Lift-and-Shift or Refactor? A 2026 Decision Framework for Hong Kong Enterprises Moving to the Cloud

Hong Kong’s cloud migration story is entering its second act. The first wave—roughly 2018 to 2023—was dominated by speed: get off aging on-premise hardware in Kwun Tong and Tsuen Wan data centres, satisfy auditors, and establish a cloud footprint. Lift-and-shift was the default, and for good reason. It was fast, low-risk, and required minimal changes to applications that had run reliably for a decade.

By 2026, the calculus has shifted. Cloud spending across Hong Kong enterprises has matured to the point where CFOs are asking harder questions: why are our cloud bills growing faster than our revenue, and why does our infrastructure still feel as rigid as the servers we retired? Meanwhile, the arrival of generative AI workloads, tighter data residency expectations under the evolving Personal Data (Privacy) Ordinance, and the government’s push toward smart city infrastructure have raised the bar for what “cloud-ready” actually means. The lift-and-shift versus refactor decision is no longer a one-time project choice—it is a recurring strategic question every Hong Kong CIO must answer with rigour.

This article offers a practical decision framework for 2026, grounded in the 6 Rs model, real cost analysis, and lessons from Hong Kong enterprises that have already made the journey.

The 6 Rs in 2026: Beyond the Binary

The 6 Rs framework—Retire, Retain, Rehost, Replatform, Repurchase, and Refactor—has been around for years, but most Hong Kong organisations still collapse it into a binary: lift-and-shift (Rehost) or rewrite everything (Refactor). That binary is expensive and often unnecessary. The value of the 6 Rs lies in its granularity, and in 2026 that granularity matters more than ever because cloud economics have changed.

Retire and Retain are the two Rs that most enterprises under-use. In our work with Hong Kong clients, we routinely find that 15 to 25 percent of applications in a migration inventory are either unused, duplicated across departments, or better kept on-premise for regulatory reasons. A mid-sized Hong Kong logistics firm we worked with in 2025 discovered that four of its 38 applications had zero active users; retiring them before migration cut the project scope by 11 percent and avoided roughly HK$2.4 million in unnecessary cloud run-rate over three years.

Rehost and Replatform are where most migration volume lands, and the distinction is critical. Rehost means moving the application as-is to cloud infrastructure—virtual machines in a VPC, essentially. Replatform means making targeted changes to leverage managed services without re-architecting: moving a self-managed MySQL cluster to a managed database service, or containerising an application without changing its code. The middle ground is where most Hong Kong enterprises in 2026 should be operating, because it captures the bulk of cloud’s operational benefits without the cost and risk of a full refactor.

Repurchase—replacing a custom or legacy system with a SaaS product—has become increasingly attractive for back-office functions like HR, payroll, and expense management, where Hong Kong-specific compliance requirements (MPF calculations, IR56B reporting) are well-handled by local SaaS vendors. Refactor remains the right choice for a narrow set of applications: customer-facing platforms where performance and scalability are competitive differentiators, or workloads that will feed AI models and therefore need modern data architectures.

The Cost Math: What Lift-and-Shift Really Costs Over Time

The standard argument for lift-and-shift is that it is cheaper upfront. That is true, and it is also incomplete. The real question is total cost of ownership over a three-to-five-year horizon, and the answer depends heavily on workload characteristics.

A typical lift-and-shift migration for a Hong Kong enterprise costs significantly less in project spend than a refactor—often 40 to 60 percent less—because it requires minimal application changes, shorter timelines, and fewer specialist engineers. For a 50-application estate, that might mean a HK$8–12 million migration project versus HK$20–30 million for a broad refactor programme. The upfront savings are real and bankable.

The problem emerges in run-rate. Rehosted applications typically run at 30 to 50 percent of their original on-premise efficiency because they are not designed to scale elastically, they carry technical debt into the cloud, and they consume reserved capacity whether or not it is needed. A Hong Kong financial services client that completed a pure lift-and-shift in 2022 found its cloud bill growing 18 percent year-on-year by 2025 despite flat transaction volumes—driven largely by over-provisioned VMs and unoptimised storage tiers.

The refactor case is the mirror image: higher upfront cost, but substantially lower run-rate and far greater elasticity. Refactored applications using serverless or containerised architectures can reduce compute costs by 60 to 70 percent for variable workloads, and they unlock the ability to scale during Hong Kong’s seasonal peaks—Lunar New Year e-commerce surges, September IPO season for financial platforms, or the summer tourist influx for hospitality systems.

The practical 2026 conclusion is that the cost comparison is not a single number but a portfolio decision. Most Hong Kong enterprises should rehost or replatform the stable, low-variance back-office estate, and reserve refactoring budget for the 10 to 20 percent of applications where elasticity, performance, or AI-readiness genuinely drives business value. A blended portfolio typically delivers 25 to 35 percent lower three-year TCO than either extreme.

Hong Kong Case Studies: Three Migration Paths in Practice

Case 1: A family-owned trading house goes Replatform. A 60-year-old Hong Kong import-export firm with 400 staff ran its order management system on an on-premise AS/400 that was increasingly difficult to maintain. A full refactor was quoted at HK$18 million and 18 months—untenable for a business with thin trading margins. Instead, the firm replatformed: the core application was containerised and moved to a managed Kubernetes service, its database migrated to a managed relational service, and its reporting layer rebuilt on a cloud data warehouse. Total cost was HK$6.5 million over nine months. Cloud run-rate is now 22 percent below the equivalent on-premise refresh cost, and month-end reporting that took three days now completes in under four hours.

Case 2: A regional insurer refactors its customer platform. A Hong Kong-based insurer with operations across the Greater Bay Area needed its customer portal to handle policy renewal spikes and support a new AI-driven claims triage feature. A lift-and-shift would have delivered no performance improvement and would have blocked the AI roadmap entirely. The firm refactored the portal into microservices on a serverless architecture, with a separate data pipeline feeding its machine learning models. The project cost HK$14 million over 14 months—roughly double the rehost estimate—but reduced infrastructure cost per policy by 40 percent and cut claims triage time from 48 hours to under 3 hours.

Case 3: A public sector-adjacent organisation chooses selective Rehost. A Hong Kong statutory body with strict data handling requirements migrated 70 percent of its estate via rehost, keeping sensitive workloads in a Hong Kong region with dedicated tenancy. The remaining 30 percent—citizen-facing digital services—were replatformed onto managed services. The organisation explicitly rejected refactoring for most systems, citing auditability and the need for minimal behavioural change during a period of regulatory review. The lesson: for some Hong Kong organisations, migration strategy is a governance decision as much as a technical one, and lift-and-shift is a legitimate, defensible choice when change risk outweighs efficiency gains.

The 2026 Wildcards: AI Readiness, Data Residency, and Talent

Three factors are reshaping the migration decision in Hong Kong specifically, and any 2026 framework must account for them.

AI readiness. Generative AI and machine learning workloads require clean, well-structured, accessible data—and legacy applications rehosted as-is rarely provide it. Enterprises that lifted-and-shifted in 2021 are now discovering that their data is trapped in monolithic databases and batch-oriented systems that cannot feed real-time AI pipelines. If your 2026 roadmap includes AI-driven customer service, fraud detection, or demand forecasting, the applications feeding those models almost certainly need replatforming or refactoring, not rehosting. This is the single biggest reason the lift-and-shift default is breaking down.

Data residency and regulatory expectations. Hong Kong’s data protection regime, combined with sector-specific rules from the HKMA and SFC, means that “move it to the cheapest region” is not an option for regulated enterprises. Cloud providers now offer Hong Kong regions with strong compliance postures, but refactored architectures that span multiple services and regions require careful data flow mapping. Rehosting into a single Hong Kong region is simpler to govern—a genuine advantage that should be weighed honestly against the efficiency benefits of refactoring.

Talent and delivery capacity. Hong Kong’s cloud engineering talent pool remains tight, and refactoring programmes demand skills—distributed systems design, infrastructure-as-code, observability—that are scarcer and more expensive than general cloud operations. A refactor programme staffed by an under-experienced team is a bigger risk than a well-executed rehost. Many Hong Kong enterprises in 2026 are therefore sequencing: rehost first to exit the data centre, then replatform and refactor in waves as internal capability matures, often supported by specialist partners.

A Practical Decision Framework for 2026

So how should a Hong Kong CIO decide, application by application? A workable framework has four steps.

First, classify every application against the 6 Rs with explicit criteria: business criticality, change frequency, regulatory sensitivity, data architecture, and AI roadmap relevance. Resist the urge to default; force a documented rationale for each application’s placement.

Second, model three-year TCO for each option, not just project cost. Include run-rate, staff time, licensing, and the opportunity cost of delayed capability. For most Hong Kong enterprises, the model will show that 60 to 70 percent of applications should be rehosted or replatformed, 10 to 15 percent retired or retained, and 15 to 25 percent refactored or repurchased.

Third, sequence by risk and value. Migrate low-risk, high-savings workloads first to build momentum and fund later phases. Defer refactors until the team has cloud operating experience and the business case is concrete—not speculative.

Fourth, build exit ramps and guardrails. Avoid architectures that lock you into a single provider or make future refactoring harder. Use infrastructure-as-code, containerisation where sensible, and managed services with portable interfaces. In Hong Kong’s fast-moving regulatory and competitive environment, optionality has real value.

Conclusion

The lift-and-shift versus refactor debate is often framed as a false choice, and in 2026 that framing is more misleading than ever. The 6 Rs framework exists precisely because real enterprise estates are heterogeneous: some applications should be retired, some rehosted, some replatformed, and a select few refactored to unlock AI and elasticity. Hong Kong enterprises that treat migration as a portfolio decision—grounded in three-year TCO, honest about regulatory constraints, and sequenced to match their talent reality—consistently outperform those that pick an ideological camp.

The first wave of Hong Kong cloud migration was about leaving the data centre. The second wave, now underway, is about making the cloud earn its keep. That means moving beyond lift-and-shift as a default, investing refactoring budget where it creates genuine competitive advantage, and accepting that the right answer differs by application, by sector, and by ambition. For Hong Kong businesses navigating this in 2026, the winners will be those who ask the harder questions early—and build a migration strategy that can evolve as fast as the market around them.

---

S.C.G.A. Limited is a Hong Kong technology company specialising in machine learning and custom software development. We help enterprises design and execute cloud migration strategies grounded in real business outcomes.