Tencent's (腾讯) Lobster Cannot Fill the Gap in the Computing Power Center

Big launch, bigger mismatch

Tencent (腾讯) pushed three OpenClaw-related moves in a single day — WorkBuddy formally launched, QClaw was reportedly exposed, and an official Enterprise WeChat integration guide went live — just as Shenzhen rolled out its “Lobster Ten Measures” (龙虾十条) of subsidies and compute support. It has been reported that the timing was not accidental. But can a vendor-side product push really solve a structural utilization problem that local governments have been trying to fix with cash and coupons?

Vouchers, idle racks and stubborn economics

Local governments have poured money into compute vouchers — Shenzhen’s “training-power” vouchers can cover up to 60% and its annual pool is reportedly RMB 500 million; Hangzhou’s compute vouchers total about RMB 250 million a year. Yet public data shows weak results: China Academy of Information and Communications Technology (中国信通院) finds average smart-compute center utilization around 32%, while IDC data suggests enterprise compute centers often run at 10–15% utilization and some domestic chips sit idle 70–80% of the time. The upshot: issuing vouchers is easy, finding steady, high-volume workloads is not. Companies often prefer paying for Amazon or Microsoft cloud rather than wrestling with local, harder-to-integrate domestic resources.

Why OpenClaw (a.k.a. “Lobster”) helps — but not enough

OpenClaw-style tools lower the barrier to consumer and developer consumption by enabling high-frequency token use — a single configured instance can trigger hundreds or thousands of API calls a day. It has been reported that WorkBuddy was already used internally by over 2,000 non-technical Tencent employees, and analysts describe QClaw’s stack as a “sandwich”: OpenClaw base, Tencent deployment and routing in the middle, and QQ/WeChat access on top. That matters: heavy developers and automation users generate sustained, predictable load. But mass-market adoption remains shallow. Many early adopters churn after the novelty fades; average users don’t create the continuous, high-volume tasks that compute centers need to raise utilization above the trough.

Political context and a hard commercial truth

Part of the push is geopolitical: U.S. export controls and broader tech decoupling have accelerated domestic build-outs of compute and chip capacity, and local authorities are under pressure to show progress. Still, the economics are stubborn. Cloud providers welcome higher-value, sticky enterprise customers left after the churn. For public compute centers whose business case depends on broad, steady utilization, the current OpenClaw-driven surge may look like a short, sharp peak rather than a durable solution. In short: Lobster can spark demand — but it cannot, by itself, turn underused data centers into profitable infrastructure.