(Yicai) Sept. 1 -- Z.AI's shares rose after the Chinese startup shrank its net loss and grew revenue from its open platform and application programming interface services more than 27 times in the first half of this year.

Z.AI, formerly known as Zhipu AI, [HKG: 2513] jumped 2.9 percent to HKD1230 (USD156.91) per share as of lunch break in Hong Kong today. The shares, which debuted on the bourse in January, have increased more than 10 times in value from their offering price of HKD116.20 (USD14.82).

Net loss narrowed 12 percent to CNY2.1 billion (USD307.7 million) in the six months ended June 30 from a year earlier, according to the Beijing-based company's earnings report released yesterday. Its revenue jumped fivefold to CNY954 million (USD142 million), surpassing last year's total of CNY724.3 million, with income from its open platform and API business reaching CNY825 million.

Z.AI's research and development spending jumped 34 percent to CNY2.1 billion, mainly on model R&D and computing infrastructure, the company noted.

The shift in revenue composition is more noteworthy than the growth, Xiao Lei, board secretary of Z.AI, said at an earnings conference call. The share of income from localized deployment fell to 14 percent from 85 percent year on year, while the API services' gross profit margin jumped to 25 percent from minus 0.4 percent, Xiao pointed out.

As of yesterday, the number of enterprise and developer users of Z.AI's Mobility-as-a-Service platform exceeded 7.4 million, the token call volume soared more than 40 times from the end of last year, the number of paid daily active users surged 603 percent year on year, and the daily call volume of its top 10 customers soared 98 times. In addition, the average selling price of APIs more than doubled.

As the model architecture and inference infrastructure are being optimized, Z.AI's unit token inference cost has dropped 80 percent from the beginning of the year, the revenue per unit of computing power investment (including training and inference) has surged 14 times, and the gross profit margin of the MaaS business has reached almost 25 percent, the company said.

Z.AI has achieved large-scale low-cost inference using a cluster of 100,000 China-made chips, the company previously told Yicai.

Z.AI's computing power sources are diversified, covering its own clusters, computing power leasing, and procurement services, used in pre-, mid-, and post-training, as well as production reasoning, Xiao stressed.

The main contradiction in the computing power industry has shifted from whether Chinese chips can work to how to verify their economic viability, Xiao noted, adding that Z.AI is more focused on "effective computing power," meaning how much computing power has been installed, can be stably scheduled, run for a long time, and ultimately be converted into effective tokens.

Artificial intelligence models are evolving rapidly, and amid fierce industry competition, any advantage on a single leaderboard is likely to be erased soon, noted Liu Debing, chairman of Z.AI. What really matters is who can sustainably deliver higher intelligence at lower cost, he stressed.

The industry will have two price curves, with the price of equivalent intelligence continuing to decline, but pricing space for models that open new task boundaries and significantly improve success rates, Liu said, noting that tokens do not have pricing power, but what matters is rather what tasks they can ultimately accomplish.

Editor: Martin Kadiev