AI Markets: Unitree soars in Shanghai, China takes on SpaceX in orbit

A double win for Chinese tech: humanoid robot maker Unitree jumped 629% in its Shanghai stock market debut, while LandSpace landed a reusable rocket and joined SpaceX’s very exclusive club. Elsewhere, rising long-term US yields rattled AI stocks on the Nasdaq, and 13F season revealed diametrically opposed bets among the world’s biggest money managers.

Unitree jumps 629% in Shanghai: Chinese humanoid robotics lands its first listed unicorn

Chinese humanoid and quadruped robot maker Unitree made a blockbuster entrance on the Shanghai Stock Exchange this Wednesday, August 19: its shares opened at 1,100 yuan, against an IPO price of 150.80 yuan, a 629% jump. That’s the biggest first-day gain for a listing in China since the start of the year.

Key points:

  • Valuation at the open: about 445 billion yuan, or nearly $66 billion.
  • Unitree raised 6.1 billion yuan ($905 million), above its initial target of 4.2 billion.
  • The retail tranche was oversubscribed more than 8,000 times, a record for a tech company on Shanghai’s STAR Market.
  • DeepSeek is among the IPO’s strategic investors, alongside Tencent and China’s social security fund.

A bet on robotics that’s “already profitable”

Unlike many robotics startups still losing money, Unitree bills itself as the world’s top humanoid robot maker by sales volume and says its business is already profitable. The company becomes the first humanoid or quadruped robotics company listed on a major global exchange: neither Boston Dynamics (a SoftBank subsidiary) nor America’s Figure AI, both still private, has crossed that line. The market is therefore valuing Unitree at roughly 219 times earnings, against a sector average of around 39 times, a multiple that reflects both enthusiasm for “physical AI” robotics and the scarcity of a listed pure play to sink your teeth into.

What to watch next

This explosive start says nothing about the medium-term trajectory: Unitree’s revenue growth already slowed sharply in the first half of 2026 (+35% to 45%, versus +332% a year earlier), and adjusted net income is expected to decline as R&D spending climbs steeply. The real question now is how the stock holds up once the euphoria fades, especially when institutional investors are able to sell their shares. For a European investor, Unitree remains an A-share listed in Shanghai, so it’s not eligible for a PEA and out of reach of a standard brokerage account without a specialized broker; indirect exposure goes through UCITS ETFs dedicated to robotics (such as KOID), which generally carry a limited Chinese weighting.

Sources:

China: LandSpace lands a reusable rocket and joins SpaceX’s exclusive club

On the same day, August 19, Chinese startup LandSpace successfully landed the first stage of its Zhuque-3 rocket after a second orbital flight, becoming the first private Chinese company — and only the third in the world after SpaceX and Blue Origin — to recover an orbital rocket stage of this size.

Key points:

  • The stainless steel Zhuque-3 rocket (66 meters tall) lifted off from the Jiuquan space center, and its first stage touched down vertically on its deployable legs.
  • A first attempt, in December 2025, had failed: the stage reached orbit but crashed in flames during the landing attempt.
  • Zhuque-3 borrows the architecture of SpaceX’s Falcon 9 (reusable stage, nine engines) but runs on methane and liquid oxygen rather than kerosene.
  • LandSpace itself is in the middle of an IPO process on Shanghai’s STAR Market — the same market where Unitree debuted this Wednesday.

A success that narrows the gap with SpaceX

After December’s spectacular failure, which industry experts called “90% successful,” this second attempt validates LandSpace’s technical mastery of guided landing — the hardest hurdle to clear in rocket reuse. SpaceX’s Falcon 9 remains the global benchmark with hundreds of successful flights, but LandSpace proves that China can replicate, a few years behind, the technology that made SpaceX’s fortune on the markets. Another Chinese player, state giant CASC, had already pulled off a similar landing, but LandSpace is the country’s first private company to do it.

The direct link to the markets: an IPO underway, but profitability far off

LandSpace updated its IPO filing in late June, aiming to raise about 7.5 billion yuan ($1.1 billion) on the STAR Market, riding a wave of about ten Chinese space startups lining up to go public. But the financials remain modest for now: 2025 revenue of just 52 million yuan ($7.7 million) against cumulative losses exceeding 4.8 billion yuan, with a profitability target pushed back to 2029. Like Unitree, LandSpace would, once listed, remain a Chinese stock outside the PEA for European investors, accessible only through a brokerage account or UCITS ETFs exposed to the broader space sector.

Sources:

Rising long-term yields sink Nasdaq AI stocks, Apple holds up

Wall Street took another hit on August 18: the 30-year Treasury yield climbed to around 5.34%, its highest level since June 2007, triggering a broad selloff in tech stocks and, in particular, semiconductors.

Key points:

  • The Nasdaq Composite fell 1.33% to 26,289.71, while the SOX semiconductor index dropped 5.5%.
  • The 10-year yield also rose, to around 4.70%.
  • CoreWeave, one of the symbols of debt-financed AI infrastructure, plunged 12.1% in the session.
  • Causes cited by the markets: worries about US government debt, inflation fears tied to tensions around Iran and rising oil prices, and the scale of borrowing linked to the AI data center race.

Why higher rates hurt AI stocks in particular

The mechanism is fairly straightforward: AI giants (hyperscalers, data center operators like CoreWeave or Nebius) fund part of their massive investments with debt, and higher rates make that financing more expensive while lowering the present value of the future profits they’ve promised investors. That’s exactly the debate the sector has been having for weeks, between believers in a compute demand “super-cycle” and those who fear a bubble fueled by circular financing. CoreWeave’s drop captures that tension: the company just raised its 2026 capex budget to $35 billion to $39 billion, largely financed by debt whose interest already weighs heavily on its results.

Apple, the unexpected safe haven of US tech

In the same session, Apple stood out by holding up better than the rest of the “Magnificent Seven,” extending a trend seen since late July: investors worried about the scale of AI capex are retreating into companies whose revenue depends less on the data center race. It’s a signal worth watching for a European investor exposed to US tech through a US equity or tech UCITS ETF (outside the PEA for most of these stocks, except for the rare European companies in the sector such as ASML): rising long-term US yields don’t hit all AI stocks the same way, and the gap between winners and losers is widening.

Sources:

13F season: hedge funds split over AI stocks

The 13F regulatory filings, which reveal the positions of major US funds as of June 30, 2026, show this week a marked rotation out of the “legacy” chipmakers toward the big buyers of AI compute — and diametrically opposed bets on some star stocks.

Key points:

  • Stanley Druckenmiller (Duquesne Family Office) fully exited five chip and photonics stocks — Micron, Intel, Broadcom, Lattice Semiconductor and Coherent — in the second quarter.
  • In exchange, he more than tenfolded his Amazon position, opened stakes in Alphabet, TSMC and Lam Research, and took positions in data center operators/former crypto miners repurposed for AI compute (IREN, Bitdeer, Riot Platforms, Hut 8).
  • Several heavyweights abandoned Broadcom in Q2: Third Point and Duquesne exited completely, as did D1 Capital, while Tiger Global cut its position in half.
  • On CoreWeave, David Tepper (Appaloosa) opened a new $107 million position, while George Soros held, as of the same date, put options with a notional value of about $65 million on the same stock.

A rotation “from the shovels to the gold diggers”

The dominant move this quarter is clear: several major managers are cutting their exposure to the chip suppliers that have benefited most from the AI gold rush over the past two years, repositioning toward the companies that buy and run that compute — hyperscalers like Amazon and Alphabet, foundry TSMC, equipment maker Lam Research, or data center operators. One possible reading is that these managers see the “easiest” part of the chip bet as already well priced in, and are looking for a new source of returns in the infrastructure that puts those chips to work.

CoreWeave, battleground between bulls and skeptics

The CoreWeave case neatly illustrates the split: Tepper sees it as a growth bet on AI infrastructure, while Soros is hedging with puts — a position that, according to analysts, looks more like a hedge than a pure bearish bet, since 13Fs reveal neither the strike price nor the premium paid. That bet took an ironic turn: CoreWeave stock fell 12.1% on August 18, a few days after these positions were published. For European investors, the key is to remember that 13F filings are a dated snapshot (positions as of June 30, published in mid-August): they say nothing about decisions made since, or about the real size of options bets, and are in no way a recommendation to follow.

Sources:

Edition prepared by our monitoring agent (data and sources as of Wednesday, August 19, 2026), reviewed under our editorial policy. General information, not investment advice: investing carries a risk of capital loss.