Zhipu AI investor meeting: leading coding capability, strong demand, with a clear path to scaling and profitability

Overview

This report is Goldman Sachs’ meeting minutes after Knowledge Atlas Technology/Zhipu AI management and investors met in Hong Kong on April 13-14, 2026. The discussion focused on GLM coding model strengths, API price trends, demand and use-case coverage, model roadmap, and compute and infrastructure capabilities.

Core views

The core view is that the company has a strong differentiated advantage in coding and agent-engineering directions. Management believes coding is the primary mechanism for model interaction with the physical world through tools, so coding capability affects not only development efficiency but also future agent commercialization. The company also shows strong demand and pricing power, but whether it can continuously expand compute, improve utilization, and convert high demand into sustainable revenue and profit remains the key proof point.

Analysis framework

The report mainly uses an investor-meeting-summary approach, synthesizing management commentary on product capability, pricing, demand, roadmap, and infrastructure, combined with external benchmark results, API price changes, and revenue/usage growth shown in charts to form a qualitative assessment.

Methodology notes

• Capability assessmentBerkeley Function-Calling Leaderboard

  Function-calling capability evaluation

  The report notes that GLM ranks high on the BFCL, which evaluates a large language model's accuracy in calling functions, and is an important reference for assessing coding and agent capability.

• Meeting minutesInvestor meeting takeaways

  Management communication point synthesis

  The report is based on exchanges with management and investors and extracts five themes: coding capability, API pricing, demand intensity, model roadmap, and compute supply.

• Goldman Sachs research frameworkGS Factor Profile

  Comparisons of growth, financial returns, valuation multiples, and composite factors

  The disclosed appendix explains that GS Factor Profile compares stocks by growth, financial return, valuation multiples, and composite percentile ranking, but this report does not provide the company-specific factor scores.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Knowledge Atlas Technology/Zhipu AI (2513.HK)

  Core research target

  Strengths

  Strong GLM coding performance, high BFCL ranking, API pricing power, demand significantly above compute supply, and experience in AI infrastructure efficiency optimization.

  Weaknesses

  The company relies mainly on third-party compute and servers, so supply bottlenecks may limit revenue release; the report does not provide full financial forecasts, ratings, or a target price.

  Comparison

  GLM 5.1 input pricing is significantly higher than domestic peers but still clearly below U.S. vendors, indicating it has a domestic performance premium while an international price gap remains.

  Risks

  Insufficient compute supply, potential demand sensitivity to higher API pricing, slower-than-expected realization of the agent roadmap, intensified model competition, and dependence on third-party infrastructure.

• AI applications and agent ecosystem

  Benefit theme

  Strengths

  Improved coding and function-calling capability helps the model connect tools, software, and the physical world, advancing autonomous agents and high-value production use cases.

  Weaknesses

  Commercial deployment requires stable model capability, reliable execution of long-cycle tasks, and sufficient compute support.

  Comparison

  Compared with lightweight applications, the company places greater emphasis on high-value production scenarios and difficult coding tasks.

  Risks

  If autonomous agents and VLM multimodal capabilities fail to produce verifiable productivity gains at scale, commercialization pace may be slower than expected.

Key data

• Report date2026-04-15The cover date is 15 April 2026 | 12:49AM HKT.
• Stock code2513.HKThe report title identifies Knowledge Atlas Technology/Zhipu AI with status Not Covered.
• API price changeRaised 100% year-to-dateThe company said the API price increase reflects pricing power driven by coding performance.
• Demand versus supplyDemand is several times compute supplyManagement repeatedly emphasized that demand is significantly above current compute capacity.
• Charted revenue/scale metricAbout US$250mn by end-March 2026, approximately 60x year-on-yearThe chart shows a sharp rise from the second half of 2025 and reaching around 250 by around March 2026.
• Current price disclosureHK$948.00Company-specific disclosure includes Knowledge Atlas Technology (HK$948.00).
• GS global rating distributionBuy 50%，Hold 34%，Sell 16%The disclosed chart shows the rating distribution across Goldman Sachs global equity coverage.

Impact & implications

If GLM's coding edge proves durable, the company may gain a higher share in enterprise-grade complex coding, tool calling, and agent applications and improve revenue quality through API price hikes. At the same time, compute supply, reliance on third-party infrastructure, and execution efficiency in high-value scenarios will determine the pace of scaling and the extent of profit realization.

Risks

• Compute supply remains below demand, constraining client delivery and revenue expansion.
• After API price increases, if clients become more price sensitive, call-volume growth may be affected.
• Intensifying domestic and global foundation-model competition may compress both performance advantage and pricing power.
• The company mainly rents third-party compute and servers, which may create risks around cost, availability, and infrastructure stability.
• Autonomous agents, VLM multimodal, and long-cycle task execution still require validation, and the product roadmap has implementation uncertainty.
• The stock is not covered by the firm, and no rating, target price, or earnings forecast was provided; investment judgment requires additional financial analysis.

What to watch

• Changes in benchmark rankings for subsequent GLM versions on coding, function calling, and agent tasks.
• Client retention, call volume, and revenue elasticity after API price increases.
• Depth of usage by top internet companies and enterprise clients in difficult coding tasks.
• Scale of compute leasing, hardware utilization, and changes in unit inference cost.
• Whether autonomous agent and VLM multimodal capabilities transition into scalable production scenarios.
• Whether the company releases more complete guidance on revenue, margins, capex, or compute supply.