AMPulse

US / China funding architecture

Two Capital Stacks

Disclosed primary funding in the tracked US and China AM ecosystems, 2023-2026. Similar Tier B totals, different source mixes, and material sensitivity to cohort scope and disclosure.

Disclosed capital by source · 2023–2026

US

$3.9B

CN

$3.9B

Strategic / corporate

Public markets

Tier-1 VC

US federal / defense

CN state-linked

Other / undisclosed

Funded firms

US165CN146

Disclosed capital

US$3.9BCN$3.9B

Median round

US$3.7MCN$10.3M

AMPULSE DATA REPORT · 2026Data snapshot: Sep 8, 2026

AMPulse Data Reports

ampulse.online · Additive manufacturing intelligence

Two stacks, different bets

In the tracked Tier B AM-ecosystem cohort, disclosed primary funding from 2023 through 2026-09-08 totals $3.86B in the United States and $3.90B in China. These are similar recorded totals, not an estimate of equal national capital formation. Only 72.3% of US rounds and 34.8% of Chinese rounds have a parsable amount. Scope matters too: Tier A totals are US $148.3B and CN $8.0B; Tier C totals are US $2.5B and CN $1.7B.

The source mix is more useful than a claim of opposite channels. Public markets are the largest identified dollar source in both countries, but carry 68% of Chinese dollars versus 18% in the US. The US Other / Undisclosed bucket is larger than any identified source. Federal awards are numerous among labeled US sources; they do not account for most disclosed US capital. Investor-name classification and missing amounts limit how far this comparison can be generalized.

The contrasts to watch are these: where the money comes from, how heavy each round is and at what stage, what fraction of each cohort sits in physical hardware and materials versus services and software, and which process families attract concentration. Most of these signals hold under three nested cohort definitions (inclusive, AM-ecosystem, strict AM); the sensitivity panel names the findings that change with scope. The cohort and disclosure conditions travel with every comparison.

Finding 1: Where the money comes from

US AM funding is a federal-grant count sitting under an unclassified plurality. 74 of the 238 US rounds are federal grants and contracts, more than any other identified source bucket, and they sum to $181.2M. The largest US dollar bucket is Other / Undisclosed: 113 rounds (47% of the count) and $1.95B (50% of the dollars) with no investor the classifier can name. Tier-1 venture firms account for 20 rounds and $447.9M, 12% of US dollars. Strategic and corporate rounds add $596.7M.

China's stack is public markets first. IPOs, post-IPO offerings and private placements (treated as quasi-public) put $2.64B into Chinese AM in the window, 68% of the CN total, across 18 transactions: STAR Market and ChiNext listings of equipment and materials makers (Farsoon on STAR in 2023, Aidite on ChiNext in 2024, Zrapid in 2024) and placements by already-listed companies (BLT's 2023-12 placement, CNMNC/OTIC's 2024-03 placement). In the United States the same bucket totals $0.7B across 8 transactions, and the two large US listings in the window are not in it: UL Solutions's 2024 IPO and 2025 offering were share sales by its parent and sit in the ownership-transfer table.

SBIR Phase I awards are sized in the low hundreds of thousands of dollars and Phase II in the low single-digit millions, so a high count and a low aggregate is the SBIR shape itself. China runs analogous early-stage public R&D support through provincial industry funds, national guidance funds and university spin-off vehicles, but those flow on different timelines and report through different channels than the single-round disclosures this report aggregates; the 14 rounds and $221.2M labeled State-linked here are the part the investor names give away, not the whole.

The first edition of this report put US Strategic / Corporate at $6.8B on the strength of one transaction, Apollo's $5.2B buyout of Arconic. That was a change of control, not a raise; typed correctly, US Strategic / Corporate is $596.7M over 23 rounds against CN's $45.9M over 23. The US corporate channel is still an order of magnitude larger in dollars; it is no longer the largest US channel.

Finding 2: Round shape and stage

CN median rounds are heavier than US median rounds at every tier: $10.3M against $3.7M in the headline cohort, 2.8x. Part of that ratio is the SBIR count: 69 of the 238 US rounds are government grants against 3 of 244 in China. Take the grants out and the gap narrows to $11.0M against $6.5M, 1.7x, and at Tier A (which admits the large AM-user rounds on the US side) it closes. The ex-grant median is the number to carry: Chinese AM rounds are heavier, by less than half again rather than by three times.

The US distribution declines from the $1-5M bucket upward (51 rounds at $1-5M, 42 at $5-25M, 24 at $25-100M, 10 at $100M+), with the sub-$1M SBIR band (45) close behind. The CN distribution peaks at $5-25M (32 rounds) and then splits: 10 rounds at $25-100M and 15 at $100M+, so the thinnest CN bucket is the one just below the megarounds. The CN shape is a venture middle plus a public-market top.

The stage chart is where the two cohorts look least alike. China is Series-A-modal: 75 Series A rounds, 37 Series B, 24 seed or angel, 17 IPO or post-IPO. The United States is grant-modal: 69 government grants, 50 seed or angel, and only 19 Series A rounds. The Chinese cohort is an early-venture population with a public-market exit ramp, not a population of late-stage industrial operators; the heavier CN median comes from the size of an ordinary Chinese Series A, not from the absence of small rounds.

Finding 3: Where the money goes: industry layer

Combining Hardware and Materials, China funds the physical stack: 89% of the CN cohort sits in those two categories versus 64% of the US cohort. Service and Platform together account for 27% of US funded AM-ecosystem companies and 8% of CN, a 3.4x US tilt toward the digital-manufacturing layer (on-demand services, marketplaces, build-process orchestration).

In absolute terms the Hardware layer is larger in China: 104 CN Tier B hardware companies against 78 in the United States, from a smaller cohort. The two countries are not racing each other on equipment formation count. They differ on which layers above hardware get funded enough to count as separate businesses.

Finding 4: Process focus

Metal AM is more dominant in the Chinese cohort (52.7% vs 37% in the US), and the CN hero-anchor list is the same story in dollars: BLT, CNMNC/OTIC, Farsoon, Yudding and Metalink are all metal equipment or powder businesses. Polymer AM is slightly more prevalent in the US (37% vs 31.5%). The cleanest divergence is in Bioprinting, where the US cohort tags at 7.3% versus China's 0.7%, an order-of-magnitude gap. Novel Process and Ceramics/Composites are close in both cohorts.

Finding 5: The years inside the window

Hero anchors: where the primary raises landed

The fifteen largest disclosed primary raises in each country, 2023 through 2026-09-08, one round per company. Each card carries an AM-relevance label and a one-sentence rationale, so the reader can weigh each round on its own terms rather than reading the AMPulse classification as the last word.

The two columns differ in three visible ways. First, 9 of the 15 CN top rounds are primary-AM companies (equipment, materials or service businesses) versus 8 of the 15 US top rounds. Second, 11 CN top rounds are IPOs, post-IPO placements or private placements versus 3 US top rounds: the dominant CN capital-raise channel is public markets, the dominant US channel is private growth equity. Third, the top of the US column is materials companies for which AM is one market among several (Titan America's printable concrete line, NatureWorks's PLA); the first US round that is AM through and through is Divergent at #3. On the CN side the largest raise, JLCPCB's 2026 listing, is an online manufacturing platform whose core is PCB prototyping, and the first pure AM round is BLT at #2.

Ownership transfers, not counted above

The table below records identified ownership transfers in the wider census. The first edition incorrectly included some ownership transactions in funding totals. These events are changes of ownership: the money went to the selling shareholders, not into the company. 7 carry a disclosed amount and 13 do not; the US rows with an amount sum to $42.5B, more than ten times the US primary total, but that ledger is drawn from the wider census, not the original Tier B population. It cannot be subtracted from the original headline to explain the revision. Desktop Metal's sale to Nano Dimension closed on 2025-04-02 and the company filed for Chapter 11 on 2025-07-28, so the $183M in that row was the last price paid for the business, not a bet on it.

Sensitivity check: does the headline survive a stricter cut?

Nine findings re-run under Tier A (inclusive), Tier B (default) and Tier C (strict AM tagging). 7 of the nine hold direction across all three tiers. The two that do not are the report's two most-changed numbers, and both fail for a stated reason. Disclosed capital reads $148.3B vs $8.0B at Tier A, because Tier A admits AM users and one launch-vehicle listing carries half of it; at Tier B it is $3.9B vs $3.9B, parity; at Tier C, $2.5B vs $1.7B, a US lead again, because the process-level tag requirement drops CN rows tagged only at the head level (BLT and Farsoon among them) faster than US rows. The ex-grant median holds at Tier B and C and closes at Tier A for the same reason: the AM-user rounds it admits are large. The three structural findings (CN's heavier physical-stack share, CN's heavier Metal-AM tilt, public markets as the top CN source) hold at every tier, and so, now that transfers are out, does public markets as the top labeled US source.

Counter-thesis

A determined skeptic could attack the headline three ways.

1. CN disclosure rate is low and the absolute capital numbers cannot be trusted. The CN cohort's round disclosure rate is 34.8% versus 72.3% for the US, so the CN $3.90B figure is a floor. That cuts against the old headline, not this one: a fuller CN ledger would put China ahead in primary dollars, and the structural findings (which sources, which stages, which layers, which processes) are computed within each country and do not move with the total.

2. AMPulse classification leakage inflates one side. The AM-relevance labels on the hero anchors and the Tier A/B/C cuts exist for this. Of the US top 15, 7 carry adjacent or edge labels (Titan America, NatureWorks, Principal Mineral, Forge Nano, 6K, Firestorm Labs, Lumafield), summing to $1.36B. Of the CN top 15, 6 carry adjacent or edge labels (JLCPCB, Hymson, Tripo AI, Makeblock, HG Laser, Finework Technology), summing to $1.56B. Scrubbing adj/edge rounds from both top-15 lists (a top-15 scrub, not a re-cohort) leaves about $2.50B US and $2.34B CN, a ratio of 1.07x. The leakage that mattered was the one the first edition missed: AM users filed as suppliers, ownership transfers filed as raises, and one investment into a non-AM subsidiary, whose combined contribution cannot be reconstructed from the retained May per-company data. The net US headline decrease is $9.17B, with changes to transaction typing, supplier scope, deduplication, currencies and the reporting window. Identified ownership transfers and excluded companies are now typed out of the headline, and the supplier-vs-user boundary is a rule in the methodology rather than a label on a card.

3. The window is too narrow, or the wrong one. The window is 2023-01 through 2026-09-08, 44 months. Finding 5 is the answer: inside the window the two countries are moving in opposite directions, so a 2023-only cut would show a Chinese lead and a 2025-only cut a US lead of four to one. The Tier B totals summarize different annual patterns, and the per-year series is the figure to cite when the direction matters more than the level.

Method appendix

Source. AMPulse's tracked-company database, US and China subset, with funding-round history scoped to 2023-01-01 through 2026-09-08. The census covers 2784 companies, of which 620 have at least one primary round in the window.

Snapshot integrity. All numerical claims in the body are computed from a single snapshot extract. The data files behind this report are frozen as of the snapshot date and versioned alongside the report, so subsequent re-extracts do not silently mutate the values cited here.

Revision bridge. The shipped May 19 Tier B US total was $13.036B and the September 8 total is $3.864B, a net decrease of $9.173B. CN moved from $3.246B to $3.899B. The observation window also extended from May 19 to September 8 and cohort membership changed. Retained artifacts do not provide the May per-round membership needed to attribute that change across exclusion, transaction typing, deduplication and currency rules. The $42.5B wider US ownership-transfer ledger is therefore not the amount removed from the original headline. Dated revision record.

Corrections policy. Corrections are applied in place to the data files and recorded, with the figures that moved, in the manifest's revision notes; earlier notes are never removed. The 2026-09-08 re-edit (transaction typing, census filter, same-round collapse, closed window) is the first post-publication correction and moved the US headline from $13.0B to $3.86B.

The September 30, 2026 correction replaces the earlier Section 849 interpretation with its enacted equipment-procurement scope. The frozen analytical data are unchanged.

Lee, S. (2026). Two Capital Stacks. AMPulse Data Reports. https://www.ampulse.online/reports/us-china-capital-stacks-2026