Glacier Capital · 庚辛资本

Looking at deals

Investors are not our downstream; they are the upstream of our understanding. Every line below is something you taught us.

To Investors

• Repeat business is the only report card we watchWe recognise only one thing: whether you are willing to look at the next deal.
• We are at the table ourselvesGlacier Capital invests its own capital in companies it has served in depth. Glacier Capital insists on co-investing its own capital in the projects it serves deeply. If the judgement is right, we gain together. If it is wrong, we carry it together.
• Do not call every company number oneIn winner-take-all sectors, we argue why a company is first; where the market is not winner-take-all, we state clearly where it stands in the first tier. We give you the judgment; the choice is always yours.
• The full picture, not just the allocationOnly when you see the full picture will you dare to place a heavier bet. Round structure, secondary arrangements, where the others stand — we lay it all out.
• What reaches your desk is what we ourselves would dare to invest inWhether orders close freely on their own, whether the founder keeps promises, whether the existing shareholders genuinely help.
• Only the right dealsWe learn your profile by heart first: sector, stage, required certainty, exit preference. A deal that does not fit the profile is never recommended — not one.
• Materials complete in one pass — verifiable, citableLogic, interview notes, and data are complete in one pass — verifiable and citable. Your time should be spent on judgment, not on tidying.
• Meet the founder before the round opensThe window on a good deal is short, and allocation has always been contested. You should know the founder before the round opens — when no fundraising is under way, we are glad to arrange a first meeting, as friends.

We built ourselves a three-stage rocket: close rate, annual table-turn rate, institutional give-back rate. The first two stages can be worked on — pick your mandates and the first goes up, grind harder and the second goes up. Only the third cannot be manufactured.

We take twenty to thirty mandates a year, half new and half existing clients coming back for the next round — which means half of a year's capacity is bought by last year's delivery. The institutions we have closed with most often are almost all repeat clients.

We hold ourselves to one rule: every sentence we say should hold up when you check it later.

This chapter also appears on the home page: index.html#trust

Three screens we run for investors before a referral

Fail one and we do not refer. These three screens are not written into any contract. They are written into our own process.

In a year an investor will open a few dozen referrals seriously, about one a week of his attention. Every referral we make takes one slot out of that denominator.

Only what passes the three screens deserves a full picture.

Full piece: Three Screens Before a Referral

02 / SKIN IN THE GAME — would we put in our own money

Glacier Capital co-invests its own capital in the projects it serves deeply. That means going in on the same round, at the same price, as the investors. The reason is not return. It is calibration.

Companies Glaciers has invested in with its own capital

• D-Robotics
• ASTRONSTONE
• DISCOVER Robotics
• CAYE
• XY
• Spirit AI
• Paxini
• Volant Aerotech
• YIMU
• Evotrex
• OBSBOT
• Yarbo
• Rino.ai
• Robosen
• Yucca DPU
• Okawa
• ARCGine
• 碳硅无界
• 第一推力

Eight years serving and accompanying more than 80 portfolio companies; the public register lists 58, and our own capital is co-invested in 19 of them — in those 19 we are sitting at the same table.

The list covers investments made directly with our own capital and through affiliated entities, in no particular order; a few stay unlisted — confidentiality agreements, or their own tempo.

Many companies served, some served deeply, and our own capital committed only where we see most clearly.

We invest only in companies we have served deeply and see most clearly. If we cannot see clearly, we do not invest. If we see clearly and the price is wrong, we do not invest either.

This chapter also appears on the home page: index.html#skin

Where we stand

Roles must be distinguished deal by deal: in most projects we act as financial advisor; in some we also co-invest our own capital; in a small number we are solely an investor and not the advisor. When citing a specific case, follow the role stated in that transaction’s public reporting.

For eight years, we have given our best attention to these companies.

The list records only completed collaborations — none still under way.

TOP 1 on Qimingpian’s August FA Ranking; Qimingpian 2026 Mid-Year Rankings · China Private Equity Series: TOP 3 Best Financial Adviser (overall), TOP 3 Financial Adviser in Artificial Intelligence, TOP 3 Financial Adviser in Embodied Intelligence. The rankings are compiled and published by Qimingpian, a third party.

Qimingpian 2026 mid-year lists

Register · 58 companies

Eight Years of Service, in Full · ALL COMPANIES SERVED

• Evotrex
• 积加科技
• 创视半导体
• 零一汽车
• YIMU
• 第一推力
• 杉木科技
• 碳硅无界
• 内卡科技
• 大漠大智控
• 万勋科技
• 浩瀚卓越
• 铁锚科技
• 汤恩智能
• 优联电气
• 可以科技
• 觅光
• 易控智驾
• Okawa
• 高仙自动化
• ARCGine
• 中雷科技
• 元萝卜
• 予智科技
• 英达视
• 数字华夏
• 九识智能
• 希沃斯
• 北电数智
• 志凌伟业
• 时空壶
• 朗博特
• 识渊科技
• 七腾机器人
• PGT

The full panorama of companies Glacier Capital has served and accompanied over eight years, in no particular order; a few stay unlisted — confidentiality agreements, or their own tempo.

Note: for the role and the amount in any one transaction, follow that period’s public reporting.

• Paxini · Another record for the largest tactile-sensing financing worldwide: Paxini raises a further RMB 1 billion strategic round ↗ 36Kr, in Chinese
• Paxini · JD.com leads a strategic round in robot tactile-sensor maker Paxini, which has raised over RMB 1 billion in four months ↗ Sina Finance, in Chinese
• MatriQ · Three rounds in half a year: MatriQ completes its Series A ↗ PEdaily, in Chinese
• DISCOVER Robotics · DISCOVER Robotics completes a US$100 million angel+ round ↗ PEdaily, in Chinese
• Evotrex · Evotrex completes a Series A series of over RMB 200 million ↗ Tencent News, in Chinese

This chapter also appears on the home page: index.html#portfolio

We lay the facts in their proper order; the conclusion, the market draws for itself.

The four layers are one and the same thing, tightened four times. Each turn is narrower, heavier, and more concrete than the last.

The Four-Layer Vortex · how a deal is won

• SCREENING · find those who already understandEvery call and every email, every day, serves one thing: screening. The core algorithm is one word — precision. The criterion is not the size of the capital but the depth of understanding: the people whose eyes light up at the first keyword. Judge, too, which layer the other side stands on — the product, or the ecosystem — and write it down. This is the first valve on where time is invested.
• CONSTRUCTING · close the logic into a loopOnce the right people are screened out, there is no rush to sell the project. The standard for the BP and the memo is coherence, not completeness: organise an asset’s value into a verifiable closed loop of logic, in which every number and every chart withstands checking in reverse. Once the chain is complete, the conclusion is drawn by the other side itself.
• IGNITING · let the information arrive at the same timeWhen only one person says a project is good, the information has not yet spread. Meetings and roadshows are not isolated points; they synchronise the same set of facts to all relevant parties within the same time window. Who sees first, who later, in what order — that is tempo, and tempo is efficiency. Once information is synchronised, the market completes the pricing itself.
• CONTROLLING · lock the details down to the last cellOnce the information has spread, the work shifts from expression to keeping commitments. Price, allocation, terms, timetable — aligned item by item; every version of the term sheet and every round of communication traceable, so whatever is called upon at the negotiating table is accurate. The finer the late stages, the steadier the win.

A financing round, increasingly, runs like a competitive tender: capital bids in hard money, and every bid counts. The record must survive re-checking, each bidder must be found one by one, and the bids must land within the same window. The Four-Layer Vortex organises that tender: facts in order, and the price set by the market.

This chapter also appears on the home page: index.html#vortex

10 / FULL-MANDATE ADVISORY — THE CADENCE OF REVIEW

A brief review after every meeting; a full one that same day; a weekly synthesis; a complete monthly retrospective; and each quarter, a capital-markets race plan for the two quarters ahead. The first metric it watches is conversion: if the right people were screened, the TS conversion rate should pass one half. If it does not, the profile or the story is wrong — we go back and fix it.

When the window had yet to come, we sat with one company through ten months on the bench; when it came, multiple rounds were closed in quick succession. Never giving up is not a slogan — it is a schedule.

This chapter also appears on the home page: index.html#mandate

18 / SCALING UNDERSTANDING — helping institutions understand assets

A company misjudges one fewer window. What is saved is what we truly deliver.

The deal is only the result. What we do is Understanding, at Scale.

This chapter also appears on the home page: index.html#scaling-understanding

A select few sectors, and heavy conviction in phenomenal entrepreneurs.

AI (foundation models); Physical AI (embodied intelligence); Space · Quantum · Fusion · BCI (commercial spaceflight, quantum, nuclear fusion, brain-computer interfaces); INFRA (compute, power, optics). Going deep on the Musk tracks, one chain end to end.

These are not four isolated tracks. They share the upstream and downstream of one industrial chain, and the same cohort of phenomenal entrepreneurs.

Note: plus mature technology companies at growth financing, Pre-IPO, and M&A stages.

This chapter also appears on the home page: index.html#focus

12 / CONVICTION — the penetration of intelligence will follow the curve of new energy.

The AI of 2026 is the electric vehicle of 2015 — EV penetration exploded after a ramp of nearly five years, far beyond outside expectations. The dividend of the 14th Five-Year Plan came from electricity; the dividend of the 15th will come from intelligence. We are putting these ten years — the 15th and 16th Plans — on that one chain: staying with a hundred companies that are still small today, until they can stand on their own.

Shift the same curve eleven years: fifteen years on, AI penetration passes 60%; fifteen years after that, humans step out of 99% of production.

• 14th | penetration 1% → 27% → 60%E.
• 15th | AI penetration starting from roughly 1%.

The above is Glacier Capital’s own assessment, not a forecast or a promise, and does not constitute investment advice.

This chapter also appears on the home page: index.html#conviction

Certainty of Money · Volume IV · Directions & Sectors (11 pieces)

• Where We Are Going
• Watch Fewer Launch Events, Go Look at the Line
• The Watershed in Embodied Intelligence Is Day One
• The Real Question in Driverless Logistics Is Operations, Not Demos
• Low Altitude Is, in Essence, Order in a Layer of Airspace
• The Complexity of a Rocket
• For Quantum, Change the Ruler
• The Longest Line
• Below Compute Is Electricity
• Hardware Starts Charging by the Month
• Before the Hong Kong Window, Ask Whether You Qualify

You can meet even when there is no round

The meeting held when you are not raising is worth the most. Why? Because with no deal pressure, neither side has to perform. The founder does not have to round out a story. The investor does not have to hold a posture. The talk is about the business itself: where it is hard, where it is stuck, what comes next.

Meet only when the round opens and the investor has to judge a stranger in a few weeks. Meet in advance and he is verifying an observation: were the targets set back then met? The first runs on impressions, the second on evidence. Time is more honest than materials.

We hold the boundaries fairly tight. First, on a major project we share at once, before it starts, without waiting for the materials to be complete. Second, for a project we do not serve, a meeting carries one line of disclosure: we have invested, no referral, no group chat. Third, the gatekeeping we do for investors is not outsourced. Whether the orders close the loop, the founder’s integrity, how far the shareholders will support — we verify these ourselves.

Full piece: You Can Meet Even When There Is No Round

Contact

Direct line for investors: invest@grandecygne.com (Hangzhou · SEAL Team). Public contact entry points are also on the official site at grandecygne.com/#contact.