Chinese custom silicon and semiconductor IP provider VeriSilicon reported first-half 2026 revenue of 1.864 billion yuan, up 91.37% from the same period a year earlier. The result underscores how quickly demand for artificial intelligence compute is reshaping the order books of design service companies that sit between fabless chip developers and foundries.
According to TechNode’s August 18 report on VeriSilicon’s first-half results, which cited 21 Finance, AI computing-related work accounted for roughly 90% of new orders that the company signed this year. VeriSilicon also disclosed that orders on hand stood at 12.449 billion yuan as of June 30, 2026, a figure that dwarfs the revenue booked in the six months just reported and points to a substantial pipeline of work awaiting execution. DIGITIMES separately corroborated the revenue growth and described AI computing demand as supporting the company’s custom silicon and mass production services.
The scale of that backlog, combined with the AI concentration in new bookings, suggests that VeriSilicon has become one of the more visible beneficiaries of the current wave of AI infrastructure spending inside China. The result places VeriSilicon within a wider domestic effort to build semiconductor design and software capabilities for AI. EastFrontier’s earlier coverage of China’s domestic AI software stack provides context for the environment in which such design-service orders are growing.
Backlog dwarfs first-half revenue and hints at multi-year visibility
The gap between the 12.449 billion yuan order book and the 1.864 billion yuan of revenue recognized during the first six months is the most striking element of the disclosure. In broad terms, the reported backlog is roughly 6.7 times the revenue booked in the period, a ratio that is unusual for a company primarily engaged in chip design services and IP licensing, where revenue recognition tends to track project milestones and mass production ramps.
A backlog of that size can be read in more than one way. It can reflect the fact that custom silicon projects, particularly those targeting AI accelerators, often involve multi-year engagements that begin with design, move through tape-out, and only later translate into higher-value mass production revenue. It can also indicate that customers are placing longer-dated commitments to secure capacity and design bandwidth, given how competitive the market for advanced AI compute has become. The sources reviewed do not provide a conversion schedule for the backlog, so any assumption about how quickly these orders will flow into reported revenue would go beyond the available information.
Analysts and investors watching the sector will likely focus on whether the mix of design services, IP licensing, and mass production within the backlog shifts materially in the coming quarters. Mass production revenue tends to arrive later in a project’s lifecycle but can be substantially larger in absolute terms than the design phase, which means the composition of the 12.449 billion yuan figure is as important as its headline size.
AI computing dominates new bookings for 2026
The disclosure that AI computing-related orders accounted for around 90% of new orders signed in 2026, as reported by TechNode citing 21 Finance, marks a significant concentration. It signals that VeriSilicon’s incremental growth is now overwhelmingly tied to the AI cycle rather than to the broader base of consumer, automotive, or industrial silicon that has historically featured in the company’s disclosed end markets.
That concentration has both upside and risk. On the upside, VeriSilicon is positioned to benefit from Chinese cloud providers, internet platforms, and specialized AI chip startups that are all searching for design partners capable of delivering custom silicon at advanced nodes. Domestic demand for AI compute has been reinforced by a broader push to build out software and hardware stacks that reduce reliance on foreign technology, a trend explored in EastFrontier’s coverage of how China is breaking free from CUDA as the domestic AI software stack gains ground. VeriSilicon’s mass production services and IP portfolio slot into that ecosystem at the silicon layer.
The risks are the mirror image. A book heavily concentrated in AI compute is exposed to any deceleration in AI capex, to shifts in customer product roadmaps, and to policy or export control developments that alter access to advanced process nodes or key EDA tools. VeriSilicon’s disclosures do not quantify these sensitivities, and no customer names or specific end applications tied to the AI bookings have been made public in the sources reviewed.
What the first-half print says and what it does not
The 91.37% year over year revenue growth in the first half is a clear indicator that VeriSilicon’s operating momentum has strengthened, while the order mix reported by TechNode citing 21 Finance places AI compute at the center of new bookings. DIGITIMES’s description of AI demand supporting custom silicon and mass production services is consistent with the picture drawn by the order book and the new order mix reported by TechNode.
Several important items remain outside the public record. The company has not disclosed a precise share of revenue attributable to AI computing customers during the first half, nor has it detailed customer concentration, gross margin trends, or a timeline for converting the 12.449 billion yuan of orders on hand into recognized revenue. Any commentary on these dimensions would go beyond what VeriSilicon and the cited outlets have confirmed.
For now, the readable signals are straightforward. Revenue nearly doubled year over year in the first half, the backlog stands at multiples of that revenue, and the overwhelming majority of new business signed this year is tied to AI compute. Those three facts together frame VeriSilicon as one of the clearer proxies for the pace of AI silicon design activity in China, even as questions about the timing, margin profile, and durability of that demand remain to be answered in later disclosures.