Xpeng cuts product lines from 4 to 2 as R&D spending climbs and deliveries fall

The Chinese EV maker merged its I and F lines into the G line, leaving only the G and D lines, 36Kr reported, citing several industry sources. CnEVPost carried the report the same day.

The reorganisation is an internal change to product management and R&D responsibilities, not a cut to the number of models the company sells. Before the move, the G line mostly handled large SUVs such as the GX and G9L, the F line covered products including P-series sedans, the I line looked after overseas products, and the D line oversaw the Mona series.

After the restructuring, Mona stays under the D line and everything else sits under G. Management roles shifted as well. According to 36Kr, the former head of the F line now oversees product definition for the G line at the same rank, while the former head of the I line has moved to overseas affairs, excluding vehicle sales in overseas markets. CnEVPost said the report cited an industry source close to Xpeng.

Why now: falling deliveries against rising R&D

The logic is straightforward. Xpeng delivered 243,111 vehicles in the first eight months of 2026, down 10.49% from a year earlier, according to data compiled by CnEVPost. August alone brought 39,107 deliveries, up 3.71% year on year and 2.84% from July. That is three consecutive months of year-on-year growth, but not enough to cover the shortfall from earlier in the year.

The financial picture adds pressure. Xpeng posted a second-quarter net loss of 1.34 billion yuan ($199 million), wider than the 480 million yuan loss a year earlier but narrower than the 1.78 billion yuan loss in the first quarter. R&D expenses rose 32.1% year on year to 2.91 billion yuan in the quarter, which the company attributed mainly to spending on new models and AI-related technologies. Deliveries in the quarter rose just 0.1% to 103,295 vehicles, and vehicle margin came in at 12.1%, flat on the first quarter but below 14.3% a year earlier, which Xpeng blamed on model transitions. Services and other businesses lifted the overall gross margin to 20.7% from 17.3%.

Consolidating the product lines could help focus R&D resources, lower costs and reduce internal competition between models with overlapping positioning and pricing, the 36Kr report noted.

That overlap point deserves attention. Xpeng has kept launching cars while demand growth stalled, and the report frames the merger as a way to stop sibling models from competing for the same buyers. Consolidating four lines into two does not shrink the range. It changes who decides what gets built.

The lineup expansion continues regardless. Xpeng launched the G9L flagship five-seat SUV on 17 September, complementing the six-seat GX. The Mona series is growing too: deliveries of the Mona L03 began in July, and chairman and CEO He Xiaopeng said on an August earnings call that the Mona L05 SUV was scheduled to launch in China in the fourth quarter. Mr He also said on that call that Xpeng aims to exceed 60,000 single-month deliveries in the fourth quarter, a target roughly 53% above its August level.

Whether consolidation helps hit that number is an open question. Cutting R&D overhead is a slower lever than shipping volume, and the company has not published a cost figure attached to the reorganisation. CnEVPost reported the changes as internal, based on 36Kr's unnamed sources, so the details of who reports to whom remain unconfirmed by Xpeng itself.

The wider pressure on compute and hardware budgets

Xpeng's AI-related R&D line is where this story touches the rest of the technology business. Inference and GPU costs have been moving in the opposite direction to what buyers hoped. Recent headlines for context, not cited as facts here, include B300 GPU pricing above $24 per hour in short-term contracts and an 18.3% inference price rise from Nebius.

For anyone trying to plan that spending, a tool published on 23 September by developer Flavio Copes offers a browser-based inference cost calculator. It compares 27 models across OpenAI, Anthropic, Google, xAI, Mistral, OpenRouter and Workers AI, and turns assumptions about daily active users, calls per user, input and output tokens, and prompt-cache hit rate into a monthly bill. Its own documentation is explicit about the limits: the figures come from published API rates, they ignore batch pricing, enterprise discounts, image or tool surcharges and any caching layer a customer builds, and the output should be treated as a planning estimate rather than a vendor quote. The calculator runs entirely in the browser, and a share link encodes only the inputs in the URL.

Hardware pricing has its own extremes. Tom's Hardware reported on 27 September that a customer ordered a special edition of the roughly 5-inch (120 x 120 x 120mm) Kubb Fanless mini PC with a pure 24-carat gold passive chassis, at a cost of around $1.7 million. The gold build weighs 28.7 pounds (13kg), up from 4.6 pounds (2.1kg) in the standard aluminium version. Fanless Tech, quoted by Tom's Hardware, described the thermal conductivity of gold as about 30% to 50% faster than aluminium, and Tom's Hardware noted that copper would have been far cheaper and offers even better thermal conductivity. The machine inside is ordinary: Intel Core Ultra 5 225H, Core Ultra X7 358H or AMD Ryzen AI 5 340 options, with standard aluminium Kubb models listed at €1,595 ($1,822) at Kubb.eu but out of stock.

None of this makes Xpeng's restructuring cheap or easy. It does show the same tension running through the industry: money spent on models, chips and thermal engineering has to be justified by returns that are not arriving on schedule.

Solar costs as a reminder that scale cuts unit prices

Away from cars and GPUs, pv magazine reported on 26 September that average photovoltaic system prices in Brazil rose 7% in the first half of 2026 for projects up to 300 kW, citing Greener's "Distributed Energy Solutions" study. Final prices ranged from BRL 2.02 ($0.38) per watt for 30 kW and 50 kW installations to BRL 3.62/W for 2 kW systems. Kit prices for 4 kW systems jumped 18.3%, from BRL 1.42/W in January to BRL 1.68/W in June, while 300 kW kits rose just 2.0%.

The historical comparison is the interesting part. Greener's data shows the average price of a 4 kW residential system fell from BRL 7.74/W in January 2017 to BRL 2.91/W in June 2026. New distributed generation connections fell 16% in the first half, from 488,000 to 411,000, and the number of new consumer units receiving credits dropped 43%, from 951,000 to 541,000. Residential systems took 65% of added capacity, up from 39% in 2019, and only 33% of integrators' sales involved financing, down eight percentage points from 2025 and the lowest share in the period analysed.

The pattern is familiar to anyone watching AI infrastructure: unit costs fall over time, but demand shifts to smaller buyers, financing tightens, and the mix changes what an average price actually means. Xpeng's two remaining product lines will be judged on the same terms.