China's AI Boom and the Indonesian Mineral Channel: Compute, Batteries, Nickel, and the Rupiah

Rupiah Stability Watch · 2026-10-16

The Signal

On 14 October 2026, Nature published two features that mark a threshold. "China's AI research is booming: what it's like to be on the front line" documents a research ecosystem that now trains 51,000 active AI researchers — nearly four times the United States' 14,000 — and produces the largest share of global AI publications. The companion piece, "China's extraordinary rise in science and how it will change the world," frames this as a systemic shift: China has moved from marginal participant to co-leader in the global knowledge order.

The signal is not the paper count. It is the infrastructure that paper count demands.

China's AI firms — DeepSeek, Moonshot AI, Zhipu, Baichuan, and the hyperscalers behind them — are building compute clusters at a pace that makes power the binding constraint. Wood Mackenzie forecasts China's data-centre electricity consumption will nearly quadruple to 774 TWh by 2030, up from roughly 200 TWh in 2020. That is not a projection of server count; it is a projection of firm, round-the-clock power. Intermittent renewables cannot meet it alone. The grid needs storage — lots of it, and increasingly long-duration — and storage needs nickel, cobalt, and lithium.

The chain is short and unforgiving: AI compute → firm power → stationary storage → battery minerals → Indonesia.

The Chain

Compute demands firm power

The 774 TWh figure is not speculative. It is the sum of announced and under-construction data-centre capacity across China's eastern seaboard and western "computing hubs" (Guizhou, Inner Mongolia, Gansu), each mandated by the "East Data, West Computing" policy to absorb surplus renewable generation. But AI workloads — training runs that last weeks, inference latency budgets measured in milliseconds — do not tolerate curtailment. They need firm power: nuclear, coal, hydro, geothermal, and increasingly, battery storage that can shift renewable output across hours and days.

Firm power demands stationary storage

China's National Energy Administration (NEA) reports 73.76 GW / 168 GWh of new-type energy storage operational at end-2024, a 130% year-on-year increase. In 2025 alone, CNESA data show 66.43 GW / 189.48 GWh commissioned — a 52% rise in power and 73% in energy capacity. The average duration climbed to 2.58 hours (from 2.11 hours in 2021), confirming a structural shift toward longer-duration configurations. Standalone storage now accounts for 58% of new capacity, overtaking renewables-paired and user-side applications.

Lithium iron phosphate (LFP) dominates >98% of new-type installed capacity. LFP cathodes require no cobalt, but they do require lithium — and for the nickel-rich NMC chemistries that still power premium EVs and high-energy-density stationary systems, they require nickel.

Stationary storage demands battery minerals

China's EV battery demand is a known quantity. Less appreciated is the stationary storage multiplier. At 189 GWh of new storage in 2025, if even one-third uses nickel-containing chemistries (NMC, NCA), that is ~60 GWh of nickel-intensive demand annually — and rising. Each GWh of NMC 811 consumes roughly 450–500 tonnes of contained nickel. The arithmetic is brutal: 27,000–30,000 tonnes of nickel per year from stationary storage alone, on top of EV demand that already exceeds 1.5 million tonnes globally.

Indonesia supplies 49% of global mined nickel (1.8 Mt of 3.7 Mt in 2024, per USGS) and holds 55 Mt of reserves — the largest national endowment. The laterite ores in Sulawesi and Halmahera feed two downstream paths: saprolite to rotary-kiln electric furnaces (RKEF) for nickel pig iron (NPI) and ferronickel (stainless steel), and limonite to high-pressure acid leach (HPAL) for mixed hydroxide precipitate (MHP) and nickel sulphate (batteries).

Indonesia is the only country that can scale nickel supply at the pace China's storage buildout requires.

The Indonesian Node

Nickel dominance, Chinese midstream control

Five HPAL plants operated in Indonesia in 2025:

Two industrial parks anchor the complex: Indonesia Morowali Industrial Park (IMIP) and Indonesia Weda Bay Industrial Park (IWIP). Both are Chinese-anchored. Tsingshan Holding Group is the common denominator — the world's largest stainless steel producer, the architect of the NPI-to-HPAL conversion bridge, and the capital partner for Huayou, GEM, Lygend, and CATL's Brunp recycling arm.

The downstream is not Indonesian-owned. It is Chinese-capital, Chinese-technology, Chinese-offtake — built on Indonesian ore.

CATL Karawang: the test case

In June 2025, CATL and partners (Indonesia Battery Corporation, PT Aneka Tambang) broke ground on a $6 billion, 6.9 GWh Phase 1 battery cell plant in Karawang, West Java — the first full value-chain integration (precursor → cathode → cell) in Southeast Asia. Phase 2 targets 15 GWh. The plant is a litmus test for three questions:

1. Can Indonesia move beyond MHP export to cell production?
2. Will Chinese technology transfer stick, or remain enclave?
3. Does the IRA foreign-entity-of-concern (FEOC) rule block US-market access, and does that matter?

The FEOC rule (effective January 2025) excludes batteries from producers ≥25% owned, controlled, or directed by China, Russia, Iran, or North Korea. CATL Karawang is a CATL subsidiary. Its output is FEOC-ineligible for US 30D tax credits. The EU Critical Raw Materials Act similarly scrutinises Chinese-controlled supply chains. Indonesia has no FTA with the US to qualify under the critical-mineral exemption. The Karawang output is structurally locked into the Chinese domestic and non-Western export market.

PP 24/2026: the state reasserts control

Government Regulation (PP) No. 24/2026, effective 1 June 2026, designates DSI (the state export enterprise) as the mandatory sole export intermediary for coal, palm oil, and ferro-alloys — including nickel products. Every shipment must pass through DSI's five digital monitoring systems (CEISA, SINSW, INATRADE, SiMoDIS, MOMS). The constitutional grounding is Article 33: natural resources controlled by the state for the greatest benefit of the people.

The regulation's stated aim is to halt under-invoicing and transfer pricing. Its structural effect is to centralise pricing power in Jakarta — but the buyer on the other side remains overwhelmingly Chinese. If DSI cannot negotiate better terms than individual smelters could, the regulation captures transparency without capturing value.

The Rupiah Channels

1. Export earnings: volume yes, value uncertain

Mysteel data (Ministry of Trade) show Indonesia exported 8.21 Mt of nickel products in Jan–Jul 2025, of which 95.7% went to China. Full-year 2024 nickel export value was ~$33.9 billion; 2025 is tracking above $40 billion. This is genuine FX revenue — a direct credit to the current account.

But the price Indonesia receives is not the LME nickel price. It is the Chinese spot contract price (MHP payable ~78–82% of contained nickel, nickel sulphate ~85–88%), negotiated by Chinese offtakers who also own the smelters. When Tsingshan, Huayou, and GEM are both the processor and the buyer, the "market price" is an internal transfer price. Indonesia captures volume; China captures the margin between MHP and cathode.

2. FDI & processing investment: dollar inflows during construction

Chinese firms have committed > $30 billion in downstream capital spending to 2025 (IMIP, IWIP, Karawang, Pomalaa, QMB Phase II, Danantara-GEM $1.42B HPAL). These are dollar inflows during construction — equipment imports, engineering services, labour — that support the rupiah while the plants are being built. Once operational, the FDI stock generates profit repatriation outflows (dividends, management fees, technology licensing) that reverse the flow.

The net present value of the FDI channel depends on:

• Local content fulfilment (how much capex stays in Indonesia)
• Technology transfer clauses (are Indonesian engineers gaining HPAL/cathode/cell know-how?)
• Profit retention (does the JV structure allow earnings to be reinvested locally?)

CATL Karawang's 6.9 GWh phase is the first real test. Early signals: Indonesian engineers are being trained in Ningde, but core cathode and cell IP remains in China. The enclave risk is non-trivial.

3. Terms of trade & pricing power: who sets the price?

Three price regimes coexist:

Indonesia is a price taker in a market where the buyer and the processor are the same entity. PP 24/2026 gives DSI the legal right to intermediate, but without an alternative buyer base (Korea, Japan, EU demand is growing but from a small base), DSI's bargaining power is asymmetric.

4. Energy collateral: the coal cushion persists

IEA data: Indonesia exported 555 Mt of thermal coal in 2024 (up 6% YoY), primarily to China and India. China's data-centre buildout (774 TWh by 2030) requires firm power in China. That firm power is still ~60% coal-fired. Indonesian thermal coal remains the marginal swing supplier to China's coastal and southern grids.

But the coal channel is fading. IEA's 2026 mid-year update notes seaborne thermal coal demand peaks mid-decade. Indonesia's 2025 coal exports fell 19.7% to $24.48 billion (from $30.49 billion in 2024). The coal cushion that historically absorbed current-account shocks is becoming a cost-of-capital question — as documented in this organisation's "Climate Accountability and the Rupiah" (Oct 8). International financiers price Indonesian coal exposure into sovereign risk premiums. The nickel channel must replace the coal channel before the coal channel disappears.

5. Geopolitical risk premium: the FEOC discount

US-China tech competition adds a structural risk premium to Indonesia's mineral assets. The IRA FEOC rule, EU CRMA, and potential Section 232 tariff investigations create three scenarios:

The Korea pivot (referenced in Deluair's "Indonesia nickel 2026" brief) is nascent: POSCO, LG Energy Solution, and Hyundai are evaluating non-Chinese HPAL and precursor routes. But Korean and Japanese capital moves slower, demands stricter ESG, and requires IRA/EU compliance — which Chinese-controlled Indonesian assets struggle to meet. The geopolitical risk premium is currently priced as a discount on Indonesian mineral assets, not a premium for scarcity.

6. Domestic value capture: the GR 24/2026 test

Indonesia's downstreaming mandates (2020 ore export ban, 2024 ferro-alloy export restrictions, PP 24/2026) have forced $30B+ of Chinese smelter investment. The question is whether the next layer — precursor, cathode, cell, pack — follows.

CATL Karawang 6.9 GWh is the test case. If it delivers:

• Indonesian engineers operating cathode and cell lines independently within 3 years

• 40% local content by value (not just casing and assembly)

• A second Korean/Japanese cell plant committing to Indonesian precursor feedstock

...then the value-capture thesis holds. If instead Karawang operates as a CATL enclave — Chinese management, Chinese core IP, FEOC-locked output, profits repatriated — then Indonesia has built the world's largest nickel refinery complex without moving up the value chain.

The Danantara sovereign wealth fund (launched Feb 2025) is the policy lever. Its $1.42B HPAL partnership with GEM signals state willingness to co-invest. But Danantara's mandate is commercial return, not industrial policy. The tension between maximising state revenue (DSI margin) and maximising domestic value capture (local content, tech transfer) is unresolved.

Data Box: Key Figures

The Watchlist: Signposts for the Next 6–12 Months

What the Evidence Supports

1. China's AI boom is real and power-intensive. The 774 TWh forecast is grounded in announced capacity and policy directives. The storage buildout (66 GW added in 2025) is measured, not projected.

2. Indonesia is the indispensable upstream supplier. 49% of mined supply, 55 Mt reserves, and the only operating HPAL-at-scale outside China. No other jurisdiction can ramp MHP/nickel sulphate fast enough.

3. Export earnings are material. >$40B annualised nickel revenue is ~10% of Indonesia's goods exports. This is a genuine FX cushion.

4. Chinese capital controls the midstream. Tsingshan, Huayou, GEM, CATL Brunp own the HPAL, precursor, and cathode capacity. DSI intermediates but does not set the price.

5. The coal channel is declining structurally. 19.7% export value drop in 2025; IEA sees seaborne thermal peaking mid-decade. The nickel channel must carry more weight.

6. PP 24/2026 centralises control but not yet value. The regulatory architecture exists; the commercial outcome depends on whether DSI can negotiate better than atomised smelters.

What the Evidence Does Not Support

• That Indonesia captures cathode/cell margins today. CATL Karawang is the first cell plant; its output is FEOC-ineligible. No precursor or cathode plant operates at scale under Indonesian control.

• That Chinese demand is price-inelastic. Chinese smelters (Tsingshan, Huayou, GEM) have cut HPAL output by ≥10% since March 2026 due to sulphur costs and weak nickel prices (SMM, idnFinancials). They respond to margins.

• That Western capital will replace Chinese capital on timeline. EU CRMA strategic projects and Korean JVs are in feasibility, not FID. The IRA FEOC rule makes Indonesian-Chinese JVs structurally ineligible for US subsidies.

• That PP 24/2026 eliminates transfer pricing. It creates the monitoring infrastructure to detect it. Enforcement capacity and political will are untested.

The Least-Harm Path

The rupiah's mineral channel is necessary but not sufficient. It delivers export volume and construction-phase FDI — both supportive. But without domestic value capture beyond MHP, the channel deepens a structural dependency: Indonesia sells ore to Chinese smelters who sell cathode to Chinese cell makers who sell batteries to Chinese data centres. The FX earned is real; the terms of trade are not in Indonesia's favour.

Three conditions would shift the balance:

1. DSI demonstrates pricing power — publishing quarterly realised prices that exceed Chinese spot indices by a measurable margin.
2. CATL Karawang (or a Korean/Japanese peer) hits >40% local content by value within 3 years — proving the enclave model can be broken.
3. A second, non-Chinese cell FID is announced in Indonesia — creating genuine offtake competition for Indonesian MHP.

Absent these, the rupiah receives a volume cushion without a value cushion. The current account improves; the terms of trade do not. The risk premium remains a discount.

What I'm Uncertain About

1. DSI's realised margin — No quarterly report has been published yet. The first (Q3 2026) will be the first hard evidence.

2. CATL Karawang's technology transfer depth — Training Indonesian engineers in Ningde is not the same as transferring cathode formulation IP. The JV agreement is not public.

3. China's 2026–2027 storage tender pipeline — CNESA reported 690 tenders in 2025 (down 10.4%). If 2026 tenders rebound, nickel demand accelerates; if they stall, the channel narrows.

4. Whether Korean/Japanese FID follows EU CRMA designation — The strategic project pipeline exists; the capital commitment does not yet.

5. The sulphur cost pass-through — HPAL operators cut output when sulphur spikes. If Indonesia cannot secure domestic sulphur supply (currently ~75% imported, per Deluair), the midstream bottleneck is Indonesian, not Chinese.

6. How the FEOC rule evolves under a potential US administration change — The 25% threshold is statutory; guidance can narrow or widen "control." A single Treasury FAQ could reclassify Indonesian JVs.

This analysis builds on prior work from this organisation: "Climate Accountability and the Rupiah" (Oct 8, 2026) on the coal-to-cost-of-capital transition; "Enhanced Geothermal and the Rupiah" (Oct 13, 2026) on firm power for compute; "Private Power for Compute" (Oct 14, 2026) on big-tech energy deals as FX infrastructure; and "Duration Is the New FX Variable" (Oct 7, 2026) on storage duration as a currency variable. The Korea contrast — "The AI Export Asymmetry: Korea's Semiconductor Boom, Indonesia's Compute Buildout, and the Rupiah" (Sep 29, 2026) — frames the alternative path where domestic chip design captures value that Indonesia's mineral channel has not yet secured.