During Tencent Cloud’s AI Industry Application Conference on June 5, Tang Daosheng, Senior Executive VP and CEO of the Cloud & Smart Industries Group, alongside Tencent’s Chief AI Scientist Yao Shunyu, broke down the company’s latest moves and strategic thinking for the next chapter of the AI race.CEOTang Daosheng, Tencent’s ChiefAIScientist Yao Shunyu walked us through exactly where Tencent stands in the next phase of the AI race—and how they’re doubling down.
In their conversation, Tang noted that a huge chunk of Tencent’s code this year is now AI-generated. That means engineers are shifting their focus toward higher-level tasks like system architecture, basically handing off the actual coding to AI and stepping in to guide and polish what the models spit out.
Yao took it a step further, breaking today’s AI landscape into three core pillars: First, nail down the fundamentals by optimizing pre-training and post-training pipelines to build a rock-solid tech foundation. Second, turn those underlying technologies into real-world products that actually deliver value to society. And third, constantly hunt for new research paradigms and untapped opportunities. According to Yao, the absolute priority for AI’s next era is establishing a long-term, dedicated AGI organization right here in China.AGI
Tencent’s October 2025 R&D Big Data Report backs this up with some serious numbers. The report highlights how AI has seamlessly woven itself into every stage of the development lifecycle—from drafting and reviewing code to testing and deployment—transforming into the new efficiency engine driving Tencent’s entire R&D machine.20252025R&D Big Data Report2025AI
The stats are pretty wild: over 90% of Tencent’s engineers are already riding shotgun with the AI coding assistant CodeBuddy. Back in 2025, AI helped generate half of all new code written at the company—which essentially means for every two lines a programmer types, one of them was AI-assisted.90%CodeBuddy202550%——
But AI isn’t just tossing out code; it’s playing referee, too. It’s showing up in nearly 94% of code review sessions. What’s more, AI directly caught and validated 28% of the issues flagged, boosting the overall detection rate for actionable bugs by 44%.“审代码”94%28%44%
With tools like WeDev merging R&D workflows and smarter automation taking root, Tencent’s development automation jumped by 67%, wiping out roughly 5.3 million manual operations every month. By 2025, the team was shipping around 16,000 daily requirements—a 25% year-over-year leap—with average completion times shaved off by 12 hours. Code quality got a serious upgrade, too: they squashed over 5.4 million bugs and security holes all year round, cutting the average fix time by 8 hours per issue. On the infrastructure side, the pipeline churned out 6.3 million code builds weekly. Thanks to faster compilation, they’ve saved up a staggering 900,000 hours—that’s roughly 102 years of compute time reclaimed.WeDev67%53020251.625%12540BUG8BUG63090——102
When you zoom in on specific product lines, the efficiency gains are even more obvious. After integrating AI, backend compilation for WeChat dropped by 50%, WeChat Pay’s delivery cycle sped up by 31% while release quality climbed 14%, and iOS builds for QQ Mobile clocked in 40% faster. Tencent Games hit a 95% automation rate for art production, with auto-generated 2D UV mapping now matching human-grade quality. Over at Tencent Cloud, 65% of fresh code rolls directly out of CodeBuddy, slashing bug rates to 31.5% fewer per thousand lines. Even fintech deployments saw reliability skyrocket to a near-perfect 99%.50%31%14%iOS40%95%2UV65%AI Codebuddy31.5%99%
Word hit the streets on June 2 that WeChat is gearing up to launch its own AI agent, with internal plans to kick off compliance reviews before it goes public as early as this month. Per financial outlets, Tencent insiders admit the exact rollout date is still up in the air. It heavily hinges on how fast regulators clear the green light. Given WeChat’s massive user base of 1.4 billion, the compliance hurdle is likely going to be tighter than what other apps face.62AIAI14
Looking at the books, last year’s capital expenditure—which covered everything from IT infrastructure and data centers to land rights, campus expansions, and IP—clocked in at 79.2 billion RMB. Q4 alone accounted for 19.6 billion RMB, most of which was funneled straight into AI infrastructure. Tencent President Martin Lau pointed out that the company is pouring 18 billion RMB into new AI products throughout 2025.IT792196AI2025180
Speaking at the March earnings call, Lau admitted that GPU shortages and supply chain constraints kept 2025 capex below initial targets. But if hardware availability opens up this year, he made it clear they’re planning to ramp up spending significantly, with AI and model investments slated to at least double.32025GPUAI
Fast forward to Q1 2026, and Tencent’s capex skyrocketed past 31.9 billion RMB, up 16% year-over-year, essentially eating up 40% of the entire 2025 budget in just three months. R&D spend followed suit at 22.54 billion RMB, climbing 19%. For context, Alibaba dropped 26.9 billion RMB and Baidu went in at 5.9 billion RMB during the same quarter.2026几乎占到2025年全年资本开支的四成；研发投入225.4亿元，同比增长19%。作为对比，当季阿里集团资本支出269亿元，百度59亿元。
The financials tell a solid story: Q1 2026 revenue hit 196.46 billion RMB, up 9% YoY, while net profit attributable to shareholders soared to 58.09 billion RMB, a sharp 21% jump.20261964.589%580.9321%
Wall Street’s watching closely. UBS just dropped a forecast predicting Tencent’s full-year 2026 capex will smash past 170 billion RMB. The bank reaffirmed a “Buy” rating, noting that Tencent is currently playing hardball with AI investment—willing to eat short-term margin pressure to guarantee a full-scale upgrade of its cloud services and foundation models. UBS stresses that this massive R&D burn won’t choke the quarterly results. Classic cash cows like gaming and ads are proving incredibly resilient at monetizing, meaning the unexpected upside from AI is actually acting as a massive financial cushion, funding this aggressive tech sprint without breaking the bank.20261700“买入”AIAIAI
