NeuroAI NEUROAINEUROAI.SITE
ESC

Two small acquisitions show where China's AI supply chain is actually being bought

A laser-equipment maker bought an optical-module coupling supplier for RMB 400 million; a juice maker bought a laminate supplier for RMB 793 million, with earnouts.

2026-11-03 · 1183 words · NeuroAI
Two small acquisitions show where China's AI supply chain is actually being bought

A laser company that cuts battery casings bought a business that couples light into optical modules. A company that sells fruit juice bought a business that laminates copper sheets for servers.

Neither acquirer had anything to do with AI before either deal. Both paid well over four times the target's net assets, wrote performance promises into the contract, and held money in escrow until those promises were met.

Key takeaways

  • Wuhan Yifei Laser (逸飞激光) agreed to pay RMB 400 million (≈US$60M / HK$467M) in cash for 100% of Shenzhen Xingqihang Automation (深圳市兴启航自动化设备), which makes coupling and packaging equipment for optical modules. The board approved it on 27 October; RMB 224 million (≈US$33M / HK$262M) comes from over-raised IPO proceeds (超募资金).
  • The valuation was set by income approach at RMB 403 million, an appraisal premium of 532.56% over the target's RMB 63.7 million net assets. A market approach came in at RMB 577 million, an 805.68% premium, and the parties chose the lower figure.
  • Yantai North Andre Juice (安德利) agreed to pay RMB 792,787,965.88 (≈US$118M / HK$926M) for 62.0611% of Ningbo Yongqiang Technology (宁波甬强科技), a high-speed copper-clad laminate (覆铜板) supplier. Completion was confirmed on 28 October, and Yongqiang becomes a consolidated subsidiary with three of five board seats.
  • Yongqiang's income-approach valuation was RMB 1.275 billion, a 354.38% premium, with about RMB 520 million (≈US$77M) of goodwill expected. It lost money in 2024 and 2025, then turned a RMB 12.36 million (≈US$1.8M) net profit in the first half of 2026 on revenue of RMB 121.69 million (≈US$18.1M), up 58.33%.
  • The earnouts are the real price. Yifei's sellers promise RMB 93 million (≈US$14M / HK$109M) of cumulative adjusted net profit over 2026–2028, with cash compensation below 85% of that. Yongqiang's sellers promise RMB 135 million (≈US$20M / HK$158M) across three years, in steps of RMB 23 million, RMB 42 million and RMB 70 million.

What the two deals have in common

Both are 能力平移型 acquisitions — buying a process and a customer list rather than a product line, the structure Chinese industrial buyers are converging on as the AI supply chain fragments.

Xingqihang was founded in 2014 and makes coupling equipment across TO-CAN, COB, BOX lens and fibre-array formats, covering laser welding, UV glue curing and thermal curing, and pushing towards co-packaged optics (CPO) equipment and coupling gear for 1.6T-plus links. It serves multiple leading optical module customers, and its founders come from precision equipment.

Yongqiang makes high-frequency high-speed copper-clad laminate, prepreg and IC substrate materials in Ningbo's Beilun cluster. Its M8 and M9 grade products have passed certification at some end customers and are shipping; M10 is in customer testing. In the first eight months of 2026 its high-speed product revenue grew 174.19% year on year — from a base so small that high-speed lines were still only 10.52% of revenue.

The buyers are the surprise. Yifei Laser supplies precision laser processing equipment, mainly for new-energy batteries. Andre is a leading global fruit juice concentrate producer. Both have cash flow, and neither had an AI business to fold the target into.

The earnout mechanics

When an acquirer pays more than four times net assets for a business, the premium is a claim about future earnings, and the only mechanism that makes it enforceable is the earnout.

Yifei's is structured around a three-year adjusted net profit floor. Payment is staged 60%, 20%, 20%, with later tranches tied to audit results and performance. The sellers separately agreed to buy at least 20% of their after-tax proceeds back into Yifei Laser stock before October 2029, locked three years — unusual, since the people paid in cash must become shareholders.

Andre's arrangement is more elaborate. Yongqiang's sellers, JIANGQIHE and QIANGYUAN, retain 13.2592% and 12.9011% after the deal, 26.1603% between them, worth about RMB 333.5 million (≈US$50M) at the transaction valuation. Their compensation obligation is capped at roughly RMB 100.7 million, and an escrow account is funded with about RMB 80.7 million (≈US$12M) after tax. A stress test the company ran puts the maximum the sellers might have to cover from their own pocket at RMB 37.39 million (≈US$5.6M), which their retained stake covers. Their remaining equity is locked until the end of 2030 and until Yongqiang's annual net profit first exceeds RMB 50 million.

Andre also discloses that over the past three years, capital injections and share transfers in the target implied a post-money valuation of about RMB 1.49 billion (≈US$222M), 16.86% above what it is paying now.

What the market is actually buying

Both targets are early. Xingqihang reported revenue of RMB 54.18 million (≈US$8.1M) in 2025 and net profit of RMB 11.16 million (≈US$1.7M), then revenue of RMB 37.41 million (≈US$5.6M) and net profit of RMB 10.34 million (≈US$1.5M) in the first five months of 2026 — a run rate that, if sustained, would put the RMB 400 million price near sixteen times annualised profit.

Yongqiang's order book at the end of August 2026 stood at RMB 47.23 million (≈US$7M), against a 2026 profit commitment of RMB 23 million. Its top five customers accounted for RMB 91.09 million of first-half revenue, and the company calls these relationships stable, with new certifications still coming: two more optical module companies for M4-to-M8 grade material, one M9 certification in small-batch supply, an M6 pass at a domestic internet company and an M8 pass at an international social technology company.

According to Li An, chief scientist at BrainNet (脑机网), the earnout is the only clause in either deal that tells you what the acquirer actually believes about the target's next three years.

Honest limitations

The valuations, appraisal premiums, earnout amounts, escrow balances and stress-test figures all come from the acquirers' own announcements or supplementary filings, and the appraisers' models are not published; where two methods gave very different answers, the company chose the lower one. The RMB 400 million and RMB 792.79 million prices are agreed transaction amounts, not post-money valuations, so multiples depend on how much of the earnout pays out. The customers behind Yongqiang's certifications are described by category, not by name. Yifei's staged payment percentages and the sellers' obligation to buy acquirer stock are as described in the announcement, but how that purchase would be executed has not been set out publicly. This article describes transactions and their contractual structures; it is not a recommendation to buy or sell the securities of any party involved.

What readers can do now

  • Read the earnout before the headline price. Two deals, three protection mechanisms, and in both the earnout does more work than the headline multiple.
  • Check what the acquirer knows that you do not. Both buyers had operating relationships or adjacent process knowledge before the transaction, which is the real source of value, not the target's standalone technology.
  • Track milestone dates against the earnout steps. The 2026 and 2027 profit numbers decide whether the premium was justified or paid for with escrow.
  • Note that the acquirers here are industrial companies with cash flow, not technology companies with stock. That distinction matters for who ends up holding the AI supply chain.

Related coverage

Cite this article

NeuroAI (脑机网), "Two small acquisitions show where China's AI supply chain is actually being bought," 3 November 2026, https://neuroai.site/a/na-capital-ma-optical-coupling-laminate

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Capital & Markets” → · Back to home · Markdown version