Zhipu AI: API demand, token prices, and model iteration all strengthen in tandem

Overview

This report updates the earnings and valuation for Knowledge Atlas Technology Joint Stock Co. Ltd. (Zhipu AI). J.P. Morgan believes the company is showing strong upward momentum in model iteration, API demand, token consumption, and API pricing, with especially rapid growth in open platform API ARR, which is simultaneously improving revenue visibility and the capacity to fund R&D.

Core views

The core view is that Zhipu AI has entered an important inflection point: on one hand, the iteration from GLM-4.5/4.6/4.7 to GLM5, together with a shift toward agentic systems, tool-enhanced reasoning, and developer infrastructure, has increased its commercial value in coding, long-context reasoning, and multi-step execution reliability; on the other hand, API demand and pricing are rising at the same time, suggesting that growth is not merely driven by low-price volume expansion, but by model capability, task completion quality, throughput, and reliability. The report therefore raises 2026-30E revenue forecasts by 46%-78% and maintains its view of long-term profitability improvement.

Analysis framework

The report combines fundamental forecast revisions with valuation re-rating: it first analyzes API ARR, token pricing, gross margin, and R&D spending structure, then updates 2026-2030E revenue, profit, and cash flow forecasts; on valuation, it uses 2030E normalized earnings, a 30x P/E, and a 15% WACC to discount to a Dec-26 target price.

Methodology notes

• Valuation methodsP/E Discounted Target Price Method

  2030E normalized earnings valuation

  The HK$950 target price is derived from 2030E adjusted EPS of Rmb43, a 30x target P/E, and discounting at 15% WACC to Dec-26.

• Fundamental AnalysisRevenue Forecast Revision

  Rising API demand and prices

  The report takes API ARR, strong token consumption intensity, and higher token prices as the core evidence for revenue upgrades, viewing them as a reflection of real model competitiveness and higher-value workloads.

• Profit QualityContribution-Level Breakeven Analysis

  Gross profit covers non-R&D operating expenses

  The 2025 Rmb3.2bn R&D investment is roughly equal to adjusted net loss, leading the report to conclude that gross profit from current models already covers SG&A on a cash basis, and that losses mainly stem from deliberate R&D investment.

Asset mapping & comparison

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

• Knowledge Atlas Technology Joint Stock Co. Ltd. (Zhipu AI) equity

  Directly covered name

  Strengths

  Fast-growing API ARR, rising token prices, rapid model iteration, local deployment base in regulated sectors, and scalable growth potential in cloud API.

  Weaknesses

  Still loss-making, high R&D spending, and near-term profitability remains under pressure.

  Comparison

  The report believes its valuation multiple should trade at a premium to China's leading internet companies, mainly because 2026-30E revenue CAGR is expected to exceed 100%.

  Risks

  Export controls, geopolitics, entity-list risk, intensifying competition, compute cost and supply availability, and uncertainty around commercialization and customer adoption.

• AI foundation model industry chain

  Thematic beneficiary direction

  Strengths

  Developer infrastructure, coding workflows, agentic systems, and long-context reasoning may expand enterprise and developer budgets.

  Weaknesses

  The industry is highly competitive, and maintaining a leading model position requires sustained heavy investment.

  Comparison

  The report emphasizes that long-term economic outcomes are primarily determined by whether the company can remain in the global top tier of model capability, rather than by short-term deployment form.

  Risks

  Failed technical iteration, price competition, external compute supply constraints, and regulatory restrictions.

Key data

• API platform ARRUS$250mnAs of March 31, 2026, up 6.4x YTD and 60x over the past 12 months; management's year-end target is US$1bn.
• Token price change+83% YTDPrices rose while demand continued to accelerate, indicating relatively strong commercialization quality.
• 2026-30E revenue forecast revision+46% to +78%J.P. Morgan raised Zhipu AI's 2026-2030E revenue forecasts.
• 2025 R&D spendingRmb3.2bnThe amount was roughly equal to adjusted net loss and is viewed as proactive investment in the next generation of model iteration.
• Open platform API gross margin3% in 2024 to 19% in 2025Reflects improvements in scale, utilization, and model efficiency.
• 2030E revenueRmb98.832bnThe 2030E revenue forecast in the valuation table.
• 2030E adjusted net profitRmb20.360bnRaised by 19% versus the previous forecast.
• Target priceHK$950Raised from HK$800, with the rating maintained at Overweight.

Impact & implications

If API demand, token prices, and gross margins continue to improve, Zhipu AI's revenue growth and narrowing losses will become more mechanical: high-growth revenue expands the gross profit pool, margin expansion improves unit economics, and R&D spending can continue to support model leadership. For investors, the key implication is that the company's valuation will depend more on whether it can continue to remain in the global first-tier model capability group and convert technological capability into developer and enterprise budgets.

Risks

• Export controls, geopolitics, and entity-list designation risk.
• Intensifying competition in the AI foundation model industry may compress pricing and market share.
• Heavy and ongoing R&D spending creates execution risk and pressures profitability.
• Commercialization progress and customer adoption remain uncertain.
• Dependence on compute infrastructure and external suppliers creates cost and availability risk.

What to watch

• Whether API platform ARR can continue to progress from US$250mn toward management's US$1bn year-end target.
• Whether the increase in token prices is sustainable and whether demand continues to accelerate in tandem.
• Developer adoption of the GLM series in high-value use cases such as coding, agents, and long-context reasoning.
• Whether open platform API gross margin can continue to improve from 19% in 2025.
• Whether the 2029 profitability inflection point and the 2030E adjusted net profit forecast of Rmb20.36bn can be achieved.