Battery Swap: The Rental Works, the Station Does Not

Two things happened in September 2026 that together answer a question the Chinese EV industry has been arguing about for eight years.

On 22 September, XPeng and Halo Energy opened a 1 MW flash-charging station in San Po Kong, Hong Kong — 16 times the power of a conventional 60 kW unit, adding roughly 450 km in nine minutes on a 5C-capable car. XPeng's vice chairman Brian Gu cut the ribbon. Six days later, NIO signed definitive agreements giving Geely 30% of NIO Power, the battery-swapping and charging subsidiary, in exchange for Geely's commercial-mobility swap business plus RMB 640 million in cash. The deal valued NIO Power at about RMB 16 billion post-money.

One company put another megawatt into charging. The other sold a third of its swapping business to fund the next 6,000 stations. Both are defensible. But they are not the same business, and the difference is worth being precise about — because "is battery swapping viable?" has three different answers depending on which of the three businesses underneath it you mean.

The three businesses wearing one name

Battery swapping as NIO practises it is not one business. It is three, with different economics, different capital requirements and different owners:

1. The swap station — a building, a grid connection, 20-odd batteries and a robotic changer. Sells electricity plus a service fee.
2. The battery rental (BaaS) — the pack is removed from the car's purchase price and rented monthly, forever, by a company that owns the asset.
3. The car sale — the network exists partly to sell RMB 300,000+ cars, and swapping is the differentiator that closes them.

The first is a grind. The second is a genuinely good business. The third is why the first two get funded at all.

Where NIO actually stands

The scale is not in question. As of the Geely announcement, NIO had invested over RMB 20 billion in charging and swapping infrastructure and operated 9,433 stations — 4,126 swap stations and 5,307 charging stations with 30,598 piles — having delivered more than 220 million energy services, including 125 million battery swaps (NIO's own release and station-count announcement). It passed 100 million swaps in February 2026. It plans 10,000 swap stations by 2030.

Investors have taken the other side of that spend. Geely's stake can be reduced to no less than 20% if it misses operational milestones, and Geely holds an option to put in a further RMB 640 million to reach 34%. That is a claw-back structure, not a growth-equity handshake: the buyer gets paid in equity for volume it has not yet delivered, and the equity shrinks if the volume does not arrive.

Meanwhile the parent is only just at the line. NIO posted its first quarterly profit in Q4 2025 — RMB 2.827 billion net on RMB 34.65 billion of revenue — then slipped back to a RMB 528 million net loss in Q2 2026 (adjusted net income +RMB 26.1 million). The company has been explicit that the swap network is a cost centre by choice: on the Q1 2026 call, William Li said plainly that "for the short term, the profitability of the power swap network is not our primary focus."

The station arithmetic

NIO prices a swap as (electricity + service fee) × kWh delivered, with a 20 kWh floor on the service fee (the pricing change NIO announced in September 2024). Service fees run around 0.3–0.4 RMB/kWh. So walk one mid-spec fourth-generation station, using NIO's own publicly stated operating costs:

• Revenue per swap: ~45 kWh × ~1.25 RMB/kWh — about 56 RMB
• Cost of that electricity at retail industrial rates (~0.8 RMB/kWh) — about 36 RMB
• Gross profit per swap: about 20 RMB. That is the service fee plus whatever discount you negotiated on power, and nothing else.
• Fixed costs — site rent ~20k, labour ~15k, maintenance ~8k, grid capacity charges ~10k, battery capacity cost ~10k — about RMB 55–60k/month, or roughly 1,900 RMB/day.

Run that:

Swaps/day	Gross profit @ 20 RMB	Fixed cost	Station net
30	600	(1,900)	−1,300
45	900	(1,900)	−1,000
60	1,200	(1,900)	−700
80	1,600	(1,900)	−300
100	2,000	(1,900)	+100

The table says a swap station breaks even around 95 swaps a day. NIO's stated breakeven is 60–70, and the CFO's earlier framing was the same. The gap between my arithmetic and theirs is the entire story, and it is not a technology gap. It is power procurement.

A station holding 20-plus packs at ~45 kWh each is a 1–2 MWh battery bank. Buy at valley rates overnight (0.3 RMB/kWh), discharge at peak (0.9–1.2), and you earn a 0.5–0.8 RMB/kWh spread on every kilowatt-hour you cycle — including the ones sitting idle in the racks. That second revenue line frequently exceeds the swap service fee itself. NIO's CFO named it directly on the Q1 2026 call: "off-peak and off-tick charging and also grid interactions and also electricity trading."

Which is the honest conclusion on stations: the same building is a losing energy-retail outlet or a decent energy-trading asset depending on how it buys power — not on how fast the robot changes a pack. Station capital cost has already fallen from ~RMB 3 million (first generation) to under RMB 1.5 million (fourth) to roughly RMB 1 million (fifth), with operations and maintenance down about 30%. Capex is not the problem. Utilisation and the price of electrons are the problems.

And utilisation is not there yet. On the Q1 2026 earnings call, Li put average stations at about 45 swaps a day at holiday peaks and about 30 on normal days — against a 45-swap breakeven for fifth-generation stations. Shanghai's mature sites clear 80+, some over 300. Elsewhere the network runs well below the line, and Chinese industry estimates put over 80% of swap stations still lossmaking, with some county-level sites under 7% equipment utilisation.

The part that actually works

Now the battery rental, which is a different animal entirely.

NIO's BaaS terms: the buyer takes RMB 70,000 off a 75 kWh car or RMB 108,000 off a 100 kWh car, and pays RMB 728 or RMB 1,128 per month respectively (NIO's published BaaS terms). The pack is not owned by NIO and not owned by the driver — it belongs to Wuhan Weineng, a separate company founded in August 2020 with CATL as a shareholder (NIO ~19.4%, CATL ~9%). More than 80% of NIO buyers now take the rental option.

Do the arithmetic on a single pack: RMB 70,000 of battery earning RMB 8,736 a year is a ~12.5% gross yield. Borrow at 3–4% — and Weineng borrows cheap, with a shareholder register now heavily weighted to local state capital — and you are earning a wide positive spread on an asset with an 8–10 year life and a real residual and recycling value. That is infrastructure finance, and it behaves like infrastructure finance.

The evidence that it is bankable rather than merely theoretical is in what it has been able to raise. Weineng's managed battery assets passed 42 GWh serving 550,000+ subscribers, up from 20 GWh and 200,000 users in December 2024 — it has doubled in a year. It closed roughly RMB 2 billion across its C round, its eighth raise. And in February 2026 it issued what the market calls the world's first "holding-type" power-battery green ABS/REIT — RMB 501 million — securitising battery rental cash flows.

That securitisation is the single most informative datapoint in the entire swapping debate. You cannot package a swap station's P&L into a rated security. You can package battery rental cash flows. The capital markets have now said so with money.

The structural elegance, and the catch, are the same fact: Weineng holds the residual-value risk and carries the funding burden; NIO gets cash upfront and keeps its balance sheet clean. Excellent for NIO. A permanent treadmill for Weineng, which must keep raising against a depreciating asset base to keep buying packs.

The natural experiment: Aulton

If you want the counter-case for standalone swapping, the HKEX prospectus of Aulton — the largest independent third-party swapper — is the cleanest evidence available.

• Own stations: 321 → 291 → 236 → 214. It has been closing faster than building.
• Cumulative losses of roughly RMB 2 billion.
• Only 38–39 of 236 stations at breakeven.
• Own-station gross margins: −16.2%, −20.1%, −21.4%, −18.6%.
• Customer base: taxi and ride-hail drivers, the most price-sensitive buyers in the market — so no pricing power.

The prospectus states the root cause plainly: Aulton has no car-sales profit to subsidise the network and no authority over battery standards, so it had to build custom packs for each automaker and never reached scale.

This is as close to a controlled trial as the industry offers. Swapping with captive demand and a car margin attached is survivable. Swapping as an unaffiliated utility is negative gross margin.

The one place independent swapping plausibly works is heavy trucks, where packs are enormous, duty cycles are predictable, routes repeat, and fleets buy on cost per kilometre rather than convenience. That is a different business from consumer swapping, and Geely's contribution to NIO Power — a commercial-mobility swap operator — is a bet on exactly that.

Where XPeng is, and why

XPeng's position has been consistent for four years and is the direct opposite bet.

He Xiaopeng, asked whether XPeng would build swap stations: "We are not considering battery swap now — everything going forward is ultra-fast charging." His framing is that these are different tools for different segments rather than rivals — and on the long-run view he argues fast charging is the better bet for global reach and energy security.

The company has put its money where that is. Roughly 3,000 self-operated charging stations in China, including around 1,800 S5 ultra-fast stations by mid-2026 — more than any automaker except Tesla. 5C is standard across the range since the 2025 G6 and G9, with the G6 charging 10–80% in about 12 minutes and the G9 in about 10 (XPeng). The Hong Kong 1 MW station adds 450 km in nine minutes. And the network is being shared: a memorandum with Volkswagen China opened both companies' ultra-fast networks to each other, 23,000+ chargers across 420 cities.

Once, XPeng left a door open. Visiting NIO's Hefei plant in May 2024, He Xiaopeng said that when NIO passed 5,000 stations, it would seriously consider joining a swap network — and Li, laughing, said he would leave an interface for them. NIO passed 4,000 in August 2026 and targets 4,600+ by year-end, so the threshold is genuinely approaching.

But read the condition rather than the sentence. What He Xiaopeng set was not a timeline; it was a free option. If NIO proves the network is dense enough, the standards settled and the model sustainable, XPeng joins and inherits infrastructure it did not pay to build. If NIO stalls, XPeng loses nothing and keeps saying "not yet." And the alliance has been thin on delivery — despite announcements with Changan, Geely, Chery, JAC, GAC, Lotus and FAW, not one member other than NIO has shipped a volume swappable model. XPeng's patience has cost it nothing so far.

What would have to be true

Swapping is not a technology question, and the eight-year argument mostly misses that. Five variables decide it:

1. Swaps per station per day. Below 45 it does not work; above 80 it is a good asset. Everything else is commentary.
2. How you buy electricity. Valley charging plus a wholesale contract is the difference between +0.5 and −0.2 RMB/kWh. Paying retail rates makes the site unwinnable.
3. Captive or contracted volume. Cars you sell, or fleets on contract. Not anonymous passing traffic. Aulton is the proof.
4. Whether you own the battery standard. If the packs are yours you get scale and a rental annuity. If you must custom-build for everyone else's packs, you get Aulton's gross margins.
5. Your cost of capital for the battery fleet. The rental annuity only clears if you fund it below ~5%. Weineng's state-capital register and its ABS access are not incidental — they are the business.

There is also a competitive clock running. BYD's second-generation Blade and "Flash Charge China" plan targets 20,000 stations by end-2026 at up to 1,500 kW (Chinese coverage of the launch). Ultra-fast charging threatens swapping least where swapping is strongest — dense urban networks with captive, high-mileage users — and most where swapping is already weakest: thin-traffic sites in lower-tier cities where the unit economics were underwater before anyone built a 1 MW charger.

The bottom line

Battery swap is viable, and battery rental is the reason — not battery swapping. The rental layer is a ~12.5%-yield, asset-backed annuity that has already been securitised, and it decouples the driver's cost from battery residual risk. The station layer is a marginal energy-retail business that only works when it is run as a behind-the-meter storage asset with cheap power and enough cars, and it demonstrably does not work for an operator without a car brand or a battery standard.

Strip away the branding and NIO's model is not really "sell swaps." It is: take RMB 70,000 of battery off the car price, finance it at 3%, rent it at 12.5%, run the station as a storage asset that arbitrages peak and valley power, and treat the swap itself as the retention mechanism that keeps the car — and the annuity — attached. Optimised that way, it works. Optimised as a standalone refuelling business, it does not, and the last eight years of independent swap operators are the evidence.

XPeng has read the same numbers and concluded that if battery rental is the good part and stations are the hard part, it would rather compete on the part where the physics keep improving every year and the capital cost per unit of range delivered keeps falling. Both companies may turn out to be right about different customers.

EVConduit analysis. Figures are taken from company filings, releases and earnings calls (NIO, Geely Holding, CATL, XPeng, Aulton) and from trade coverage linked in the text, including CnEVPost, Reuters, Xinhua's Economic Reference and OFweek. Currency figures are RMB unless stated. Unit economics are our own estimates from publicly stated prices, costs and utilisation, not company guidance.