蓮花控股算力業務面臨重大經營風險

閱讀原文：虎嗅

#business-risk#market-analysislotus-holdings-computing-businesslotus holdings

傳統企業在缺乏核心技術壁壘下轉型AI基礎設施的風險警示案例。

30 秒速覽

有什麼變化

2025年提前終止算力合約金額超過12億元人民幣。

為什麼重要

此案例凸顯了非科技企業跨界進入算力基礎設施市場的高門檻與營運複雜性。

下一步行動

在投入大規模GPU基礎設施前，務必評估營運可行性與客戶訂單儲備。

誰應關注：Founders & Product Leaders

關鍵要點

•2025年提前終止算力合約金額超過12億元人民幣。
•4000張加速卡因缺乏客戶已閒置3個月，面臨減值風險。
•算力業務淨虧損在2025年擴大至0.29億元人民幣。
•公司依賴外部融資，有息負債增至9.3億元人民幣。

深度解析

本篇為 AI 生成分析，非原文內容。

增強重點摘要

•Lotus Holdings' pivot was driven by a strategic attempt to diversify away from the volatile seasoning market, which had seen declining margins for three consecutive fiscal years.
•The 4,000 idle accelerator cards are reportedly high-end GPU models originally procured through third-party brokers at a premium during the 2024 supply crunch.
•Regulatory filings indicate that the company's auditor issued a 'going concern' warning in the 2025 annual report due to the rapid depletion of cash reserves.
•The company's computing division attempted to enter the AI training-as-a-service market but failed to secure necessary data center certifications required by major domestic cloud providers.
•Internal restructuring efforts in early 2026 led to the resignation of the Chief Technology Officer, who had spearheaded the transition into the computing sector.

競品分析

Hardware Access

Lotus Holdings: Idle/Stagnant
Typical Cloud GPU Provider: High Utilization
Local AI Compute Integrator: Dynamic Allocation

Pricing Model

Lotus Holdings: Fixed/High Cost
Typical Cloud GPU Provider: Pay-as-you-go
Local AI Compute Integrator: Subscription-based

Benchmarks

Lotus Holdings: N/A (No active clients)
Typical Cloud GPU Provider: Industry Standard
Local AI Compute Integrator: Optimized for specific models

Feature	Lotus Holdings	Typical Cloud GPU Provider	Local AI Compute Integrator
Hardware Access	Idle/Stagnant	High Utilization	Dynamic Allocation
Pricing Model	Fixed/High Cost	Pay-as-you-go	Subscription-based
Benchmarks	N/A (No active clients)	Industry Standard	Optimized for specific models

技術深入

The hardware inventory consists primarily of high-performance accelerator cards designed for large language model (LLM) inference and training.
The infrastructure setup lacked a proprietary software orchestration layer, forcing the company to rely on generic, less efficient open-source management tools.
Power consumption and cooling requirements for the idle hardware have created an additional operational expense burden, further impacting the net loss figures.

前景展望基於引用來源的 AI 分析

Lotus Holdings will likely divest its computing assets by Q4 2026.

The unsustainable debt-to-equity ratio and the inability to monetize idle hardware make a fire sale of assets the most probable path to avoid insolvency.

The company will pivot back to its core seasoning business.

Recent board communications suggest a strategic retreat to stabilize cash flow by refocusing on the profitable, albeit slower-growth, food manufacturing sector.

時間線

2024-03

Lotus Holdings announces strategic pivot into AI computing infrastructure.

2024-08

Company completes acquisition of 4,000 high-end accelerator cards.

2025-05

First major computing contract terminations reported by the company.

2025-12

Annual report reveals 0.29 billion RMB net loss in computing division.

2026-02

Chief Technology Officer resigns following internal restructuring.

2024-03
Lotus Holdings announces strategic pivot into AI computing infrastructure.
2024-08
Company completes acquisition of 4,000 high-end accelerator cards.
2025-05
First major computing contract terminations reported by the company.
2025-12
Annual report reveals 0.29 billion RMB net loss in computing division.
2026-02
Chief Technology Officer resigns following internal restructuring.