How Can Chinese Enterprises Navigate the US GPU Embargo and Build an Independent IT Stack?

Background

Recent U.S. export controls that prohibit companies such as Nvidia from shipping high‑end GPUs and CPUs to China have accelerated discussions about a long‑term “silicon war” between the two countries. Analysts predict that the conflict will evolve into a prolonged struggle rather than a quick resolution.

Potential Divergence of IT Stacks

At a recent technology forum a university expert warned that the global IT ecosystem could split into two mutually incompatible stacks:

Traditional stack – based on existing Western hardware, software, and development tools.

Domestic‑China stack – built around Chinese‑origin components, operating systems, and middleware.

While a fully independent stack is unlikely in the short term, critical domains (e.g., high‑performance computing, data center networking) may become isolated as domestic alternatives mature.

Domestic Replacement (Xinchuang) Strategies

Enterprises are adopting varied approaches to reduce reliance on foreign technology:

Pause or abandon – some organizations deem the effort infeasible and maintain the status quo.

Wait for diplomatic improvement – others hope that relations will ease and postpone major changes.

Comprehensive overhaul – a minority redesign their entire IT architecture, migrate workloads to cloud platforms, and replace foreign components with domestic equivalents.

Incremental substitution – the most common path, focusing first on low‑risk, easily replaceable components while keeping core services running.

Key Challenges

Cost pressure : Post‑COVID‑19 financial weakness makes large‑scale investment in new hardware difficult.

Maturity and capacity of domestic products : Many Chinese alternatives lack the performance, reliability, or production volume of their foreign counterparts.

Missing ecosystem : Application‑service layers, management tools, and integration frameworks for domestic hardware are not yet complete, risking operational instability if a wholesale swap is attempted.

Recommended Tactical Approach

Adopt a “wait‑and‑endure” posture similar to a strategic retreat: give domestic vendors time to improve while gradually evolving the enterprise architecture. Specific steps include:

Audit the existing stack to identify components with viable domestic replacements.

Prioritize substitution of low‑risk items (e.g., peripheral devices, non‑critical servers).

Maintain current cloud platforms and data‑center operations for critical workloads.

When a domestic alternative reaches sufficient maturity, replace the corresponding foreign component and validate stability before proceeding.

Case Study: SAN Switch Dependency

A client designing a “risk‑free IT stack” discovered that the only globally available SAN switch was produced by a U.S. vendor (Brocade). The client demanded a fully cloud‑native solution and rejected any bare‑metal deployment.

Technical workaround proposed:

# Example architecture using bare‑metal + JBOD
- Deploy bare‑metal servers in the data center.
- Attach JBOD (Just a Bunch Of Disks) enclosures for block storage.
- Expose the storage via iSCSI or NFS to cloud workloads.
- Use cloud‑native orchestration (e.g., Kubernetes CSI drivers) to manage the bare‑metal storage.

Key observations:

Cloud providers already run database services (e.g., RDS) on bare‑metal servers, demonstrating that bare‑metal resources can be managed within a cloud control plane.

By migrating workloads that require the SAN switch to bare‑metal + JBOD, the organization can keep the rest of the stack cloud‑native while awaiting a domestic SAN solution.

Long‑Term Outlook

Given the expected duration of the technology conflict, enterprises should:

Replace components where mature domestic products exist.

Defer substitution for items lacking viable alternatives, monitoring the market for emerging competitors.

Recognize that monopolistic positions (e.g., Brocade’s dominance in SAN switches) will erode over time as Chinese vendors invest in R&D and scale production.

This incremental, pragmatic methodology balances operational continuity with the strategic need to reduce foreign dependency.