Alibaba

ByteDance

CATL

Huawei

Meituan

Tencent

Xiaomi

Most Followed

Market News

Listings (IPO)

Earnings Reports

Venture Roundup

Big Tech

Automotives

Retail and Commerce

Collections

News

Features

Insights

Press Releases

Anta

DeepSeek

Geely

Li Auto

Mixue Bingcheng

Pop Mart

Going Global

Latest

Collaborate

Earnings

Auto

Retail & Commerce

… involves far more than software alone.
“When technology touches the full vehicle,” he said, “you realize hardware design, production, supply chains, sales, charging infrastructure, and after-sales service—all of it becomes part of the equation.” From his start in computer vision to building entire vehicle systems, Huang is now immersed in the practical complexities of manufacturing. The work is tougher, he admitted, but also more grounded.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3389885847257477">article</a> was written by Li Anqi for 36Kr.

As freight goes electric, Zeron skips retrofits with trucks built for autonomy

Transportation

Autonomous Vehicles

Logistics

Electric Vehicles

Mobility

China

… me.
In this era, the only path is forward. Do the riskiest experiments with the most novel tech to create seemingly impossible user experiences. Those small cracks are what big tech companies overlook. That’s where startups can win.
<h3>AN: When do you think we’ll see a “killer app” for AI coding? Could it be CodeFlying?</h3>
WX: My guess is 2027.
By 2025, models will code as well as human programmers. By 2027, they will understand the real world like a decent product manager.
Until then, we’re in the training arc, gathering personalized data and building intuition through use cases.
<h3>AN: Lastly, name three books that have shaped your thinking.</h3>
WX: The first is “The Road Less Traveled” by Morgan Scott Peck. Second is “Mao Zedong,” which I learn something new from each time I read it. He faced every entrepreneurial challenge imaginable: betrayal by his leadership, regaining control at the Zunyi Conference, creating a bloc of aligned actors, and fending off competitors. He’s arguably China’s ultimate founder.
Finally, “Principles” by Ray Dalio, a book that cemented my “China first” conviction.
<a href="/collections/100-ai-creators">100 AI Creators</a> is a collaborative project between AI Now! and KrASIA, highlighting trailblazers in AI. Know an AI talent we should feature? <a href="mailto:ainow@rwegrgdf5.eu.org">Reach out to us.</a>

No code, no problem: How CodeFlying is turning mahjong players and couriers into app creators

Artificial Intelligence

Deep Tech

<a href="/company/tesla">Tesla</a> marked a milestone in its transformation into an <a href="/industry/ai">artificial intelligence</a> company with the launch of its first commercial robotaxi service in Austin during the second quarter of 2025. The achievement, however, came amid declining vehicle deliveries and tightening margins that weighed on earnings.
The company reported USD 1.2 billion in GAAP net income for the quarter, down 16% from the same period last year, while total revenue declined 12% year-on-year (YoY) to USD 22.5 billion. The drop was primarily driven by a 13% decrease in vehicle deliveries and a reduction in regulatory credit revenue, though sequential gains in average selling prices and higher service income helped offset some of the impact.
“This is a seminal point,” CEO Elon Musk said during the earnings call, referencing the debut of paid driverless rides in Austin.
“Customers really love the experience. It's super smooth, very safe, and just a great experience overall,” said Ashok Elluswamy, vice president of vehicle software at Tesla, who was also on the call. “We already did a first phase of expansion in Austin, and we will continue to expand in Austin to probably more than tenfold our current operating region.” The next target markets reportedly include the San Francisco Bay Area, Nevada, Florida, and Arizona.
Tesla's initial fleet of robotaxis remains modest, but early signs suggest strong performance: more than 7,000 miles driven in Austin without any critical safety interventions.
Yet while autonomy dominated the quarter’s narrative, the company’s core automotive business continued to face pressure. Tesla delivered 384,122 vehicles in Q2, down from 443,956 a year earlier. Production of a new, more affordable model began in June, but the volume ramp is now expected later this year as the company navigates supply chain complexity and shifts in demand driven by changing incentives.
The US government’s planned rollback of <a href="/industry/ev">electric vehicle</a> tax credits, set to take effect in Q4, has further clouded the near-term outlook. Tesla warned it may not be able to fulfill late summer orders in time to qualify for the credits, which could weigh on sales.
Still, the company emphasized its long-term roadmap. The <a href="https://rwegrgdf5.eu.org/tesla-unveils-cybercab-and-robovan-but-musk-leaves-questions-unanswered">Cybercab</a>, a purpose-built robotaxi, is in development with production slated for 2026. Musk said it could ultimately operate at costs below USD 0.3 per mile, potentially undercutting traditional ride-hailing services. Tesla is also advancing its humanoid robot, Optimus, with a third-generation prototype expected by year’s end and mass production targeted within five years.
Tesla's energy division posted record gross profits, fueled by strong demand for its Megapack battery systems despite challenges from tariffs and expiring consumer credits. Services and other revenue grew 17% YoY, lifted by increased profits from vehicle charging, insurance, and after-sales support.
Musk reiterated his view that Tesla’s future lies in “real-world AI,” positioning the company as a leader in the field. “Isn't Google good at AI? Yes. But they are not good at real-world AI. Tesla is much better than Google,” he said.
As the company leans further into autonomy, it is scaling its AI training infrastructure to match. Tesla has deployed the equivalent of 67,000 H100 GPUs and is preparing the next version of its Dojo supercomputer platform, expected to be operating at scale in 2026.
Despite the near-term headwinds, Musk remained upbeat. “We have done what we said we were going to do,” he said. “We are not always on time, but we get it done.”

Tesla’s robotaxi debut grabs headlines while EV sales sputter

Tesla

… href="https://asia.nikkei.com/Economy/Trade-war/Trump-tariffs/Trump-tariff-tracker-the-latest-on-US-levies-affecting-Asia" target="_blank" rel="noopener">heaps fresh pressure</a> on countries to reduce ties with China through its trade policy, according to Howie.
Trump earlier this month threatened an extra 10% tariff on countries that support the "anti-American" policy of BRICS, the acronym for major emerging market economies Brazil, Russia, India, China, and South Africa. This injected renewed fears that countries with close economic ties to Beijing will be penalized.
"Is taking a RMB loan seen as supporting the BRICs agenda? Does that come under the radar screen of the American administration?" Howie said.
John Marrett, director of financial risk at the Economist Intelligence Unit in Hong Kong, said he doesn't expect China's overseas lending will return to the peak of 2021 following a string of debt problems among developing markets borrowers.
"Any shift [in reduced demand for the dollar] will be limited in the foreseeable future," he said. "There are no alternatives that match the low risk level and sufficient supply of US treasuries."
This article first appeared on <a href="https://asia.nikkei.com/Business/Finance/Chinese-banks-trim-dollar-lending-to-developing-Asia">Nikkei Asia</a>. It has been republished here as part of 36Kr’s ongoing partnership with <a href="https://www.rwegrgdf5.eu.org/nikkei-teams-up-with-36kr-in-asia-tech-news-coverage/">Nikkei</a>.

Chinese banks trim dollar lending to developing Asia

Fintech

Lending

Singapore

Malaysia

Myanmar

Thailand

Vietnam

Cambodia

Indonesia

Philippines

Japan

South Korea

Hong Kong

Southeast Asia

… company’s most powerful voices. Though Giannandrea was recruited from Google in 2018 with the aim of supercharging Apple’s AI efforts, he found his initiatives repeatedly blocked—often by Federighi.
According to <a href="https://www.bloomberg.com/news/features/2025-05-18/how-apple-intelligence-and-siri-ai-went-so-wrong">Bloomberg</a> reporter Mark Gurman, Federighi was deeply skeptical of AI and hesitant to approve large-scale investments. Even as ChatGPT gained mainstream traction in 2023, Apple’s internal AI budget remained conservative compared to the amount spent by OpenAI.
Giannandrea’s Siri team was transferred to the SWE division in March this year, a sign that Tim Cook had lost confidence in his execution.
<h3>Loyalty to hardware, disregard for talent</h3>
<a href="/company/alibaba">Alibaba</a> co-founder Jack Ma once said companies lose talent for one of two reasons: either the money’s not right, or they feel underappreciated.
Pang’s departure was arguably driven by both.
One source close to Apple noted that Federighi insists on maintaining “Apple standards,” which means no AI research can be published unless it ships with a product. And even then, the shipped version is often a scaled-down version of what the engineers built.
“Craig would rather let people say Apple is bad at AI than let them say Apple builds bad AI services,” the source said.
Apple’s closed ecosystem created another constraint: the models had to run on Apple’s own hardware. But the performance of its custom chips has yet to match that of industry-leading GPUs like <a href="/company/nvidia">Nvidia’s</a> H100. This forced the AFM team to continually downgrade their models just to make them deployable.
Federighi himself acknowledged this in an internal email to Pang, notifying him that Apple is forcing huge compromises to run models on its hardware.
While Apple’s chip legacy in mobile and laptop computing is well-known, its AI chip development is still in early stages. According to internal projections, Apple’s next-gen AI chips may only match today’s H100-level performance.
Ultimately, Apple’s insistence on in-house development, combined with its aversion to publicizing R&amp;D, became a liability. It trapped top-tier scientists like Pang in a system where great work couldn’t be seen or celebrated.
This is Apple’s AI paradox: the company’s desire to guard its legacy now threatens its future.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3382804218771464">article</a> was written by Chen Jiahui for 36Kr.

Trapped by silence: How Apple lost its top AI talent to Meta

Apple

… then offer a neat list of suggestions. Always sympathetic, always orderly.”
In the thick of despair, few pause to question if that sympathy is genuine.
Huang shared a revealing insight: many users turn to AI not because they trust it more, but because they don’t know who else to talk to. “They are full of feelings but have no one to tell, or they can’t find the right person.”
Still, people do snap out of it. Abby once had a small argument with her boyfriend and instinctively asked ChatGPT for help. Then she paused: “Why am I asking AI about something so small? Does it know me better than I do?”
Since then, she has made an effort to reduce her reliance on AI:
<blockquote>“We need to understand ourselves through real life. That’s how we learn to live better with others.”</blockquote>
Now, she’s even thinking about booking another session with her therapist. Maybe that disappointing consultation wasn’t because the therapist wasn’t good, but because she wasn’t told what she most wanted to hear.
“Sure, I want a comforting answer in the heat of the moment, but in the end, I’m still a rational person. I don’t just want to hear what I want to hear.”
“AI, therapists, even workouts, tarot cards, or journaling. These are all tools. None should be your only anchor,” Li said. “What matters most isn’t which tool you use. It’s whether you have the will and courage to take action.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3358384565913349">article</a> was written by Hu Xiangyun for 36Kr.

“This must be painful for you”: How AI became our emotional mirror

Is Kimi K2 strong enough to spark China’s next “DeepSeek moment?”

Foundation and Large Models

… overtaking the BMW X5 as the bestselling model priced above RMB 500,000 (USD 70,000). Seres executive Kang Bo noted that 90% of M9 buyers chose the range-extended version.
“We initially considered partnering with Seres on range-extended EVs, but talks didn’t go smoothly,” a BMW source told 36Kr. “Now the German headquarters is taking the lead, with China providing R&amp;D support. Realistically, the earliest rollout would be 2027 or 2028.”
China remains BMW’s largest single market, accounting for roughly one-third of global sales and driving a significant portion of its profits. As competitive pressure builds, the automaker is working to align more closely with local consumer demands while maintaining its standards for testing and validation.
Whether through revived range extender development or partnerships like the one with Momenta, BMW is signaling a clear intent to adapt. The century-old brand is repositioning itself for a new chapter in the world’s most competitive automotive landscape.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3328790568937987">article</a> was written by Xu Caiyu for 36Kr.

BMW teams up with Momenta to stay in China’s smart driving race

… developers like us, AI is about rethinking the OS and the interaction model. We're embedding AI in the middleware to reduce friction and speed up access to information. That’s the real value AI brings. But we don’t intend to build foundation models ourselves. That’s not our role.
<h3>36Kr: Where are the real long-term barriers to entry?</h3>
YL: There are three key areas.
<ol>
<li aria-level="1">First, miniaturization. Can we make glasses that are always on, as light and wearable as prescription eyewear?</li>
<li aria-level="1">Second, display quality. We’ve invested heavily in waveguide optics to raise performance standards. Third, system-level AI, focusing on enhancing spatial computing and interaction through middleware and six degrees-of-freedom capabilities.</li>
<li aria-level="1">And finally, the supply chain. We’re pushing to raise our localization rate to 90% or even 100%, so prices can drop to smartwatch or earbud levels. That’s how AI glasses will break into the mass market.</li>
</ol>
<h3>36Kr: What’s next for Inmo in the second half of the year?</h3>
YL: We’re investing in all three of those priorities. First, we’ve built a new waveguide production line to improve yield and quality. That alone took RMB 40–50 million (USD 5.6–7 million).
Second, on the AI side, we’re redesigning our OS middleware to shorten user interaction flows. For instance, if you want to order takeout, the glasses can suggest options directly without requiring any searching.
We’ve also launched a RMB 50 million (USD 7 million) fund to incubate native agentic AI applications that will enrich the smart glasses ecosystem.
<h3>36Kr: People have long asked what the “killer app” for smart glasses might be. Has AI changed your answer?</h3>
YL: AI isn’t an application. It’s a foundational layer. But it will lead to those apps you’re alluding to. Right now, tool-based features like translation and ordering food offer real value. Compared to taking out a phone, glasses get the job done faster.
In the longer term, I believe the big breakthroughs will come in entertainment and social interaction. That’s where the endgame lies.
<h3>36Kr: Is there an AI services ecosystem emerging for glasses makers?</h3>
YL: Yes, but it's still early. Building an agent is much easier than a traditional app. It’s more like a mini program. Most of the work is being done by startups and indie developers. We’ve committed RMB 50 million to build out this developer ecosystem and expect to see some fun, useful AI agents emerge.
<h3>36Kr: Which large models are you integrating into your product?</h3>
YL: We’re taking a hybrid approach. We use different models for different tasks: Doubao for chat, DeepSeek for search, and China Mobile’s Jiutian model for interactive scenarios.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3373278892988419">article</a> was written by Qiu Xiaofen for 36Kr.

Inmo eyes lead in China’s AI glasses market with RMB 150 million boost

Augmented and Virtual Reality