What it is in one line
Turning a "warehouse for files" into a "database you can ask" — dormant files
from years ago can finally be found by content, summarized, and read by an agent.
Who built it
Baidu. The drive launched in 2012 and is among China's largest personal cloud
storage services. Since 2025 it has been operated jointly with Baidu Wenku under
the shared GenFlow agent; on 2026-08-14 GenFlow was renamed "Kuku AI." Model
capability underneath is Baidu's Ernie line.
Read: the drive is an old asset rediscovered by AI — its value is not storage
(storage is a commodity) but the private data users have already stored in it.
What it actually does
- Storage base → serves 1B+ registered users, 100B+ GB stored, up to 30TB
per user, multi-device sync
- AI search → files found by content, no need to remember names or folders
- AI summaries → videos, research reports, documents summarized in one click;
AI notes and mind maps
- GenFlow agent → 100M MAU, 200M task deliveries/month (official figures)
- Backup → photos, documents, contacts, WeChat/QQ files auto-backup
- Kuku AI → the office-agent entry merged with Wenku; cross-device, tasks
keep running with the computer off
What it deliberately does not do: no social, no content community. The drive's
AI serves only one job — managing your own files.
What old behavior it replaces
Finding a file on a cloud drive used to depend on "remembering which folder it
is in and what it is called" — forget, and it might as well not exist. Watching a
video in the drive meant watching it end to end. A report stored two years ago,
its location long forgotten, was never going to be found again.
AI changes "person finds file (browses folders)" into "person asks, agent finds."
Dormant files become searchable, summarizable, and readable by an agent within
authorized bounds — the drive shifts from warehouse to personal knowledge-base
infrastructure. That is the largest role change this category has seen in twenty
years.
Business model
SVIP membership (capacity, download speed, transfers), with AI capability bundled
in as a membership benefit. Since the Wenku merger, the Kuku AI standalone line
charges an office subscription.
Read: the drive has always monetized "capacity anxiety + speed anxiety." AI gives
it a second billing reason — "your data is being put to work" — and also the
leverage for ARPU increases.
Hard numbers
- MAU 101.03M, -1.35% MoM (traffic board, June 2026 global app ranking, #8)
- Official: 1B+ registered users, 100B+ GB stored, storage growing 30%+ YoY
- Official (April 2026): Baidu Drive AI MAU over 80M
- GenFlow: 100M MAU, 200M task deliveries/month (official figures)
- traffic board's 101M MAU and the official "1B registered users" are different
dimensions and must not be compared directly
Four-way read
| Dimension |
Call |
| Founder-product fit |
A corporate incumbent product, not a founder's project |
| Product insight |
AI makes the drive's "private data asset" value visible — a category-level cognitive upgrade |
| Execution quality |
Baidu's model and infrastructure; supply is not the constraint |
| Timing |
Personal knowledge bases are becoming an AI-era necessity; the drive is the most natural carrier |
The call
The drive is an old asset being repriced by AI, and Baidu moved fastest. Storage
itself is a commodity with no differentiation; but "the data users already
stored" is a private asset with high switching costs. AI makes that data
searchable, summarizable, and agent-readable — the drive shifts from a utility
into AI workflow infrastructure. Whoever sediments AI artifacts and lets agents
safely access personal data wins the next round of the drive war.
The base is stable; the story is ARPU. MAU 101M, -1.35% MoM is a classic
incumbent curve — flat, users do not leave, but new-user growth will not break
out either. The real story is whether 80M AI MAU converts into SVIP renewals and
price increases, and whether the Kuku AI office subscription becomes a second
revenue curve.
The competitive landscape is shifting: Tencent's cloud drive entered in 2026,
skipping the price war entirely and pitching an agent ecosystem, cross-app
context, and a permission system. Quark Drive has also folded AI search and AI
video summaries into its browser entry. The drive war moved from "compare
download speed" to "compare who lets agents use data well" — Baidu holds the
user-stock advantage, but Tencent's narrative is clearer.
A data-integrity flag: the official "80M AI MAU" and "GenFlow 100M MAU" are
different dimensions from the third-party verifiable 101M app MAU — the official
figures are larger and fuzzier; cite with care.
What to watch next
① Whether AI MAU converts into SVIP renewals and price increases — whether the
drive's AI-ification closes a commercial loop
② The impact on mindshare once Tencent's drive officially launches — the
entrant's narrative is clearer; will incumbents' stock get pried away
③ How the drive positions itself between "storage" and "office" after the Kuku
AI standalone launch — warehouse, or upgraded work entry point
What you can take from it
Product logic: do not treat AI as a feature list; treat it as the key that
gets stale assets reused. For any tool holding historical data, the highest-value
move is making old data searchable, summarizable, and agent-callable — cheaper
than acquisition, more useful than adding features.
Pricing structure: nothing special to take. SVIP membership is an incumbent
model; AI has not yet formed an independent charging structure.
Verdict
Worth watching. Data is real and the base is solid — a typical case of a stale
asset being repriced by AI. The real fork is whether AI turns from "feature" into
"revenue" — and whether Tencent's entry rewrites the rules of the drive war.