Z.ai Revenue Jumps Fivefold as API Business Surges

Z.ai revenue nearly quintupled in the first half of 2026 as the Chinese artificial-intelligence company shifted from project-led deployments toward a cloud platform and API business. The Hong Kong-listed company reported 953.9 million yuan ($142 million) of revenue, up 399.7% from a year earlier. Open-platform and API sales reached 825 million yuan, making recurring developer usage the central commercial engine.

Everyone else is reporting a 400% jump; we are explaining the business-model change behind it. Z.ai, known domestically as Zhipu AI, did not simply sell more of the same service. Cloud and API revenue expanded 2,735.7%, while local-deployment revenue fell 20.5% to about 129 million yuan. The company’s mix moved from customised installations toward metered access that developers can buy repeatedly.

What drove Z.ai revenue growth

The company’s open platform and application programming interface let developers send requests to Z.ai models and pay for usage. This model can scale faster than a separate on-premise deployment for each customer because the same cloud infrastructure serves many users.

API and open-platform revenue rose from roughly 29 million yuan to 825 million yuan. That business represented 86.5% of first-half revenue, according to the interim results. Caixin, South China Morning Post and Reuters independently reported the same commercial shift on August 31.

Local-deployment revenue moved in the opposite direction, falling 20.5% to about 129 million yuan. Local deployment installs models and supporting software inside a customer’s own infrastructure or private cloud. It can serve regulated or security-sensitive customers, but projects are slower to deliver and less standardised.

This mix change is more important than the headline growth rate. A platform business can generate repeat usage from a large developer base, while customised projects may require new sales and implementation work for each contract. Z.ai said its MaaS platform had more than 7.4 million enterprise and developer users.

Why losses tell two different stories

Z.ai’s total loss for the six months narrowed 12.1% to about 2.07 billion yuan from roughly 2.36 billion yuan a year earlier. That indicates faster revenue growth improved the gap between sales and overall spending on a reported basis.

However, adjusted net loss increased 12.1% to 1.96 billion yuan. Adjusted figures remove specified items to show management’s preferred view of recurring performance. When reported and adjusted measures move in different directions, readers should use both rather than saying simply that “losses narrowed.”

Gross profit rose 163.7% to 252 million yuan. Gross margin was about 26.4%, based on the reported gross profit and revenue figures. That means most revenue still flowed into the direct cost of delivering services before research, sales, administration and other expenses.

H1 2026 measure	Reported result	Year-on-year change
Revenue	953.9 million yuan	+399.7%
Open platform and API revenue	825 million yuan	+2,735.7%
Gross profit	252 million yuan	+163.7%
Total loss	2.07 billion yuan	Narrowed 12.1%
Adjusted net loss	1.96 billion yuan	Increased 12.1%
R&D expense	2.13 billion yuan	+33.6%

The cost of building the cloud engine

Research and development expense rose 33.6% to 2.13 billion yuan, more than twice first-half revenue. Advanced AI companies spend on model training, inference systems, data, engineers and computing clusters. The expense can remain high even when usage grows rapidly.

Each API call also carries an inference cost. Scale improves economics only if the revenue per call and infrastructure efficiency exceed the incremental computing cost. Price cuts can attract users but may weaken margin; price increases can improve unit economics but encourage developers to switch providers.

Z.ai has expanded the use of domestic chips as US export controls constrain access to some advanced processors. A domestic supply chain may reduce geopolitical dependence, but customers will judge the result by performance, reliability and price rather than origin alone.

What annual recurring revenue adds—and what it does not

Z.ai management said annual recurring revenue reached $1.6 billion by the end of August, up from $1 billion in early July. ARR annualises current recurring monthly income; it is not the same as audited revenue already earned during the year.

The gap between first-half reported revenue and August ARR is therefore not automatically a contradiction. ARR is a run-rate measure at a point in time, while financial statements record recognised revenue over a completed period under accounting rules. Rapidly changing API usage can make the annualised number much larger.

Investors should still ask how ARR is defined, which customers and products it includes, and how usage volatility affects it. A weekly spike cannot be assumed to persist for twelve months. Reported revenue, cash flow and margin remain necessary checks.

How Z.ai compares with China’s AI cloud race

Alibaba, Baidu, Tencent and independent model developers are all competing for enterprise and developer workloads. Large technology groups can subsidise AI investment with established businesses, while independent labs must prove that model access itself can become a durable commercial engine.

Z.ai’s first-half result is evidence that developers will pay for its platform at scale. It is not yet proof of sustainable profitability. The company needs high utilisation, reliable infrastructure and a margin that can fund continuing research without losses growing faster than revenue.

Lapaas Voice has tracked the company’s product and infrastructure strategy through the GLM-5.3 launch, its Chinese-chip data-centre plan and a real-world GLM-5.2 cost comparison. Those developments help explain why the company is spending heavily while pushing lower-cost access.

The primary source is Z.ai’s interim-results announcement filed with Hong Kong Exchanges and Clearing. Its issuer filings are available through the HKEXnews disclosure portal, while product information is published on the Z.ai platform.

What the result means for AI buyers

A rapidly growing API provider can give developers more competition on price and capability. Teams may gain access to strong coding, agent and language models without operating the full infrastructure themselves. Competition also pressures every provider to improve latency and reliability.

Buyers should not select a provider from revenue growth alone. They need tests on their own workloads, clear data-retention terms, service-level commitments and a fallback plan. For regulated uses, deployment location and auditability may matter as much as benchmark performance.

API economics should be measured across the complete application. A cheaper token price may be offset by more retries or longer outputs. A more expensive model may cost less overall if it completes tasks accurately in fewer calls.

Z.ai revenue growth shows that its commercial centre has moved decisively to cloud APIs: the segment supplied 86.5% of first-half sales after growing more than 27-fold. Yet R&D expense exceeded revenue and adjusted loss increased, so the business has demonstrated demand—not profitability.

What investors should watch next

The first metric is open-platform growth after the extraordinary comparison period. Maintaining 2,700% growth becomes mathematically harder as the base expands. Absolute revenue, user retention and usage per paying customer will become more informative.

The second is gross margin. Revenue that grows faster than gross profit can indicate pricing pressure or heavy delivery costs. Z.ai must improve infrastructure efficiency while funding model development and domestic-chip deployment.

The third is the relationship between reported loss, adjusted loss and cash use. A narrower accounting loss does not remove financing needs when adjusted loss and R&D spending remain high. Investors should reconcile every measure to the filing.

Finally, watch whether the ARR claim becomes recognised revenue over subsequent periods. If August usage persists, financial statements should eventually reflect it. If it was driven by promotions or temporary demand, the run rate may fall.

Why the comparison base matters

A 399.7% increase means current revenue is about five times the prior-year level, not that Z.ai added four times its current revenue. The earlier base was approximately 191 million yuan, so relatively small absolute gains in the comparison period create a very large percentage.

Future growth rates will probably slow even if the platform continues adding substantial revenue, because the comparison base is now much larger. Investors should focus on absolute additions, gross profit and cash requirements instead of expecting another fivefold increase every half-year.

The same discipline applies to API growth. A 27-fold expansion proves rapid adoption from a small base, while the next phase must demonstrate durable customer usage. Retention, workload diversity and revenue concentration will reveal whether the cloud business is resilient or dependent on a few unusually large customers.

FAQs

How much did Z.ai revenue grow in the first half of 2026?

Revenue rose 399.7% year on year to 953.9 million yuan, or about $142 million.

What drove Z.ai revenue growth?

Open-platform and API revenue grew 2,735.7% to 825 million yuan and represented 86.5% of total revenue.

Is Z.ai profitable?

No. Total loss narrowed, but the company still reported a loss of about 2.07 billion yuan and an adjusted net loss of 1.96 billion yuan.

Is the $1.6 billion ARR the same as annual revenue?

No. ARR annualises current recurring monthly income, while reported revenue is recognised under accounting rules for a completed period.