Sinovatio posts 129 million yuan loss as AI security stalls

Sinovatio, a state-controlled network security and data analytics firm listed in Shenzhen, reported a net loss of 129 million yuan (about $18 million) for the first half on revenue of roughly 235 million yuan, even as "AI security" hype lifted its shares.

The results expose how far the company's core business has slipped during a period of tighter government and enterprise budgets — and how little its AI products currently contribute. AI applications accounted for less than 2% of revenue, just a few million yuan, undercutting the concept-driven rally in its stock.

Sinovatio Technology was founded in February 2003 with registered capital of about 171 million yuan and is listed on the Shenzhen Stock Exchange main board, according to corporate records database Tianyancha. It began inside ZTE's early incubation system and was later taken over by state-owned platforms including Shenzhen Capital Group. Its core expertise has long been data extraction, network content security auditing and big-data analytics.

Revenue tied to government budgets

The main buyers of its broadband and mobile network data extraction systems are government cyberspace, police and state security agencies, along with major telecom carriers. Such contracts are large, highly customized and slow to be accepted and paid for, and their timing is tightly bound to budget cycles.

As local governments came under fiscal pressure and procurement of IT infrastructure was tightened, big-ticket, high-margin orders were delayed, scaled back or slow to clear acceptance.

At the same time, research staff pay, routine computing costs and depreciation are fixed and hard to cut. With revenue shrinking and costs holding steady, a loss of more than 100 million yuan became unavoidable.

AI security still early stage

The AI security and AI WorkForce platform themes that speculative money has chased are a case of distant water that cannot quench present thirst.

As large language models move into enterprise and government use, real security gaps have emerged: managing hallucinations, preventing data poisoning, guarding against prompt injection and reviewing content for compliance. Sinovatio's plan to carry over its deep packet inspection (DPI) and network content auditing technology into AI agent deployment and AI security governance is technically plausible.

But model security remains at an very early commercial stage. Government and enterprise clients are still cautiously testing private model deployments, and few are willing to set aside dedicated budgets for AI security software. With AI at under 2% of revenue, the AI WorkForce platform is mostly a bundled add-on in bids for core projects or a small pilot, unable to scale as a standard software product — let alone fill a loss of more than 100 million yuan.

The filing sends a blunt signal to the network security and enterprise services sector: markets can rally on imagination, but valuations ultimately return to signed contracts and cash collection. Sinovatio has kept to compliance by owning up to the numbers. The harder task for management is staunching the bleeding in its core business as the government procurement boom in network visualization fades and AI security has yet to prove it can sell.