What does a state-anchored deep tech fund look like once it stops being an announcement and starts moving money? On 26 December the State Council published the disbursement structure of China’s National Venture Capital Guidance Fund, and it is the clearest public account of how a vehicle of this kind gets built: which tier holds the mandate, which tier invests, and what sits on the other side of the balance sheet. Latin America already has the development banks, the innovation agencies, the accelerators and the patient capital. What it has not assembled is the wiring between them.

A Mandate, a Balance Sheet and Three Tiers

The fund was reported by The Quantum Insider on 7 March 2025 as a US$138 billion public-private vehicle covering quantum computing, advanced semiconductors, artificial intelligence and next-generation renewables. We cited it in our report at page 104 and have since re-verified the headline: RMB 1 trillion, about US$138 billion, over twenty years, at least 70% earmarked for seed and early stage, financed through ultra-long special treasury bonds, and organised in three tiers, national, regional and sub-fund.

Two features do more work than the dollar figure. The first is the funding source. Ultra-long sovereign paper gives the vehicle a liability profile matched to the asset class, which is the first wall every patient-capital proposal in Latin America runs into: deep tech demands long holding periods, and short money cannot wait for them. The second is the 70% floor. An early-stage allocation written into the instrument cannot be quietly renegotiated by managers who find later-stage deals easier to underwrite. Most regional strategies put that commitment in a document, where it lasts as long as the minister who signed it.

The three tiers are the part worth copying most carefully. The national tier sets the mandate and provides the balance sheet. The regional tier sits inside the city clusters where allocation decisions have to be made, rather than in a capital. The sub-fund tier invests. Nothing in that shape requires a Chinese balance sheet. It requires somebody to decide which tier does which job before the money is authorised.

How Much Has Moved, and Where It Sits

The December notice gives the numbers to track. Three regional sub-funds are now stood up: Beijing-Tianjin-Hebei at RMB 29.6 billion, or US$4.26 billion; the Yangtze River Delta at RMB 47.1 billion, or US$6.77 billion; and the Greater Bay Area at RMB 45.05 billion, or US$6.47 billion. Together they come to roughly US$17.5 billion committed as of December 2025, nine months into a twenty-year authorisation. The notice does not say how much has been committed to named managers, nor how much has reached companies, and we put no number on either.

That distinction is the design lesson for anyone building a comparable vehicle here. An authorisation, a commitment to a sub-fund and a cheque to a founder are three different events, and a headline total is readable only when the line it sits on says which. US$138 billion is the authorisation; US$17.5 billion is what has been committed.

One clarification allocators in the region should carry with them: this is a domestic programme, and all three sub-funds sit inside China. Our report notes at page 105 that two months after the fund was announced, China rolled out a US$9 billion credit facility for Latin American governments, citing Eduardo Baptista’s Reuters reporting from 2025; that figure we extracted from the report and have not re-verified, and our record does not say how much has been drawn. Sovereign credit lands on a finance ministry’s balance sheet and finances procurement, infrastructure and trade. Venture capital lands on a company’s cap table. A founder in São Paulo or Mexico City raising a seed round should size the two separately.

The fund is also the sequel to a reallocation rather than the start of one. GPCA’s 2023 Emerging Trends in Asia reports deep tech rising from 15% to 71% of total domestic tech investment in China between 2017 and 2022, a series that ends before the guidance fund existed. The capital had already turned; the fund is the form the turn took.

Latin America Already Runs the Small Version of This

The mandate and sub-fund tiers are not missing here, and Uruguay is the cleanest example. It enacted Law 20.075 in 2022, formally approved in 2023, prioritising advanced digital platforms, biotechnology and green tech, then inaugurated the Uruguay Innovation Hub in May 2024. Alongside it the government allocated US$10 million for a 1:1 matching-funds programme under the Hub’s management, in which vetted venture firms and angels co-invest through convertible notes, with disbursements tied to technical and commercial milestones. That is milestone-gated disbursement at the lowest tier, already written and running.

Argentina’s 2017 Entrepreneurs Law established FONDCE, a fund of funds underwriting science-based accelerators and early-stage venture firms with reimbursable loans and partial operating grants. Within the law’s first two years, five deep tech accelerators backed by FONDCE had invested in 79% of the country’s investor-funded deep tech ventures, on a 2021 account we carry from our report and have not re-verified. Chile’s CORFO has run matching-fund programmes since the early 1990s, with reimbursable grants covering 40 to 65% of private R&D costs and FONDEF co-financing pre-competitive research between universities and firms; Chilean programmes count NotCo, which applies AI to plant-based food, and Autofact, an automotive analytics platform, among their alumni. In Brazil, FINEP is offering US$500 million with BNDES to get multinational and national firms to set up R&D hubs in the country.

The regional tier has begun to appear too. In 2025 IDB Lab committed US$3 million to GridX Fund II, a US$30 million vehicle led by Matías Peire targeting up to 75 deep tech startups; IDB Invest now manages the US$1 billion JICA-TADAC fund with Japan’s JICA, another figure we carry from the report without re-verification. What this tier compounds into is on record: Israel’s 1993 Yozma programme injected US$100 million into ten venture funds at 40% public and 60% private, on the IDB’s account, seeding an industry now managing over US$10 billion.

The Tier That Is Missing Is the Balance Sheet, and the Region Has One

China chose ultra-long treasury bonds. Latin America has a different long-dated pool in plain view. On the OECD’s numbers, pension assets under management in Chile, Mexico, Brazil, Peru and Uruguay total roughly US$1 trillion, and redirecting even 1% would mobilise US$10 billion for domestic deep tech; that is a figure we extracted from our report and have not re-verified. Our own calculation puts that 1% at about four times the US$2.5 billion deployed into Latin American deep tech since 2018. Pension liabilities and venture fund lives sit far closer together than a bank’s book ever will.

The precedents are live. Under the first phase of France’s Tibi initiative, from 2020 to 2022, institutional investors committed EUR 6.4 billion to approved funds, per DG Tresor. The UK’s Mansion House Accord of May 2025 pledged to allocate 10% of pension provided portfolios to private markets by 2030, with at least 5% ring-fenced for UK assets, voluntary and fiduciary-aligned; it runs to 2030, it is early, and we have not re-verified its terms. Recommendation 12 of our report sets out the regional version:

Governments should take the lead in mobilizing long-horizon capital for Deep Tech. A practical approach is establishing a public-anchored Fund-of-Funds.

The design difference we would keep is competitive mandates to independent managers, awarded in the open, which is the part the December notice says nothing about. Public capital sets the floor and takes the first risk; independent managers pick the companies.

What We Are Watching, and What to Do With It

Committed is not disbursed. US$17.5 billion nine months into a twenty-year vehicle establishes capacity and intent, and publishing a structure is the easy part of a fund of funds; manager selection is harder, and capital reaching founders slower than either. Several regional figures above carry the same caution in another form: the Uruguayan allocation, the Argentine 79%, the CORFO range, the Yozma figures, the FINEP and BNDES US$500 million and the OECD pension total we extracted from the report and have not re-verified since, and we mark them so rather than round them off.

The measurement we are building next follows from that. Recommendation 16 of our report is a mapping of the region’s funders, and the field it most needs is the one the Chinese notice omits: for every public vehicle in the region, how much is authorised, how much is committed to named managers, how much has reached a company. Nobody publishes that series today. We intend to.

For a ministry or a development bank, the transferable design costs nothing: write the early-stage floor into the instrument rather than the strategy, match the funding instrument to the holding period, and push allocation decisions down to a tier inside the ecosystem. Uruguay has the milestone gating, Argentina the statute, Chile the matching mechanics, Brazil the scale. Assembling them is smaller work than inventing them.

For allocators, the number worth tracking through 2026 is how much of China’s authorisation converts into commitments to named managers, since that is where a structure becomes competition for deals. For pension trustees and supervisors here, the nearer question is what a fiduciary-compatible vehicle would look like, answerable in 2026.

Our pensions chapter is open on the report site, and we are convening trustees, supervisors and deep tech managers around this design. If you are building one of these vehicles, tell us how you are specifying the tiers.