Tencent exec admits AI lag, says computing power is ‘severely insufficient’

Tencent Senior Executive Vice President Tang Daosheng publicly acknowledged on Tuesday that the Chinese tech giant has fallen behind in AI, citing "severely insufficient" computing power that has hampered model training and product development. The candid admission, published in the company's internal publication, marks a notable shift for a company long known for its caution.reuters+1

Tang Daosheng framed the AI race as a marathon, arguing that "we may have only completed the first kilometer" and that endurance matters more than an early start. But he did not shy away from conceding that competitors have held an advantage in AI visibility and momentum over the past few years.tencent+1

A Spending Spree to Close the Gap

The rhetoric is backed by hard numbers. Tencent's capital expenditure surged to RMB 52.8 billion in the second quarter of 2026, up 176% year-over-year, after RMB 31.9 billion in Q1. The first-half total of RMB 84.7 billion already exceeds the company's full-year 2025 capital expenditure of approximately RMB 79 billion. Reuters reported that last year's spending fell short of internal targets partly due to supply constraints on high-end chips.tencent+1

The investment push has come at a cost. Tencent reported negative free cash flow of RMB 13.8 billion for the second quarter, and disclosed that excluding revenue, costs, and expenses related to new AI products — including Yuanbao, CodeBuddy, and WorkBuddy — its operating profit would have been roughly RMB 10.5 billion higher.kucoin+1

Application Scenarios as the Path Forward

Rather than competing purely on model benchmarks, Tang Daosheng pointed to Tencent's ecosystem — 1.439 billion monthly active WeChat accounts, Enterprise WeChat, Tencent Meeting, Tencent Docs, and cloud services — as its core advantage. "Application scenarios are Tencent's strongest asset in AI," he wrote, arguing that algorithms determine the ceiling but engineering determines how quickly a company reaches it.news.futunn

Yet the consumer AI gap is real. According to QuestMobile data from June 2026, ByteDance's Doubao led with 382 million monthly active users, followed by Alibaba's ↘ 4.10% Qwen at 167 million and DeepSeek at 130 million. Tencent's Yuanbao trailed at roughly 50 million.kucoin

The Road Ahead

Tang Daosheng highlighted WorkBuddy and CodeBuddy — AI-powered productivity tools that underwent more than 40 iterations within three months of launch — as evidence that Tencent can translate its infrastructure into working products. Whether the next-generation Hy4 model can narrow the gap, and whether billions in computing power can convert into revenue, remain the open questions that will determine whether Tencent's expensive catch-up phase delivers results.news.futunn