Profet AI uses Domain Twin to drive the implementation of AI agents in manufacturing

Overview

This report comes from Goldman Sachs' China Private Tech Tour and records the key discussion points from an exchange in Hong Kong with Profet AI's co-founder and CEO. Headquartered in Taipei and founded in 2018, Profet AI provides AI agents and a digital workforce for manufacturing. Its Domain Twin includes three key AI modules: AutoML, AI Lifecycle Management, and AI Studio. The report focuses on how AI agents can preserve manufacturing know-how, accelerate global factory replication, and improve yields in new product manufacturing.

Core views

Goldman Sachs' core view is that Profet AI's business progress validates the trend of AI moving from foundation model training toward model monetization, end-user productivity improvement, and manufacturing efficiency gains. Geopolitical pressure is driving manufacturers to diversify production capacity geographically, while AI-related product migration is accelerating, boosting demand for Domain Twin. Profet AI entered the digital workforce field relatively early, has leading manufacturing clients, a longer usage track record, and participation in standard-setting, which may enhance customer stickiness and market penetration.

Analysis framework

The report is mainly based on management discussions during the Private Technology Tour, company introductions, and product interface demonstrations. It qualitatively analyzes Profet AI's product modules, manufacturing scenarios, competitive advantages, customer expansion, and implications for the AI infrastructure investment chain, while mapping these to asset ideas based on Goldman Sachs-covered stocks.

Methodology notes

• Company researchPrivate Tech Tour management discussion

  Validate product demand, customer expansion, and industry trends through discussions with company management.

  Based on Profet AI management's descriptions of Domain Twin, AI agents, customer deployments, and industry standards, the report assesses demand for manufacturing AI software and its support for AI infrastructure investment.

• Product and industry analysisDomain Twin / AI agent use-case analysis

  Assess product value through scenarios such as production-line monitoring, yield forecasting, parameter recommendations, report generation, and multilingual factory replication.

  Domain Twin is described as a digital employee equipped with senior engineers' know-how, which can reduce hiring and training pressure for new factories and lessen the impact of senior engineer turnover on the transfer of manufacturing experience.

• Goldman Sachs internal frameworkGS Factor Profile

  Provide investment context for stocks across four dimensions: Growth, Financial Returns, Multiple, and Integrated.

  The disclosure section states that this framework is used to compare key stock attributes with the market and industry peers, but the main text does not provide specific factor scores for Profet AI.

Asset mapping & comparison

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

• Profet AI

  Core research subject

  Strengths

  Entered the digital workforce field relatively early, serves leading manufacturers, participates in potential industry standard-setting, and offers products covering AutoML, AI model lifecycle management, and AI Studio.

  Weaknesses

  The report does not provide financial data, valuation, profitability, or listed security information.

  Comparison

  Compared with traditional enterprise software, Profet AI is positioned as an AI agent generator and platform, enabling AI agents to work with systems such as ERP, CRM, PLM, and MES.

  Risks

  AI agent security requirements, the ability to keep data and know-how within the enterprise, maintaining model accuracy, and the pace of large-scale customer deployment.

• Hon Hai

  AI server buy idea

  Strengths

  Listed by Goldman Sachs as a buy idea related to AI servers.

  Weaknesses

  This report does not elaborate on company-specific fundamental details.

  Comparison

  Along with Wistron, Wiwynn, and Lenovo, it belongs to the beneficiary theme related to AI servers.

  Risks

  Volatility in AI infrastructure capex, changes in customer orders, and manufacturing cycle risks.

• Wistron

  AI server buy idea

  Strengths

  The disclosure page lists a Buy rating and a target price of NT$144.50.

  Weaknesses

  The main text does not provide more granular financial or order data.

  Comparison

  Together with Hon Hai, Wiwynn, and Lenovo, it maps to the AI server ODM and manufacturing chain.

  Risks

  AI server demand cadence, supply chain execution, and valuation volatility.

• Wiwynn

  AI server buy idea

  Strengths

  The disclosure page lists a Buy rating and a target price of NT$4,220.00.

  Weaknesses

  The report does not provide the assumptions behind this target price.

  Comparison

  Benefits alongside other AI server-related names from the continuation of AI infrastructure investment.

  Risks

  Customer concentration, volatility in AI server orders, and high-valuation risk.

• Ruijie Networks

  Switch buy idea

  Strengths

  Listed as a buy name in the switch segment, with the disclosure page showing a Buy rating and a target price of Rmb92.99.

  Weaknesses

  The main text does not elaborate on switch business data.

  Comparison

  Compared with AI server ODMs, Ruijie Networks is more exposed to the beneficiary direction of networking equipment and switches.

  Risks

  The pace of data center network buildout, competition, and pricing pressure.

Key data

• Meeting date2026-04-20Profet AI's co-founder and CEO met in Hong Kong during Goldman Sachs' Private Technology Tour.
• During the Tour2026-04-20至2026-04-28The report is labeled as China Private Tech Tour.
• Company founding date2018Profet AI was founded in 2018 and is headquartered in Taipei.
• Number of Domain Twins per single client2到10个Management said a single client can develop multiple Domain Twins, each responsible for different manufacturing processes.
• Core AI modulesAutoML; AI Lifecycle Management; AI StudioAutoML is a no-code platform, and AI Lifecycle Management is used to govern and manage thousands of AI models across production lines.
• Served industriesSemiconductors; PCB; Panel; Automotive component; EMSThe company provides AI agents and digital workforce solutions to various types of manufacturers.
• Buy ideasAI servers: Hon Hai, Wistron, Wiwynn, Lenovo; Switch: RuijieThe report links demand from Profet AI to the sustainability of AI infrastructure investment.
• Disclosed ratings and target pricesHon Hai Buy NT$211.00; Ruijie Networks Buy Rmb92.99; Wistron Buy NT$144.50; Wiwynn Buy NT$4,220.00The disclosure page lists the ratings and target prices of the relevant covered names.

Impact & implications

The investment implication of this report is that manufacturing AI agents are not just a software application theme, but may also extend the AI infrastructure investment cycle. If Domain Twin can help manufacturers rapidly replicate overseas capacity, shorten new product introduction cycles, and improve yields, it would increase demand for AI servers, optical networking, switches, and related manufacturing supply chains. At the same time, enterprise software may shift from being used by human employees to also being invoked by AI agents, potentially reshaping integration standards for ERP, CRM, PLM, MES, and other systems.

Risks

• The report does not disclose Profet AI's financial data, valuation, order scale, or profitability, making it difficult to quantify investment value.
• Manufacturing AI agent deployment involves data security, know-how protection, and cross-system permission management; if it cannot meet the security requirements of leading clients, penetration may be limited.
• Domain Twin's actual ROI, model stability, and yield improvement effects require validation through long-term production data.
• If AI infrastructure investment slows, related buy ideas in AI servers and switches may come under pressure.
• Cross-regional factory construction and supply chain relocation are affected by geopolitics, regulation, customer certification, and execution capability.

What to watch

• Progress in Profet AI's adaptation to the security standards of globally leading foundries, and whether these requirements become industry standards.
• The rollout progress of enterprise AI agents that Profet AI and a globally leading GPU supplier plan to jointly launch in mid-2026.
• The speed at which a single client expands from 2 to 10 Domain Twins to more manufacturing processes.
• The quantifiable impact of Domain Twin on new-country factory ramp-up, new-spec product introduction, and yield improvement.
• The progress of connecting enterprise software such as ERP, CRM, PLM, and MES to AI agent platforms.
• Whether investment demand for AI servers, switches, and optical networking continues to expand along with AI application monetization.