Editor’s observe: This text is republished from Jumpstart, a TechNode Group publication.
In 2023, AI software program was DataCanvas’s enterprise. By the primary half of 2026, it was nearly absent from its gross sales.
The change is unusually stark. AI foundational software program companies generated 76.5% of the Chinese language firm’s income in 2023. Three years later, their contribution had fallen to 0.1%.
This was greater than an outdated division being overshadowed by a faster-growing one. Software program-service income itself contracted—from RMB81.3 million in 2023 to lower than RMB500,000 within the first six months of 2026. In the meantime, whole annual income climbed from RMB106.2 million in 2023 to RMB1.10 billion in 2025, a compound annual development price of 221.5%.
DataCanvas had constructed a special enterprise.
The Beijing firm, recognized in China as 九章云极, started redirecting its assets towards computing infrastructure in early 2023. It subsequently added cloud companies, turning from a provider of software program instruments into an organization that helps assemble computing infrastructure and sells entry to its capability.
On September 29, 2026, it filed for a Hong Kong itemizing. The applying opens a window onto that reinvention: an organization based greater than a decade in the past, approaching the general public market with its unique income engine successfully switched off.
Earlier than the Servers
DataCanvas was established in 2013, lengthy earlier than generative AI made computing infrastructure a mainstream enterprise story.
Its early proposition was to assist organizations work with information and construct machine-learning fashions. A visible modelling platform adopted in 2014, with subsequent merchandise extending throughout mannequin improvement and deployment.
These instruments addressed a clumsy hole inside enterprises. Having information didn’t imply an organization might readily flip it right into a working mannequin. Constructing a mannequin didn’t imply it may very well be reliably included into on a regular basis operations.
DataCanvas provided software program to bridge these steps. Its merchandise helped clients put together information, assemble analytical workflows and put fashions to work.
However constructing the expertise was not the identical as discovering a enterprise.
In 2022, founder and chairman Fang Lei recalled that the workforce had spent its early years concentrating on code whereas underestimating the market. The founders had been skilled engineers; understanding what clients would truly pay for took longer.
Banks grew to become an vital entry level. They’d substantial information, established expertise budgets and causes to enhance how data knowledgeable selections. Working with such clients gave DataCanvas a business setting for its instruments—and a clearer understanding of what it meant to promote them.
The corporate provided enterprises the means to develop and use machine studying. Prospects nonetheless had to supply the computing environments through which these instruments would function.
That boundary would turn into the following enterprise alternative.
The Resolution to Change the Enterprise
By 2023, massive AI fashions had been making the equipment behind software program a lot tougher to disregard.
Prospects wanted substantial computing assets, coordinated throughout processors, networks and storage. They needed to get hold of that capability, make the elements work collectively and hold workloads operating.
For DataCanvas, this opened a selection: proceed promoting software program into these environments, or take accountability for extra of the infrastructure itself.
It selected the latter.
The corporate started offering computing infrastructure options in 2023 and signed its first venture contract that September. These had been bigger assignments than delivering a standalone software program platform. They concerned designing the computing surroundings, procuring and integrating {hardware}, and deploying techniques to help clients’ workloads.
DataCanvas redirected assets towards the brand new operation whereas progressively withdrawing from its unique software-service enterprise.
The business penalties appeared rapidly. Infrastructure options generated RMB24.9 million in 2023. By 2025, they introduced in simply over RMB1 billion and accounted for 91.2% of whole income.
The outdated enterprise moved in the other way. Its income fell in each 2024 and 2025.
This was the decisive break in DataCanvas’s historical past. Promoting instruments had given it a task inside clients’ expertise operations. Delivering infrastructure made it answerable for extra of the surroundings these operations relied on.
The corporate had moved nearer to the equipment—and brought on a bigger share of the work.
Construct It, Then Maintain It Working
The subsequent step got here when DataCanvas launched its cloud service in September 2024.
An infrastructure venture helps a buyer set up computing capability. A cloud service lets clients entry capability with out arranging a complete set up themselves.
For DataCanvas, that meant shifting past venture supply towards an ongoing working relationship.
The primary cloud-service income was tiny: RMB196,000 in 2024. It rose to RMB77.6 million in 2025 and reached RMB153.8 million within the first half of 2026. By then, cloud companies equipped 26.1% of income, alongside the 73.8% generated by infrastructure options.
Its cloud-service enterprise buyer rely elevated from two in 2024 to 75 in 2025 and 84 within the first half of 2026.
That development provides the reinvention two distinct phases. Infrastructure initiatives first established the size of the brand new enterprise. Cloud companies then started altering how the corporate maintained and monetized its relationship with clients.
A accomplished set up might produce venture income. Offering entry to computing capability created a cause for purchasers to maintain shopping for.
The software program firm had turn into an infrastructure supplier. It was now additionally changing into an operator.
What Survived the Reinvention
DataCanvas didn’t cease writing software program when software program companies stopped driving its gross sales.
It launched Alaya NeW OS in April 2024, inserting a software program management layer on the centre of its newer enterprise. The system coordinates computing assets and the workloads utilizing them. Its subsequent cloud platforms construct on that basis.
That is the revealing stress within the transformation. DataCanvas was withdrawing a method of promoting software program whereas retaining the engineering that made its new companies work.
Beforehand, the platform was one thing a buyer purchased. More and more, it grew to become the mechanism by which DataCanvas delivered one other product: usable computing capability.
The corporate additionally developed its personal computing measurement unit, DCU, which it markets in China as “one diploma of computing energy,” echoing the on a regular basis expression for a unit of electrical energy.
The concept is to make computing simpler to measure and buy throughout totally different assets, with prices linked to consumption. It expresses the corporate’s new ambition in sensible phrases: clients ought to have the ability to purchase computing as a service with out assembling each ingredient wanted to provide it.
For DataCanvas, this provided a special reply to a query it had encountered early in its life: what ought to an enterprise purchase from the corporate?
Its first reply was software program that helped clients develop and deploy fashions. Its newer reply included the assets on which these fashions run.
The Value of Beginning Once more
Changing a enterprise can produce a spectacular development chart. It additionally adjustments what development prices.
DataCanvas reported an total gross margin of 40.6% in 2023, falling to 11.7% in 2024 as infrastructure options grew to become the dominant supply of gross sales. The margin recovered to 17.4% in 2025 and 31% within the first half of 2026.
The preliminary decline exhibits the gap between the 2 companies. An infrastructure contract can include substantial {hardware} procurement and deployment work. A a lot bigger order can subsequently deliver a a lot bigger value of supply.
Working computing companies provides one other set of calls for. Capability should be obtainable when clients want it, supported by techniques that allocate assets and hold workloads functioning. Profitable an order is just a part of the job.
DataCanvas combines preparations with infrastructure house owners and companions with funding of its personal. Transferring into computing supply nonetheless brings tools prices, capability commitments and the persevering with process of turning obtainable assets into paying utilization.
Profitability stays unfinished. Its adjusted internet loss narrowed from RMB194.3 million in 2023 to RMB91.3 million in 2025, then to RMB14.9 million within the first half of 2026.
Its reported first-half loss was a lot bigger, at RMB1.09 billion, principally due to RMB1.07 billion in fair-value adjustments in monetary liabilities. That accounting cost explains many of the hole between the reported and adjusted figures.
DataCanvas has demonstrated that it might substitute its outdated income. It’s nonetheless establishing how a lot it might earn from the alternative.
Taking the New Firm Public
The Hong Kong IPO arrives three years into the transformation.
DataCanvas plans to make use of the proceeds to strengthen analysis and improvement, increase computing assets for its data-centre initiatives, develop gross sales channels and help working capital. {Hardware} procurement varieties a part of the proposed capability growth.
These plans prolong the choice made in 2023. The corporate is in search of capital to deepen a enterprise that has already reshaped its operations and income.
For potential shareholders, the duty is to know what that enterprise has turn into. DataCanvas carries the historical past and technical capabilities of a software program firm. Most of its income now comes from infrastructure supply, whereas cloud operations account for a rising share.
Its founding identification not describes the way it makes most of its cash.
What makes DataCanvas distinctive is the extent of the selection it has already made. It spent a decade constructing instruments for enterprises to make use of AI, then intentionally stepped away from promoting these instruments as its fundamental enterprise. Of their place, it constructed the infrastructure and companies wanted to maintain these techniques operating.
The income transformation is basically full. The Hong Kong itemizing brings the following query into public view: whether or not the enterprise DataCanvas selected to turn into can flip its reinvention into sturdy earnings.
