Huawei Ascend chief scientist Liao Heng on sanctions, Silicon Valley and self-reliance

Share:
Audio Loading voice…
Huawei Ascend chief scientist Liao Heng on sanctions, Silicon Valley and self-reliance

Synopsis

Huawei's Ascend AI chip architect Liao Heng — a child prodigy who joined HiSilicon sceptical of China's chip ambitions in 2016 — says US sanctions became a forcing function, not a ceiling, and that HiSilicon's flawless tape-out record converted him into a believer.

Key Takeaways

Liao Heng , chief scientist of Huawei's semiconductor business, is the chief architect behind the Ascend AI processor family.
Liao enrolled at Tsinghua University in 1987 at age 14 and later conducted postdoctoral research at Princeton University .
He joined HiSilicon in 2016 as a sceptic but was converted after witnessing the unit tape out dozens of chips annually with zero mass-production failures.
Liao warned that extended stints working abroad can breed 'blind arrogance' — the false belief that only Western firms are capable of true innovation.
The Ascend programme is positioned as China's primary alternative to Nvidia GPUs and the CUDA ecosystem amid escalating US export controls.
Liao cited the slowing of Moore's Law as an industry-wide opening for architectural differentiation beyond pure process-node competition.

Liao Heng, chief scientist of Huawei Technologies' semiconductor business and the principal architect of its Ascend artificial intelligence processors, has spoken candidly about overcoming a 'blind arrogance' he observed among engineers trained exclusively in the West — and how that realisation shaped his strategy for building China's most consequential AI chip programme. The remarks, made on a business podcast that has since gone viral, offer a rare window into the mindset driving Huawei's push for technological self-reliance under sustained US export controls.

From Princeton to HiSilicon: an unlikely convert

Liao is no ordinary chip engineer. A former child prodigy, he enrolled at the prestigious Tsinghua University in Beijing in 1987 at the age of 14, before pursuing postdoctoral research at Princeton University. He subsequently spent more than a decade in the United States, watching what he called a 'long sunset' for American chip start-ups. When he joined Huawei's chip-design unit, HiSilicon, in 2016, he was openly sceptical that China could ever produce world-class semiconductors.

That scepticism evaporated quickly. 'Reality taught me a quick lesson,' Liao said in the podcast, hosted last month by a reporter from a mainland news organisation. He pointed to HiSilicon's extraordinary execution record — taping out dozens of new chip designs every year without a single failure to reach mass production. 'Such a high hit rate shows that every team [at the company] is highly qualified,' he said.

Why it matters: Ascend at the centre of China's AI stack

As the chief architect behind the Ascend AI processor family, Liao sits at the epicentre of China's effort to build a domestically controlled AI computing infrastructure — one that does not depend on Nvidia hardware or the CUDA software ecosystem. With successive rounds of US chip export restrictions cutting off access to leading-edge Nvidia GPUs, Ascend chips have become the de facto alternative for Chinese cloud providers, research institutions, and AI developers, including those building on models comparable to DeepSeek.

The strategic importance is amplified by Huawei's simultaneous effort to develop its own chip-manufacturing pathway, reducing dependence on Taiwan Semiconductor Manufacturing Company (TSMC) and navigating the constraints imposed by the absence of cutting-edge EUV lithography tools.

The competitive backdrop: sanctions as a forcing function

Liao revealed in the interview how he drew on his Silicon Valley experience to import best practices while identifying where domestic strengths could be leveraged differently. He described the period of working abroad as one that inadvertently bred a form of intellectual insularity — an assumption that innovation was the exclusive preserve of Western firms. Witnessing HiSilicon's tape-out discipline firsthand dismantled that assumption.

His framing echoes a broader pattern across China's semiconductor sector, where US sanctions — far from crippling development — have accelerated investment in indigenous design capability, toolchains, and talent pipelines. Liao adopted what he described as a war-hero mindset, invoking the ethos of sacrifice and mission focus associated with figures such as Dong Cunrui, a celebrated Chinese military martyr, to galvanise his engineering teams.

What's next: the limits of Moore's Law and the road ahead

Beyond the immediate competitive battle with Nvidia, Liao acknowledged the industry-wide challenge posed by the slowing of Moore's Law — a constraint that affects every chipmaker globally and opens space for architectural innovation rather than pure process-node competition. Huawei's ability to close the gap with leading-edge AI accelerators will depend not only on design ingenuity but on progress in domestic fabrication, advanced packaging, and software ecosystem maturity.

As China's AI industry scales and the geopolitical pressure on semiconductor supply chains intensifies, the trajectory of the Ascend programme — and the strategy of engineers like Liao Heng — will be among the most closely watched developments in global technology.

Point of View

Intended to contain China's AI ambitions, have instead concentrated talent, capital, and political will into domestic programmes like Ascend — producing a forcing function that Western policymakers did not fully anticipate. What mainstream coverage often misses is the organisational discipline angle: HiSilicon's near-perfect tape-out record is not a lucky streak but a systems-engineering culture that rivals the best fabless houses globally. The deeper risk for Nvidia and the broader Western AI supply chain is not just hardware substitution — it is the gradual maturation of an alternative software ecosystem that could eventually reduce CUDA's lock-in within the Chinese market. As DeepSeek-class models proliferate on domestic infrastructure, the question is no longer whether Ascend is 'good enough' but whether the gap is closing faster than export-control regimes can widen it.
NationPress
4 Aug 2026

Frequently Asked Questions

Who is Liao Heng and what is his role at Huawei?
Liao Heng is the chief scientist of Huawei Technologies' semiconductor business and the chief architect of its Ascend AI processor line. A former child prodigy, he enrolled at Tsinghua University at age 14 in 1987 and later did postdoctoral work at Princeton University before joining HiSilicon in 2016 .
What did Liao Heng say about Western chip innovation?
He said that working abroad often bred a 'blind arrogance' — a belief that only Western firms were capable of genuine innovation. He credited first-hand exposure to HiSilicon's execution discipline with rapidly changing that view, saying 'Reality taught me a quick lesson.'
How does Huawei's Ascend chip compete with Nvidia?
Huawei's Ascend processors are positioned as China's primary domestic alternative to Nvidia GPUs and the CUDA software ecosystem, particularly after successive rounds of US export controls restricted Chinese access to leading-edge Nvidia hardware. The programme aims to support a self-sustaining AI computing infrastructure within China .
What is the significance of HiSilicon's tape-out record?
HiSilicon reportedly tapes out dozens of new chip designs every year without a single failure to reach mass production — a hit rate Liao described as evidence that 'every team [at the company] is highly qualified.' This level of execution discipline is rare even among the world's top fabless chip designers.
How have US sanctions affected Huawei's chip development?
US semiconductor export restrictions have cut Huawei off from leading-edge manufacturing nodes at Taiwan Semiconductor Manufacturing Company (TSMC) and from advanced Nvidia AI accelerators. According to Liao's account, the sanctions have functioned as a forcing mechanism, accelerating investment in indigenous design capability, domestic fabrication pathways, and alternative software toolchains rather than halting progress.
Nation Press
The Trail

Connected Dots

Tracing the thread behind this story — newest first.

8 Dots
  1. Latest 3 days ago
  2. 1 week ago
  3. 1 week ago
  4. 1 month ago
  5. 1 month ago
  6. 2 months ago
  7. 2 months ago
  8. 2 months ago
Google Prefer NP
On Google