Crazy Rich Hong Kong | Data Centres, Cloud and the AI Computing Infrastructure Economy begins with a modern paradox: artificial intelligence looks weightless on a screen, but underneath it sits land, power, cooling equipment, fibre networks and very expensive computers.
Search for “Hong Kong data centre”, “Hong Kong cloud computing”, “Hong Kong AI infrastructure”, “Sandy Ridge data centre”, “Hong Kong supercomputer” or “AI computing Hong Kong” and the physical foundation of the digital economy appears.
Hong Kong’s Digital Policy Office said in April 2026 that the city had about 5,000 PFLOPS of aggregate computing power, including 3,000 PFLOPS at Cyberport’s AI Supercomputing Centre. The new Sandy Ridge Data Facility Cluster, awarded in March 2026 and now under construction, is targeted to reach 180,000 PFLOPS by 2032—around 36 times current Hong Kong capacity—with estimated cumulative investment of at least HK$23.8 billion.
The Crazy Rich story is not “Hong Kong wants more servers”. It is that computing power is becoming infrastructure in the same way ports, roads and electricity networks became infrastructure in earlier industrial eras.
Did You Know? AI Has a Physical Address
Ask an AI system a question and the response appears instantly.
That speed can hide the machinery.
Somewhere, processors are performing enormous amounts of computation.
Those chips need electricity.
Electricity creates heat.
Heat needs cooling.
The data needs networking.
Digital intelligence depends on physical systems.
Why Data Centres Matter to More Than Technology Companies
Hong Kong’s Digital Policy Office describes data centres as infrastructure supporting financial services, trading, logistics and many other sectors.
That is important because cloud infrastructure has become a general-purpose layer underneath modern business.
Banks use it.
Hospitals use it.
Retailers use it.
Universities use it.
Government uses it.
The data centre is the invisible building inside many visible businesses.
Cyberport’s 3,000-PFLOPS AI Supercomputing Centre
Cyberport’s AI Supercomputing Centre reached 3,000 PFLOPS by the end of 2025.
That creates high-performance computing capacity for artificial-intelligence research and applications.
PFLOPS measures floating-point operations per second.
The numbers become enormous because modern AI models require vast parallel computation.
The digital economy is increasingly measured in mathematical throughput.
Sandy Ridge: The 180,000-PFLOPS Leap
The Sandy Ridge Data Facility Cluster represents a much larger planned expansion.
The site covers around ten hectares and is intended for advanced data facilities and related industries.
The Government awarded the site in March 2026.
By 2032, the planned computing capacity is 180,000 PFLOPS.
That is not an incremental upgrade.
It is a change in scale.
HK$23.8 Billion: Why Computing Requires Capital
The 2026 Policy Address estimates cumulative investment in Sandy Ridge at no less than HK$23.8 billion.
High-performance data centres are capital-intensive because they require:
- advanced processors;
- specialised power systems;
- cooling equipment;
- backup generation;
- high-capacity fibre;
- cybersecurity;
- physical security; and
- highly reliable buildings.
The cloud is expensive because reliability is expensive.
The Electricity Problem
Computing power consumes electricity.
As AI workloads grow, power supply becomes one of the biggest constraints on data-centre expansion globally.
Hong Kong’s 2026 policy explicitly frames its direction as “high efficiency computing power, stable electricity supply, low-carbon transition”.
That sentence captures the central engineering problem:
more computation without making the energy system unsustainable.
Cooling Is the Hidden AI Industry
A processor that performs trillions of operations generates heat.
Too much heat reduces performance and can damage equipment.
Traditional air cooling is increasingly supplemented by liquid-cooling and other advanced systems for dense AI racks.
This creates new markets in thermal engineering, facility design and energy management.
AI does not only create software jobs.
It creates plumbing jobs for very intelligent machines.
Why Hong Kong’s Financial Sector Needs Data Infrastructure
Banks and exchanges process enormous volumes of transactions with strict requirements for latency, cybersecurity and uptime.
A financial centre therefore needs digital infrastructure capable of operating continuously.
If the server stops, the bank does not become slightly slower.
The financial service may stop functioning.
Reliability becomes economic continuity.
Cloud Computing Changes the Economics of Starting a Company
A start-up once needed to purchase much of its own computing hardware.
Cloud services allow companies to rent computing, storage and databases when needed.
That converts capital expenditure into more flexible operating expenditure.
Small companies gain access to infrastructure once available mainly to large organisations.
The cloud lowers the minimum scale required to experiment.
AI Changes the Demand Curve Again
Traditional cloud workloads often scale predictably.
Training large AI models can demand enormous bursts of specialised GPU or accelerator capacity.
Inference—the process of running trained models—can then create continuous compute demand when millions of users interact with those models.
The result is a new computing economy built around scarcity of advanced chips and power.
Why Fibre Connectivity Matters
A data centre without network connectivity is an expensive room full of isolated machines.
Hong Kong benefits from extensive international telecommunications links and submarine cable connectivity.
That allows digital information to move rapidly between Hong Kong, Mainland China and global markets.
Computing creates value when results can travel.
Latency: The Geography of Milliseconds
Latency is the time data takes to travel between systems.
For ordinary web browsing, a small delay may be annoying.
For financial trading, real-time gaming, AI agents or industrial control, milliseconds can matter.
This creates a new geography where physical distance still matters inside a digital world.
Why Cybersecurity Grows with Computing Power
More digital infrastructure creates more valuable targets.
Data centres must therefore protect both physical equipment and digital systems.
Security includes access control, network segmentation, monitoring, redundancy, encryption and incident response.
A powerful computer system without security is powerful for the wrong person too.
The AI Subsidy Scheme
Hong Kong is using public policy to help researchers access expensive computing resources.
The 2026 Policy Address announced another HK$1 billion injection into the Artificial Intelligence Subsidy Scheme to support frontier research.
This addresses a modern education and research inequality:
the ability to think of a good experiment does not automatically mean you can afford the computation required to run it.
Words-to-Technology UI Becomes Compute-to-Technology Infrastructure
AI lowers the barrier between human intention and computation.
People can increasingly describe a problem in natural language and ask machines to perform complex digital work.
But that apparent simplicity increases pressure underneath the interface.
More people can ask more complicated things.
The words-to-technology interface becomes easier while the infrastructure required to satisfy those words becomes larger.
Data Centres Are Real Estate with a Very Different Tenant
A conventional office values windows, meeting rooms and human comfort.
A data centre values power density, fibre routes, cooling, redundancy and security.
That creates a specialised real-estate category.
The building is designed around machines rather than people.
Singapore and Hong Kong: Computing Hubs in Land-Scarce Cities
Singapore and Hong Kong both face the same awkward constraint:
data centres require land and electricity, and both cities have limited quantities of each.
Singapore has imposed strict efficiency requirements while expanding capacity selectively.
Hong Kong is concentrating future expansion through Cyberport and Sandy Ridge.
The common lesson is that digital growth still has physical limits.
What Students Can Learn from Data Centres
Physics
Electricity, heat transfer and cooling determine whether computers can run safely.
Mathematics
PFLOPS, bandwidth, storage and latency quantify different dimensions of computing systems.
Economics
Cloud computing changes fixed costs into scalable services.
Geography
Power supply, land, cable routes and proximity to users determine where digital infrastructure can grow.
Civilisation mechanics
Modern civilisation increasingly depends on computational substrate that must remain operational even when users cannot see it.
Ten Vocabulary Words for AI Infrastructure
1. Data centre
A specialised facility housing computer systems, networking and storage infrastructure.
2. PFLOPS
Peta floating-point operations per second, a measure of computing performance.
3. Cloud computing
On-demand delivery of computing resources over networks.
4. Latency
The time delay between sending and receiving information.
5. Redundancy
Backup systems designed to keep operations running when one component fails.
6. Accelerator
Specialised computing hardware designed to perform certain tasks, including AI workloads, very efficiently.
7. Bandwidth
The amount of data that can be transferred through a connection over time.
8. Inference
Using a trained AI model to produce predictions or outputs.
9. Cooling
Systems that remove heat generated by computing equipment.
10. Uptime
The proportion of time a system remains available and operational.
Frequently Asked Questions
How much computing power did Hong Kong have in 2026?
The Digital Policy Office said aggregate capacity was about 5,000 PFLOPS in April 2026.
How much is at Cyberport?
Cyberport’s AI Supercomputing Centre had reached 3,000 PFLOPS by the end of 2025.
What is Sandy Ridge?
A new advanced data-facility cluster in the North District being developed for data centres and related industries.
How large will Sandy Ridge become?
The 2026 Policy Address targets up to 180,000 PFLOPS of computing power by 2032.
How much investment is expected?
At least HK$23.8 billion in estimated cumulative investment.
Why do AI systems need so much infrastructure?
Training and running modern AI models require specialised processors, electricity, cooling, networking and storage at large scale.
Helpful Reading Across the Hong Kong and Singapore Graph
- Crazy Rich Hong Kong | Cyberport, FinTech and the Digital Assets Economy
- Crazy Rich Hong Kong | Science Park, Biotech and the Deep-Tech Innovation Economy
- Crazy Rich Singapore | Data Centres, Cloud and AI Infrastructure
- Crazy Rich Singapore | Quantum Computing, Quantum Networks and Deep Tech
References and Current Sources
- Hong Kong Digital Policy Office, Data Centre Facilitation.
- Hong Kong Digital Policy Office, Digital Infrastructure, Data Ecosystem and Governance, 13 April 2026.
- Hong Kong SAR Government, 2026 Policy Address: Sandy Ridge Data Facility Cluster.
Crazy Rich Hong Kong Builds the Machine Under the Machine
AI feels magical because the infrastructure disappears behind the interface.
But civilisation does not run on magic.
It runs on power, cooling, networks, chips and people who know how to keep them alive.
Did you know? The richer the digital world becomes, the more valuable the physical machine underneath it becomes too.