipment to midstream compute operation and downstream applications. For VNET, CATL's endorsement helps lower financing costs and access industrial resources, but converting cooperation into actual performance takes time. On July 23, the company also signed a strategic cooperation with Haier Group, covering green data centers, liquid cooling technology, etc.

After the earnings report, the stock fell like this, and the market attitude is written clearly: it acknowledges the expansion speed and order backlog, but has no confidence in growth quality, cash flow pressure, or the realization pace of industrial cooperation. H2 faces concentrated capacity delivery on one side and a multi-billion funding gap on the other, plus the pending CATL cooperation. The 含金量 (gold content) of this growth answer sheet still needs time to verify.