Boxed Out: Unitree Goes Public in Shanghai the Same Week the U.S. Shuts Its Door

Capability & Integration

• OpenAI cut GPT-5.6 Luna’s API price 80% (input $1→$0.20, output $6→$1.20 per million tokens) and Terra’s 20%, leaving flagship Sol unchanged, just three weeks after the family’s July 9 launch — an unusually fast repricing OpenAI attributes to serving-cost efficiency gains, arriving as cheaper Chinese models pressure margins, per CNBC.
• Anthropic named Mariano-Florentino “Tino” Cuéllar — a former California Supreme Court justice and outgoing president of the Carnegie Endowment for International Peace — as its first Chief Global Affairs Officer on August 4, to lead engagement with heads of state and policymakers, per Anthropic.

Robotics

• Update: Unitree’s H1 Pro began Asian commercial deployment August 5, following its European launch July 22. Its planned August 12 North American launch has no path forward: the FCC’s July 28 Covered List expansion — flagged in yesterday’s brief — bars new Chinese humanoid models from the equipment authorization required to sell in the U.S., closing the launch for any new model while pre-ban inventory stays legal, per Tech Times.

  Unverified

  Reported unit counts for the Asia rollout (500+ across logistics and manufacturing sites) come from industry trackers, not Unitree itself.

• Unitree opens IPO subscription on Shanghai’s STAR Market August 10, seeking roughly $621 million at a $5.9 billion valuation — the first humanoid maker to list on a mainland China exchange, after clearing regulatory review in a record 104 days, per the South China Morning Post. It’s a public listing on a real order book, not a pilot — a sign Beijing is treating the U.S. ban as reason to deepen domestic capital markets, not negotiate.

  Review

  “$621 million at a $5.9 billion valuation” — these are pre-pricing estimates; Unitree priced the IPO on Aug. 6, the same day this brief is dated, at a reported ~$9.04B valuation raising ~$904M, per same-day CNBC/Bloomberg/Global Times coverage — suggested action: fix to final priced figures.

Hardware & Supply Chain

• Microsoft raised Xbox Series X pricing to $799.99 (+$150) and Series S to $499.99 (+$100) on August 1, explicitly citing console memory and storage costs that have more than doubled on AI data center demand, with Microsoft warning of another doubling by fall 2027, per Variety. It’s the clearest sign yet the DRAM shortage behind AI buildouts — last covered here July 31 — is reaching retail shelves.

• Sony raised the PS5 to $649.99 earlier this year, its second 2026 increase, and told the SEC it does not expect to resume discounting, citing the same memory-cost pressure, per Forbes.

  Review

  “its second 2026 increase” — imprecise: Sony’s first hike was August 2025, not earlier in 2026, so this is the second increase in roughly a year, not two hikes within calendar 2026 — suggested action: fix wording.

Environmental & Cultural Impact

• Amazon published its first comprehensive data center water disclosure: 2.5 billion gallons withdrawn globally in 2025, at a water-usage effectiveness of 0.12 liters per kilowatt-hour — down from 0.15 in 2024 and about seven times better than the 0.84 industry average Amazon cites — with the company saying it’s 75% toward a 2030 goal of replenishing more than it draws, per Data Center Dynamics.

  Review

  Amazon’s water-usage figures are the company’s own self-reported disclosure, single-sourced — suggested action: add [!unverified] flag.

• The EU AI Act’s Article 50 transparency rules took effect August 2, requiring visible, human-perceptible labeling of AI-generated content depicting real people plus machine-readable markers on AI outputs generally; penalties reach €15 million or 3% of global turnover, with a grace period for pre-existing systems running to December 2, per the European Commission.

  Review

  “a grace period for pre-existing systems running to December 2” — legal-analysis sources indicate the Dec. 2 grace period applies only to the machine-readable marking requirement, not the human-perceptible deepfake-labeling obligation (effective immediately from Aug. 2 with no grace period) — suggested action: clarify scope.

AI in the Wild

TikTok’s “history POV” trend — AI-generated first-person videos dropping viewers into the Black Death, a 1900s coal mine, or the 1986 Chernobyl disaster — has produced individual videos with 20 million-plus views apiece since accounts like @timetravellerpov popularized the format. It has drawn academic study and, in January, an open letter from German memorial sites objecting to AI content trivializing Holocaust-adjacent history; the format runs on regardless, per Fast Company.

Takeaway