Model

China clears DeepSeek and MiniMax to license AI abroad while $150 billion sits in the red

Six days after China proposed export controls on 23 AI technologies, state media says DeepSeek and MiniMax are the first labs cleared to license models overseas. The approval lands while the only two listed Chinese pure plays have erased $150 billion in value.

Vincent JiangVincent Jiangtwo-minute read
Share
The DeepSeek logo, the whale-shaped brand of the Chinese AI company whose export applications were approved
1 / 5Slide 1 of 5
DeepSeek, one of the two companies cleared this week to license its models abroad, alongside MiniMax

Six days after Beijing proposed export controls on 23 AI technologies, two names came through the gate: DeepSeek and MiniMax 14. State broadcaster CCTV's influential Yuyuantantian account confirmed on the night of October 22 that the two labs had completed registration and are cleared to open-source and license their models overseas, the first approvals under rules that took effect on October 16 14.

Who bleeds

The clearance lands on broken stocks. MiniMax Group and Z.AI, the only listed Chinese pure-play model developers, have shed a combined $150 billion in market value from their highs since their January debuts 3. MiniMax slid below its IPO price on Thursday, wiping out what had been a 650% gain at its March peak; Z.AI has tumbled 81% from its June high 3. New models did nothing: MiniMax's M3.1 Flash and Zhipu's GLM-5.3-FlashX got cold receptions 35.

The complication

JPMorgan strategists said this week they prefer Z.AI over MiniMax for its coding capability 3. The state's first pass went the other way: MiniMax, which already earns most of its revenue outside China, is cleared, while Z.AI has no announced registration 13. For the first time, a hard regulatory fact sits between the two tickers, and it points against the bank's pick.

The receipts

The rules are narrower than the panic suggested. MOFCOM says API access, websites, apps and already-public open-source releases do not count as regulated exports, and expedited "automatic licensing" applies to open-sourcing; only technology ownership transfers, overseas licensing deals and foreign ownership changes require registration first 14. Yuyuantantian framed the message plainly: most AI technology remains freely exportable 1.

The counterpoint

Approval is not profitability. Both listed labs remain unprofitable, global price wars with OpenAI$1.18T — OpenAI, private, latest valuation $1.18T and Anthropic$350B — Anthropic, private, latest valuation $350B squeeze margins, and JPMorgan−2.51% — JPMorgan, down 2.51 percent today does not expect Chinese AI models to break even before around 2028 35. January brings lockup expiries, about 178 million Z.AI shares and 25 million MiniMax shares, plus possible DeepSeek and Moonshot AI IPOs competing for the same investor cash 35.

What to watch

The gate is now an instrument of allocation. Every other Chinese lab's overseas licensing ambition waits on a state approval that can be granted or delayed, and the next registration filings will show whether Beijing is running a queue. Deep dive below for the draft rules and the filings behind each number 14.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading JPM

Reader comments

0 comments

    Model

    Meta's Q3 earnings fell to $3.83 against $5.46 expected, on a $9.89 billion teen-harm legal charge

    Excluding the charge Meta beat, and Businessweek's settlement pricing implies only about half the pending teen-harm claims are covered, with 2,000 school districts unpriced. The reserve note in January's filing decides.

    Model

    Nvidia shows how to run Z.AI's GLM-5.3 on $190,000 of Nvidia hardware, with no per-token bill to Z.AI

    The tutorial walks enterprises through serving GLM-5.3 on two DGX Stations, roughly $190,000 at listed reseller prices, with no per-token bill to Z.AI. For a stock carrying 118.6 times sales, that reads as a price list for everything Z.AI sells.

    Model

    Anthropic's Claude models tie at 58 on a coding index Cognition partly built, days before its Nasdaq IPO

    A third of the refreshed index runs on FrontierCode, a benchmark built by Cognition, whose Devin agent announced one of the two tied winners, Claude Fable 5.2, on the day of its release. Anthropic's model lists on Nasdaq on November 19; Cognition's Devin has offers from $70 billion on the table.

    Model

    Anthropic's Fable 5.2 passes 41.3% of real freelance jobs before its Nasdaq debut

    The one independent test of finished freelance work puts Anthropic's Fable 5.2 first at 41.3%, and the model still fails most of it, days before a Nasdaq debut priced at up to $1.4 trillion.

    Model

    Hedra cut Krea 3's edit time from 22.9 to 3.3 seconds on AMD, and the app doesn't own the speed

    The eight-second edit Krea advertises is manufactured in the serving stack, and a partner's measurement shows the app does not own the speed. Inference efficiency accrues to the optimizer and the chip vendor, which caps a per-image app's margin and hands AMD a demand channel beyond hyperscalers.

    Model

    xAI ships AI teammates at $30-$200 a month, ten days after Musk said serving them loses money

    The published card sets a chatbot price against an always-on machine the CEO says loses money on every subscriber; SpaceX shareholders fund both sides of that trade in one stock.

    Subscribe to TheInference for free