Is 01.AI profitable in 2026?

No audited profitability data exists. 01.AI reports contracted order volume of about RMB 500M (~$70M) in 2025, rising to a claimed RMB 1.5B (~$210M) for 2026, but these are unaudited company disclosures ahead of its planned 2027 Hong Kong IPO.

Did 01.AI give up on building its own AI models?

Yes. In March 2025, 01.AI stopped pre-training new large foundation models from scratch after DeepSeek's cheap open-weight releases made proprietary training uneconomic, pivoting instead to fine-tuning and deploying third-party models for enterprise and government clients.

Who is 01.AI's biggest competitor?

Zhipu AI (Z.AI), already Hong Kong listed and positioned as 'China's Palantir' with a reported 12,000 enterprise clients as of May 2026, is executing the same enterprise-deployment strategy 01.AI adopted in 2025, at far greater scale.

Could DeepSeek replace 01.AI's business?

DeepSeek already forced 01.AI out of frontier-model training in March 2025 by making pretraining uneconomic, and its V4 Flash model, priced around 2 cents per task as of July 2026, continues to erode margins in the fine-tuning layer 01.AI now depends on.

Is 01.AI going public?

01.AI is targeting a Hong Kong IPO in 2027 and, as of July 2026, is unwinding its offshore holding structure and raising a pre-IPO round, with its first audited annual results expected around the round's close.

How many employees does 01.AI have?

Sources conflict: Bloomberg-sourced reporting from July 2026 puts headcount at approximately 240, described as lean for an AI startup, while other estimates put it lower; no single audited figure has been published by the company.

AI companies and labs · China

01.AI

01.ai ↗

Chinese AI lab that pivoted from building foundation models to deploying them.

Protection under test

Assessed 20 July 2026 · Evidence as of 20 July 2026

Something worth protecting exists and is being tested right now.

On the recordSince 20 July 2026 · Checkable from year-end 2026 · Every call on the record Checkpoint: Pre-IPO funding round expected to close with first audited annual financials disclosed

The call

01.AI now competes as a deployment specialist in a market where Zhipu is already ahead and base-model prices keep falling out from under it.

01.AI used to build cutting-edge AI models but switched in early 2025 to helping Chinese companies and government agencies customize and run AI software on their own computers, after a rival called DeepSeek made that original business too expensive. The company claims growing contracts worth $210 million for 2026, but the work is becoming a commodity: another Chinese AI company called Zhipu is already doing the exact same thing for 12,000 clients and is already publicly traded, while costs for basic AI tasks keep dropping (now around 2 cents per task). The biggest wildcard is whether 01.AI can win bigger, stickier government and defense contracts than Zhipu before the price of AI software falls so low that anyone can do this work.

Why

AI pressure: DeepSeek. DeepSeek's cheap open-weight models ended 01.AI's own frontier-model business and now undercut its fine-tuning margins too. Could AI do the core job? Only partly.

Bigger players: Zhipu AI (Z.AI). Zhipu already runs the same enterprise-deployment playbook at 12,000 clients and is already listed in Hong Kong. Could a giant copy it? Yes, easily.

What protects it. Existing government and enterprise on-premises contracts that are slow and costly for clients to rip out.

What would change this call

01.AI's pre-IPO audited financials (expected late 2026) show gross margin trend. Reveals whether deployment contracts scale like software or stay linear/project-based.
Zhipu AI reports next earnings or client count update. Shows whether the 'China's Palantir' incumbent is pulling further ahead.
01.AI's Hong Kong IPO timing and valuation set in 2027. Tests whether investors see durable differentiation versus Zhipu and Moonshot.

Key numbers

2025 contracted orders: ~$70M (RMB 500M) (2026-06-24)
Zhipu AI: 12,000 enterprise clients (2026-05-29)
DeepSeek V4 Flash: ~2 cents per task (2026-07-20)

Named risks

Deployment-layer business commoditizing as third-party base model prices keep falling monthly. Severity: high
Zhipu already executing identical 'China's Palantir' strategy at far greater scale and already HK-listed. Severity: high
Gartner warns ontology-led approaches don't scale in China's fragmented, project-based IT market. Severity: medium
Project-based contracts create linear, not software-like, economics per company's own admission. Severity: medium
Conflicting headcount and revenue figures across sources undermine pre-IPO financial credibility. Severity: low

Rivals, as read in July 2026

Rival	Direction	Evidence
Zhipu AI (Z.AI)	losing to	Already HK-listed with 12,000 enterprise clients as of May 2026 vs. 01.AI's smaller, unaudited client base
DeepSeek	losing to	Forced 01.AI out of pretraining in March 2025; V4 Flash priced at ~2 cents per task erodes 01.AI's deployment margins
Moonshot AI	losing to	Targeting HK IPO within six months at up to ~$30B valuation after July 2026 Kimi K3 launch, ahead of 01.AI's 2027 timeline
Alibaba	contesting	Investor in 01.AI but also owns Qwen and Alibaba Cloud, with no structural bar to competing directly in enterprise AI
StepFun	contesting	Raising ~$2.5B and restructuring for its own HK IPO as of mid-2026, competing for same investor attention

Questions people ask

Is 01.AI profitable in 2026?: No audited profitability data exists. 01.AI reports contracted order volume of about RMB 500M (~$70M) in 2025, rising to a claimed RMB 1.5B (~$210M) for 2026, but these are unaudited company disclosures ahead of its planned 2027 Hong Kong IPO.
Did 01.AI give up on building its own AI models?: Yes. In March 2025, 01.AI stopped pre-training new large foundation models from scratch after DeepSeek's cheap open-weight releases made proprietary training uneconomic, pivoting instead to fine-tuning and deploying third-party models for enterprise and government clients.
Who is 01.AI's biggest competitor?: Zhipu AI (Z.AI), already Hong Kong listed and positioned as 'China's Palantir' with a reported 12,000 enterprise clients as of May 2026, is executing the same enterprise-deployment strategy 01.AI adopted in 2025, at far greater scale.
Could DeepSeek replace 01.AI's business?: DeepSeek already forced 01.AI out of frontier-model training in March 2025 by making pretraining uneconomic, and its V4 Flash model, priced around 2 cents per task as of July 2026, continues to erode margins in the fine-tuning layer 01.AI now depends on.
Is 01.AI going public?: 01.AI is targeting a Hong Kong IPO in 2027 and, as of July 2026, is unwinding its offshore holding structure and raising a pre-IPO round, with its first audited annual results expected around the round's close.
How many employees does 01.AI have?: Sources conflict: Bloomberg-sourced reporting from July 2026 puts headcount at approximately 240, described as lean for an AI startup, while other estimates put it lower; no single audited figure has been published by the company.

Timeline

2026-07-20Reports say 01.AI targets 2027 HK IPO, unwinding offshore structure, ~240 employees per Bloomberg sourcing
2026-07-19Moonshot AI targets HK IPO within six months at up to ~$30B valuation after Kimi K3 launch
2026-06-30Reporting confirms 01.AI's March 2025 exit from foundation-model pretraining
2026-06-24Kai-Fu Lee discloses 2025 contract volume of ~RMB 500M and 2026 target of ~RMB 1.5B
2026-05-29Zhipu AI (Z.AI) reported at 12,000 enterprise clients and 80M endpoints, HK listed as 'China's Palantir'
2026-01-26Gartner warns Palantir-style ontology approaches struggle to scale in China's fragmented IT market
2025-11-27Latka estimates 01.AI revenue at $20M (2023) growing to $30.5M (2024)
2025-0301.AI stops pre-training new large foundation models, pivots to enterprise deployment
2023-1101.AI reaches $1B+ unicorn valuation after ~$300M Series A
2023-0501.AI founded by Dr. Kai-Fu Lee and Ken Qi in Beijing

Revision history

20 July 2026

First assessed. Protection under test. The text above is fixed as of this date and is not edited afterwards.

In the same band

Anthropic Stability AI Darktrace