Connect with us

NEWS

Tencent Cloud Leads APAC Gaming MongoDB as Agents Arrive

Tencent Cloud led APAC gaming MongoDB at 20.6% in 2025, with MongoDB Official third, while it rebuilt the stack for AI agents.

Published

on

Tencent Cloud took first place in 2025 share, growth, and technical scores for MongoDB used by Asia-Pacific game studios. Frost & Sullivan and LeadLeo Research Institute put that result in their 2026 gaming database study, and TencentDB for MongoDB held 20.6% of the slice while growing 51.8%.

MongoDB’s own cloud sat at 14.1%. The week the study appeared, Tencent Cloud vice president Wang Yicheng said the company would treat AI agents as a new kind of database user.

Tencent Cloud Took 20.6% of APAC’s MongoDB Gaming Slice

Frost & Sullivan and LeadLeo posted the joint 2026 gaming database report on May 28, 2026. Tencent Cloud issued the English ranking statement on June 12. The score is 2025 revenue in a narrow track, MongoDB and Mongo-compatible services sold into Asia-Pacific games, not the whole game-cloud bill.

Tencent Cloud’s recap of the core findings lists 20.6% share and 51.8% growth for that year, with five other named lines behind it. Lead analysts Anqi Song and Qing Li scored vendors on five dimensions and 25 sub-indicators, from protocol fit and elastic scale to backup, AIOps diagnosis, and regional rules.

APAC GAMING MONGODB VENDORS IN 2025

Vendor Share Revenue growth
Tencent Cloud 20.6% 51.8%
Major domestic rival 16.3% 26.8%
MongoDB Official 14.1% 32.6%
AWS 12.4% 21.3%
Azure 9.2% 23.4%
Other domestic rival 6.6% 16.4%

The recap still withholds the domestic runner-up’s brand. Competitive notes published with the paper describe Alibaba Cloud as the other local platform with gaming coverage, then AWS for multi-region delivery. The six named lines add to 79.2%, so a fifth of the tracked slice sits with everyone else.

Player Files, Loot, and NPC Memory Do Not Fit Rows

Relational stores still take the largest piece of Asia-Pacific game database spend, 36.0% in 2025. Document stores are next at 25.0%, then key-value at 20.0% and time-series at 7.0%. Mainland China is 52.0% of the geographic mix, Japan and Korea 28.0%, and Southeast Asia 14.0%.

The paper’s argument is that game data stopped looking like accounts and payments. Version drops, live events, and AI features keep changing the shape of a player record, and a fixed table fights that. Tencent Cloud’s international product page sells a schema-free mode for game backends as a way to skip table rewrites and cut version work by 60%, a vendor claim, not a studio audit.

WHAT THE REPORT SAYS GAMES NOW STORE

  • Player files: Profiles, role stats, task progress, and social graphs that change with every patch.
  • Live ops: Event configs, battle logs, chat, and ranking writes that cannot wait on a batch job.
  • Generated content: UGC and AIGC assets, tags, and rights clues that arrive as nested documents.
  • Agent context: NPC memory, retrieval for dialogue, and recommendation features that sit beside the live session.

That mix is why the authors call MongoDB-style stores one of the faster core tracks in the category. They also sketch a three-step path studios already know: first keep saves safe, then make schema changes cheap, then feed real-time ops and AI from the same store.

Wuthering Waves Already Stress-Tests the Store

Tencent Cloud says TencentDB for MongoDB already covers open-world RPGs, SLGs, and Weixin mini games, and it names three live titles as proof. Kuro Games’ Wuthering Waves is the heavyweight. Hortor’s Fish Kingdoms: Idle Arena and EasyGame’s Sheep a Sheep are the other two on the English list.

A Tencent Cloud customer note on Kuro says Wuthering Waves drew more than 32 million pre-registrations and stores data in the hundreds of terabytes, with a container setup across six regions. That page is about launch load and global rollout, not a public MongoDB schema. The ranking paper still uses the title as a large-game primary-storage example, then uses hit mini games for the opposite job, short traffic spikes that arrive and vanish with a social wave.

The split is the product story Tencent wants. One service is supposed to hold the long-running open world and also absorb a mini game that suddenly needs a lot of writes. June 2026 release notes added CPU elastic scale-out to TencentDB for MongoDB for that second case, festival peaks and campaign days, billed on how long the extra cores stay on.

Third Place Went to MongoDB’s Own Cloud

MongoDB Official grew 32.6% in the same 2025 cut, faster than AWS at 21.3% and Azure at 23.4%, and still finished behind Tencent Cloud and the unnamed domestic rival. The category carries MongoDB’s name. The share leader is a protocol-compatible managed service on a Chinese cloud.

Tencent already sold that compatibility as a feature. In April 2025 its official cloud account pitched a MongoDB partnership for games, commerce, and social apps. The 2025 scoreboard says the compatible cloud store, not the official line, took the gaming lead in Asia-Pacific.

That is a local-cloud outcome as much as a kernel outcome. Song and Li’s method weights migration tools, multi-AZ backup, regional rules, and ops diagnosis beside raw throughput. Tencent Cloud’s recap leans on game-scene fit, local compliance, and horizontal scale. The paper’s AWS note, by contrast, leans on global regions and serverless elasticity. Studios that ship first in China and then abroad are shopping that difference, not a pure engine bake-off.

What Tencent Shipped the Same Week as the Report

On May 29, 2026, Wang Yicheng stood up an Agent-native database stack and said the next user of these systems is not only a human developer.

Over the past ten-plus years, Tencent Cloud databases have been through the core tests of internet-scale concurrency and homegrown, controllable technology. Databases are now moving toward a native AI 3.0 era. Tencent Cloud will treat the Agent as the new user and redesign the product and capability system.

Wang Yicheng, Vice President, Tencent Cloud, May 29 briefing

The briefing bundled more than MongoDB. Agent Memory, DatabaseClaw, and a TDSQL-B rebuild that folds transactions, vectors, full-text, and graph into one base were the named pieces, and Tencent said those tools already run inside Yuanbao, QClaw, CodeBuddy, and Claw Pro.

CLAIMS FROM THE MAY 29 BRIEFING

  • Agent Memory: Tencent says long-task success rose 30% and token cost fell as much as 60%, while a PersonaMem score moved from 48% to 76%, and the open-source repo passed 4,000 stars in its first week.
  • DatabaseClaw: One agent is said to cover 14 Tencent Cloud database products and 1,600-plus OpenAPIs, with skills drawn from 100,000-plus DBA tickets; Tencent’s internal test puts CPU diagnosis at 15 times faster.
  • Kernel add-ons: Hunyuan in the optimizer is claimed to cut slow-SQL latency by 60%+, and a predictive scaler is said to hot-change capacity before a periodic peak.

TencentDB for MongoDB is the NoSQL piece Tencent says will keep moving toward unified Agent multi-model data, agentic app work, and intelligent ops. Vector search, real-time analytics, and multi-model processing are the features it lists for AI NPCs, recommenders, AIGC, and customer service. By August it had also published TencentDB Agent Memory on GitHub as a team-level memory hub, a later drop that sits on the same May argument rather than on a new ranking.

Extra Cores for Gacha Day and Mini-Game Spikes

The ranking paper’s growth story is not only open worlds. It says TencentDB for MongoDB rose because it can hold a large game’s primary store and also take the short spike of a breakout mini game. That second load is ugly for a cluster sized for yesterday’s traffic.

Tencent Cloud’s MongoDB guide now documents custom CPU scale-out for surges. You pick extra cores, they come on immediately, and they stay on, and on the bill, until you turn them off. Instances must run MongoDB 4.0 or later, original cores cannot exceed 32, and a single add is capped at the lesser of the smallest shard’s cores and 24. The host still needs spare CPU, or the call fails.

That is a blunt tool, not a new query engine. It exists because gacha rotations, festival events, and social mini games still create the kind of peak the 2025 share table is capturing. Tencent Cloud’s product page also quotes 30,000-plus queries per second on a tuned kernel; treat that as a spec-sheet ceiling, not a Wuthering Waves measurement.

The MongoDB Track Is Still the Smaller Bill

The trophy sits on a thin slice. The same paper puts the whole Asia-Pacific gaming database market at RMB 1.95 billion in 2021 and RMB 3.51 billion in 2025, a 15.76% compound rate, and at RMB 8.08 billion in 2030 at 18.19%. The MongoDB track inside that total is RMB 410 million in 2025, about 11.7% of the 2025 bill, rising to RMB 1.29 billion in 2030 at nearly 26% a year.

Document stores as a class are already 25.0% of 2025 spend, larger than the MongoDB-compatible cut, which means other document products still share that quarter. Tencent Cloud is first in the named MongoDB racing line, at 20.6%, with a growth rate the recap calls the fastest. It is not first in every game database, and relational systems remain the biggest cheque.

FROM THE PAPER TO THE BILL

  1. May 28, 2026: Frost & Sullivan and LeadLeo post the Asia-Pacific gaming database paper that ranks 2025 MongoDB vendors.
  2. May 29, 2026: Wang Yicheng presents Agent Memory, DatabaseClaw, and the wider Agent-native database upgrade.
  3. June 12, 2026: Tencent Cloud issues the English statement that it led 2025 share, growth, and technical scores.
  4. June 19, 2026: DatabaseClaw moves onto commercial billing, with a free experience tier and a paid enterprise tier.

The 2025 table is closed. The spend line it describes is not. If the paper’s MongoDB path holds, a RMB 410 million racing line becomes a RMB 1.29 billion one by 2030, and Tencent Cloud is already selling memory, vectors, and an ops agent into that climb rather than waiting for the next scorecard.

Harry is the editor of Oton Technology, an independent site he owns and edits, covering the part of technology that people actually have to act on. After ten years in journalism, first reporting and then editing, he works from primary material by habit: the advisory rather than the write up of it, the filing rather than the press release, the changelog rather than the launch video. Every figure in an article carries its source and its date, and where a number comes from a vendor or an analyst model rather than a count, he says so plainly instead of letting it stand as established fact. What he leaves out is anything he could not verify himself, which on a beat full of unnamed supply chain claims removes a great deal. That standard applies across all the sections the site publishes for an international audience, from artificial intelligence and security to phones, computers, gaming, crypto and the software businesses depend on. He corrects errors in the open and labels them, because a site that hides its mistakes is asking readers to trust the rest on nothing.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending