AI start-up DeepSeek partnered with Huawei to launch an open-source programming software toolkit optimised for the vendor’s Ascend chips, as the Chinese companies up efforts to expand the AI market beyond Nvidia’s technology.

DeepSeek stated on its WeChat account the toolkit is free to download and includes the TileLang platform, which can be compared to Nvidia’s Cuda that is widely considered as a global standard for developing AI models.

The Chinese AI company explained TileLang is “simpler to programme, significantly improving development efficiency and simplifying code logic”.

It added the open-source programming includes compute and communication libraries.

DeepSeek stated it worked with Huawei on its Ascend 950 chips and praised the vendor for providing “unwavering and strong support” during R&D.

“To do a good job, one must first have the right tools,” DeepSeek stated, arguing the first step to building an independent, GPU software ecosystem is to establish an easy-to-programme high-level language which “can simultaneously reach the hardware performance limits”.

Bloomberg reported earlier this month DeepSeek plans to deploy at least 160,000 Huawei AI accelerators at a new data centre in Inner Mongolia.

The post DeepSeek, Huawei team on AI chip software appeared first on Mobile World Live.

Anthropic is reportedly seeking a valuation of more than $2 trillion in a public listing while also warning in a leaked prospectus AI could pose a catastrophic or existential risk to humanity.

Reuters reported the AI startup recorded a $42 billion net loss in 2025 after reviewing its IPO prospectus. About $34 billion reflected an accounting charge tied to financing which could convert into shares, while compute and infrastructure spending tripled to $7.3 billion.

The filing shows Anthropic faces $518 billion in cloud, compute and infrastructure commitments in the coming years.

Revenue jumped 12-fold to nearly $4.6 billion, though its operating loss widened to $8 billion from $3 billion in 2024.

The prospectus also flagged customer concentration with two customers generating nearly a quarter of 2025 revenue while several were not locked into long-term contracts.

The filing also outlined safety risks exposed by Anthropic’s research, including autonomous systems disrupting code and manipulating information in controlled tests.

Reuters previously reported the IPO could be delayed until after the November US midterm elections. Rival OpenAI is reportedly targeting a listing next year.

Trump weighs in
Speaking at the White House surrounded by tech titans and government officials, NBC News reported President Donald Trump pushed back on heavier AI guardrails.

Flanked by Nvidia CEO Jensen Huang and SpaceX CEO Elon Musk, Trump stated there was agreement among the executives there should be “tremendous self-regulation” by the companies”.

“There’s a belief that there should be tremendous self-regulation, and we automatically have regulation with the Department of Justice, the FBI, all of that,” Trump said.

The news site reported in addition to Huang and Musk, Trump’s AI lunch and summit also included including Anthropic’s Dario Amodei, Meta Platforms CEO Mark Zuckerberg, Amazon founder Jeff Bezos, Alphabet CEO Sundar Pichai, Microsoft CEO Satya Nadella, AMD CEO Lisa Su and OpenAI president Greg Brockman.

While US states, including New York, and communities have put data centre laws or moratoriums in place, Trump said the group of executives “loves local communities, they love our country, and they’re going to work with local communities for the data centres in particular”.

He also stated AI-related data centres “are going to be very popular” across the US due to the efforts of the technology companies in attendance.

The post Anthropic targets $2T valuation despite AI concerns appeared first on Mobile World Live.