Thesis · Semiconductors & Compute Silicon
Will US-headquartered frontier model providers (OpenAI, Anthropic, Google, Meta) still be unavailable for direct commercial service to mainland China customers as of 2027-09-30?
Regulatory and sanctions barriers keep Western frontier models out of China while Chinese models (Doubao, Qwen, DeepSeek, Kimi) hold the domestic market, locking in two separate AI product markets.
Who is on each side
The two sides.
Each side in DeepStack’s words, then the voices who argue it, attributed and paraphrased, dated where the date is known. A voice links to its page on DeepStack where one exists.
The case for yes
US and China split into two AI markets
Chinese local-hosting, licensing and content-compliance rules combined with US export and service restrictions keep US frontier labs out of direct commercial service in mainland China through the window. The result is parallel, non-overlapping model ecosystems with separate developer bases, pricing and infrastructure, mirroring how sanctions pushed Russia toward Yandex's Alice and Sber's GigaChat while China relies on Doubao, Qwen, DeepSeek and Kimi. Mutual unease over military AI applications and control of semiconductors and data flows gives both sides reason to keep the ecosystems separate rather than merge them.
The case for no
A US frontier lab wins approval to serve mainland China
No means a US-headquartered frontier model provider obtains the regulatory clearance needed from Chinese authorities and launches a direct commercial service for mainland China customers before the deadline. That single approval and launch would show the barrier was a removable licensing hurdle rather than a durable split, ending the two-market setup the yes side describes. No voice currently on record argues this outcome directly; the side rests on the observable event itself.
2 voices
US and China split into two AI markets
Olivia Moore
Western sanctions restricted access to US AI tools in Russia since 2022, and China requires local hosting and censorship compliance. These policy barriers created gaps that local products like Yandex's Alice, Sber's GigaChat, and China's Doubao and…
Nations seeking to gain an upper hand in military AI technology invest heavily to control critical inputs like semiconductors and data flows. This competition creates mutual unease and incentivizes each nation to reduce reliance on others for…
1 voice
A US frontier lab wins approval to serve mainland China
Steel Team
Once the manufacturing knowledge behind N+2 and N+3 is transferred to HLMC/Hua Hong and the same learning feeds accelerator programs, no single entity remains the bottleneck, so an entity-list approach aimed at SMIC alone stops controlling the…
What would settle it
The dated tests.
Each test names the public source that settles it, the rule and the day. The Call’s number, the readers’ median and the machines’ numbers stand beside it, as on the Docket.
TSMC, monthly revenue growth from a year earlier, October 2026: whether it comes in at or above 50%.
On the Docket Pendingth048-t04Open question
Specification pending
The machinesClaude 22% · DeepSeek 25%
Your numberSign in by email to forecast
TSMC, monthly revenue growth from a year earlier, October 2026: whether it comes in at or above 54%.
On the Docket Pendingth048-t05Open question
Specification pending
The machinesClaude 22% · DeepSeek 12%
Your numberSign in by email to forecast
TSMC, monthly revenue growth from a year earlier, October 2026: whether it comes in at or above 58%.
On the Docket Pendingth048-t06Open question
Specification pending
The machinesClaude 15% · DeepSeek 5%
Your numberSign in by email to forecast
NVIDIA, revenue growth from a year earlier, quarter ended October 25, 2026: whether it comes in at or above 86%.
On the Docket Pendingth048-t01Open question
Specification pending
The machinesClaude 11% · DeepSeek 5%
Your numberSign in by email to forecast
NVIDIA, revenue growth from a year earlier, quarter ended October 25, 2026: whether it comes in at or above 91%.
On the Docket Pendingth048-t02Open question
Specification pending
The machinesClaude 7% · DeepSeek 8%
Your numberSign in by email to forecast
NVIDIA, revenue growth from a year earlier, quarter ended October 25, 2026: whether it comes in at or above 96%.
On the Docket Pendingth048-t03Open question
Specification pending
The machinesClaude 5% · DeepSeek 8%
Your numberSign in by email to forecast
In The Buildout
The Parts on the same question.
The stories that take up this thesis, with their own Calls and tests.
Part 15: Oil Is the Headline. The Real Yield Is the Bill.
The barrel gave a Fed already facing hot core prices one more reason to hike, and the borrowing of the Treasury and the hyperscalers helped lift the real yield that long-dated earnings answer to. DeepStack reads the three sides, splits the 10-year on public data and names the tests, starting with the Fed's decision of October 28.
A thesis page states one question and both sides of it; DeepStack takes no side in words, and its view is the number it gives each test before the date. Analysis and opinion, not investment advice. Voices are paraphrased from public writing and attributed by name; DeepStack shows no score, ranking or accuracy for any person. Pre-specified means the words were approved on or before the test’s date. Specified after the date means they were written afterwards: such a test is counted and shown, and never rated. An open question has no words yet. The Call’s number is the chance, from 0 to 100, that the Call gives the yes answer to a test’s question, recorded before the test’s date; it is DeepStack’s own reading, not a market price and not advice. The readers’ mark on the same line is the readers’ median, the middle of the numbers signed-in readers gave the test, shown from 25 readers. The machines: Each model read this page and nothing else. The number that counts is the latest one each model gave before the test closed.