Realtime AI News
Musk confirms TSMC talks for chip production as Intel remains Terafab's only named partner
Musk has confirmed he is in discussions with TSMC about chip manufacturing for his Terafab project, adding a heavyweight potential partner while Intel remains the only officially named collaborator, set to supply 14A-class process technology. One reported scenario is a hybrid split, with Intel handling the advanced front end and TSMC covering fab operations, yield and packaging.
Musk's Terafab chip project may be about to add another heavyweight player. According to a report, Musk has confirmed he is discussing chip manufacturing collaboration with TSMC, while Intel remains the only partner officially named so far, planning to supply 14A, roughly 1.4nm-class, process technology to the project.
The report notes the talks are still at the discussion stage and may not lead anywhere. Even so, bringing TSMC to the table already gives Terafab's manufacturing path far more room to imagine.
The interest stems from the fact that Intel and TSMC are longtime rivals at the leading edge. The report sketches a possible hybrid: the advanced front end continues to be supplied by Intel using 14A, while the back end brings in TSMC to cover fab operations, yield and packaging.
The logic is not hard to follow. Leading-edge process is only one part of chipmaking; whether a product reaches volume often hinges on yield ramp, fab operations and advanced packaging, areas where TSMC has built deep experience. Intel is strong on the process roadmap, while TSMC is strong at turning a process into reliable volume production.
The context is what Terafab is meant to be: not a foundry taking outside orders, but a vertically integrated project to build custom chips for Tesla, SpaceX and related businesses, aimed at securing a steady, sufficient supply of in-house silicon and reducing reliance on outside manufacturing. That also explains why Musk has repeatedly stressed cutting dependence on outside chip supply.
For the industry, an Intel-plus-TSMC arrangement would mean two rival foundries potentially splitting work around one very large project, a notable shift in the story of US domestic chipmaking and AI compute supply.
What to watch next is whether the talks move from discussion to contract: whether Intel's 14A is finalized, what role TSMC takes, and whether such a hybrid can satisfy both capacity and yield requirements.
Why it matters
If Intel and TSMC end up splitting front-end and back-end work on the same mega-project, it would reshape a rivalry long defined by competition and give Musk's push for a self-built compute supply chain a more concrete footing.
Nearby Updates
All10/04, 13:47
Report: Anthropic's Olah Clashed With Vatican Over AI Consciousness Before Encyclical
According to Forkast, Anthropic's Olah clashed with the Vatican over the question of AI consciousness before the launch of a papal encyclical on artificial intelligence. The dispute shows how the debate over machine sentience is moving from research labs into religious and public institutions.
10/04, 16:25
Anthropic launches Frontier Academy with $100M plan to train 10,000 engineers
Anthropic has introduced Frontier Academy, a training initiative backed by $100 million aimed at educating roughly 10,000 engineers. The move extends the company's ambitions beyond model development into the engineering talent pipeline that AI deployment increasingly depends on.
10/04, 10:37
A malicious email could hijack the Manus AI agent, Salt Labs finds
Salt Labs, the research arm of Salt Security, disclosed that a malicious email could have hijacked the general-purpose Manus AI agent through indirect prompt injection and exposed credentials for connected services. The vulnerability has since been fixed, and the researchers stress that detecting a malicious instruction is not enough when an agent can execute it before anyone intervenes.
10/04, 10:00
Google Releases Gemini 4 Argon With 1M-Token Single-Output Limit
Google has released a new model called Gemini 4 Argon with a single-output limit of one million tokens, according to Chinese-language Google News aggregation citing a CSDN post. If accurate, the figure shifts competition from how much a model can read toward how much it can generate in one pass, reshaping long-document and code workflows.