AI Server Immersion Cooling: 100G QSFP28 & Extender Solutions | Optech
OptechTWShare
AI Server Immersion Cooling Connectivity: Optech 100G QSFP28 and Extender Solutions
AI infrastructure is rapidly increasing power density inside modern data centers.
As GPU servers, high-performance CPUs, accelerators, network adapters, and high-speed switches consume more power, traditional air cooling is becoming increasingly difficult to scale efficiently.
For this reason, immersion cooling is becoming an important thermal-management option for AI servers and high-density computing environments.
But cooling the server is only part of the challenge.
The network connections inside and around the immersion tank must also operate reliably.
For Taiwan manufacturer Optech, immersion cooling is not simply a server-cooling application. It is also a new high-speed connectivity requirement.
Based on the solution shown in the image, Optech provides two important product directions:
Immersible 100G QSFP28
and
Immersion Liquid Cooling Extenders
These products help customers maintain high-speed optical connectivity while adapting network architecture to the physical and thermal requirements of immersion-cooled AI infrastructure.
Optech can also work with customers on customized project requirements, from materials and cable lengths to mechanical design, platform compatibility, labeling, and TAA-related procurement needs.

Why AI Servers Are Moving Toward Immersion Cooling
Modern AI servers can generate significantly more heat than traditional enterprise servers.
A dense AI rack may include:
- multiple GPUs
- high-performance CPUs
- NICs and DPUs
- high-speed storage
- power supplies
- 100G, 200G, 400G, or higher-speed network interfaces
As these systems become more powerful, the amount of heat generated per rack continues to increase.
Immersion cooling addresses this challenge by placing equipment directly into a dielectric cooling fluid.
The fluid absorbs heat from the electronic components and transfers it to an external cooling system.
The image illustrates a typical architecture:
AI Server → Immersion Tank → Cooling Liquid → Heat Exchanger → Facility Water / Dry Cooler / Adiabatic Cooler
This can help support higher compute density while reducing dependence on conventional airflow-based cooling.
Networking Is a Critical Part of Immersion-Cooled AI Systems
A server may be submerged in cooling liquid, but it still needs to communicate with the rest of the data center.
That means network connectivity must pass between:
equipment inside the immersion tank
and
the external network infrastructure
This creates several engineering questions.
Should the optical transceiver itself be submerged?
Should the transceiver be moved outside the tank?
What cable materials are compatible with the cooling fluid?
How should the optical connector be serviced?
What happens when a transceiver must be replaced?
Optech can support different architecture approaches depending on the customer's system.
Optech Immersible 100G QSFP28
One of the key products highlighted in the image is an Immersible 100G QSFP28 solution.
QSFP28 is a widely used form factor for 100G networking and remains important in AI, enterprise data centers, storage networks, HPC systems, and cloud infrastructure.
For an immersion cooling environment, however, a conventional air-cooled QSFP28 module may not automatically be suitable.
An immersion-oriented QSFP28 solution may need to consider:
- dielectric-fluid compatibility
- module housing materials
- cable jacket materials
- connector stability
- adhesives
- labels
- pull tabs
- thermal behavior
- long-term exposure
- optical connector protection
Optech can work with customers to evaluate these requirements according to the actual immersion system.
Advantages of an Immersible 100G QSFP28 Solution
Direct High-Speed Connectivity Inside the Immersion System
An immersible QSFP28 can allow the network interface to remain directly connected to the submerged server or switch.
This can simplify the physical architecture when the customer prefers to keep the optical interface close to the submerged equipment.
100G Bandwidth for AI and HPC Networks
100G remains an important connectivity speed for many:
- AI server clusters
- storage systems
- HPC platforms
- cloud networks
- research computing environments
Immersion-compatible 100G connectivity can therefore help customers modernize cooling systems without abandoning established 100G network infrastructure.
Reduced Dependence on External Adapters
In some designs, using an immersible transceiver can reduce the need for additional extension hardware.
The best architecture depends on the system layout, maintenance strategy, and coolant.
Optech Immersion Liquid Cooling Extenders
The second major solution shown in the image is the Immersion Liquid Cooling Extender.
Instead of requiring the actual optical transceiver to remain inside the cooling fluid, an extender can move the optical interface away from the submerged equipment.
A simplified architecture can look like:
Immersed Server / Switch Port → Optech Immersion Extender → External Optical Transceiver → Fiber Network
This approach creates a separation between the liquid-cooled equipment and the optical transceiver.
Why Use an Immersion Liquid Cooling Extender?
1. Move the Optical Transceiver Outside the Liquid
One of the biggest advantages is the ability to keep the optical transceiver in a dry, accessible location.
This can reduce direct fluid exposure for the optical module.
2. Easier Maintenance
If the optical module is located outside the tank, technicians can replace or inspect it without accessing the submerged equipment.
This may simplify:
- module replacement
- compatibility changes
- fiber cleaning
- optical power troubleshooting
- spare management
3. More Flexibility in Optical Module Selection
The external optical module can potentially be selected according to the required:
- distance
- fiber type
- wavelength
- remote equipment
- application
subject to complete link validation.
4. Reduce Optical Connector Exposure
Keeping the optical connector outside the liquid can make inspection and cleaning easier.
Optical connector cleanliness is especially important in high-speed fiber links.
Immersible QSFP28 vs Immersion Extender
Both architectures can be useful, but they solve the immersion connectivity problem differently.
| Design | Immersible 100G QSFP28 | Immersion Extender |
|---|---|---|
| Optical module location | Inside immersion environment | Outside immersion environment |
| Direct fluid exposure | Yes | Reduced for external optical module |
| Maintenance access | Inside tank | External / dry side |
| Optical module flexibility | Depends on immersed module | Potentially easier to change |
| Mechanical simplicity | Direct connection | Additional extender required |
| Key engineering focus | Material and fluid compatibility | Signal integrity + material compatibility |
| Best fit | Direct immersed optical architecture | Separated liquid/dry architecture |
The correct design depends on the customer's system.
Optech Can Support Both Architecture Approaches
A major advantage of working with Optech is that customers are not limited to only one immersion connectivity concept.
Optech can support projects based on:
Immersible optical transceivers
or
Immersion Extenders with external optical transceivers
This gives system designers more flexibility to choose an architecture based on:
- cooling-fluid type
- server layout
- maintenance strategy
- cable routing
- optical reach
- equipment density
- network topology
Typical Application Scenario 1: Immersion-Cooled AI Servers
GPU servers are one of the most important applications for immersion cooling.
An AI training cluster may require large numbers of high-speed server-to-switch links.
Optech Immersible QSFP28 or extender-based solutions can support network connectivity while the AI servers remain inside the cooling tank.
Typical Application Scenario 2: HPC and Supercomputing
High-performance computing systems combine:
- high processor density
- high network bandwidth
- intensive storage traffic
- continuous workloads
Immersion cooling can help manage the thermal load, while Optech connectivity products support communication between compute nodes and the network fabric.
Typical Application Scenario 3: GPU Cloud Infrastructure
GPU cloud providers need to deploy large quantities of accelerators while maintaining flexibility.
Immersion cooling can support greater compute density, while modular connectivity solutions can help simplify network upgrades and maintenance.
Typical Application Scenario 4: Research and AI Computing Centers
Universities and research laboratories are increasingly building AI and HPC clusters.
These systems may operate for extended periods under heavy workloads.
Immersion cooling combined with 100G optical connectivity can be a practical option for dense research computing facilities.
Typical Application Scenario 5: High-Density Data Centers
Immersion cooling is also relevant outside pure AI applications.
Cloud platforms, hosting providers, enterprise data centers, and high-density computing environments may all benefit from improved thermal management.
Optech's immersion connectivity solutions can help integrate these systems into existing fiber network infrastructure.
Material Compatibility Is Essential
One of the most important considerations in any immersion project is cooling-fluid compatibility.
Different immersion systems can use different dielectric fluids.
These fluids may interact differently with:
- plastics
- cable jackets
- adhesives
- labels
- elastomers
- coatings
- connector materials
A product suitable for one cooling fluid should not automatically be assumed compatible with another.
Therefore, Optech recommends that customers provide the exact:
coolant manufacturer + coolant model
during project evaluation.
Customization Is a Key Part of Optech's Immersion Strategy
Immersion cooling is rarely a completely standardized environment.
Different customers may have different:
- tank dimensions
- server orientations
- switch positions
- cable exit directions
- coolant types
- optical distances
- connector requirements
For this reason, Optech can provide project-specific customization.
Custom Cable Lengths
Cable length can be optimized for the actual tank dimensions.
This helps reduce unnecessary cable loops and improves system organization.
Custom Mechanical Design
Depending on the project, Optech can evaluate:
- connector housings
- mounting brackets
- pull-tab designs
- strain relief
- cable exit direction
- external mounting structures
Custom Materials
Materials exposed to cooling fluid can be reviewed according to the specified coolant.
Connector Configuration
Optech can evaluate different optical and electrical connector arrangements according to the target equipment.
EEPROM and Compatibility Coding
Where applicable, modules can be configured for compatibility with specific networking platforms.
Custom Labels and Serial Numbers
Customers can request:
- customer P/N
- serial-number formats
- QR codes
- barcodes
- customer logos
- country-of-origin markings
Supporting AI Network Expansion Beyond 100G
The image specifically highlights Immersible 100G QSFP28, but AI networking continues evolving toward higher speeds.
Optech's broader portfolio covers multiple generations of high-speed connectivity, including:
100G → 200G → 400G → 800G → 1.6T
For customers planning next-generation immersion infrastructure, Optech can discuss future high-speed requirements together with current 100G deployments.
This helps customers build a more scalable long-term network architecture.
Why Choose Optech for AI Server Immersion Cooling?
Immersion cooling projects require cooperation between cooling, server, network, optical, and cable engineering.
Optech's value is its ability to support the connectivity portion of that system.
Customers can work with Optech on:
- Immersible 100G QSFP28
- Immersion Liquid Cooling Extenders
- optical transceivers
- high-speed cables
- custom cable lengths
- coolant-related material evaluation
- mechanical customization
- platform compatibility
- EEPROM coding
- labels and serial numbers
- OEM / ODM projects
- sample qualification
- volume manufacturing
- Taiwan-origin products
- TAA options
This allows Optech to support customers from initial design through production deployment.
From Engineering Sample to Volume Order
For a new immersion cooling project, Optech recommends beginning with project evaluation and engineering samples.
A typical process can include:
Requirement Review → Product Selection → Material Review → Custom Design → Engineering Samples → Immersion Testing → Platform Compatibility Testing → Customer Qualification → Volume Production
This helps confirm the actual system configuration before moving to mass deployment.
Taiwan Manufacturing and TAA Options
Country of origin is an important consideration for some government, public-sector, research, education, and regulated infrastructure projects.
Optech is a Taiwan-based optical transceiver and high-speed connectivity manufacturer.
For eligible product configurations, Optech can support Taiwan-origin and TAA-oriented solutions.
Project documentation can be discussed according to customer requirements, including:
- country of origin
- serial-number records
- customized labeling
- product traceability
- manufacturing information
- applicable compliance documentation
Final TAA status should be confirmed according to the specific product configuration, manufacturing process, and procurement requirements.
Ready for Your AI Server Immersion Cooling Project?
AI server cooling and network connectivity increasingly need to be designed together.
Optech provides two practical approaches to immersion connectivity:
Immersible 100G QSFP28 for direct optical connectivity in the immersion environment
and
Immersion Liquid Cooling Extenders for moving the optical module to an external serviceable location.
Combined with Optech's customization capability, Taiwan manufacturing, platform support, and TAA options, these solutions can help customers design reliable high-speed networks for next-generation liquid-cooled AI infrastructure.
Optech welcomes:
- engineering sample requests
- custom immersion projects
- AI server projects
- HPC projects
- coolant compatibility discussions
- OEM / ODM projects
- TAA requirements
- production quotations
- volume orders
Contact Optech today for samples, customization, project evaluation, pricing, and mass-production support.
FAQ
1. What is an Immersible 100G QSFP28?
An Immersible 100G QSFP28 is a 100G optical connectivity solution designed or customized for use in immersion cooling environments where equipment may be exposed to dielectric cooling fluid.
2. Does Optech provide Immersible 100G QSFP28 products?
Yes. The Optech solution shown in the product image includes an Immersible 100G QSFP28 for AI server and immersion cooling applications.
3. What is an Immersion Liquid Cooling Extender?
An Immersion Liquid Cooling Extender is a high-speed extension solution that allows a network port inside the immersion system to be extended toward an external location.
This can allow the actual optical transceiver to remain outside the cooling liquid.
4. Why move the optical transceiver outside the liquid?
Moving the optical transceiver outside the liquid can make module replacement, fiber cleaning, troubleshooting, and optical upgrades easier while reducing direct cooling-fluid exposure.
5. Which design is better: Immersible QSFP28 or an extender?
There is no single correct choice.
An immersible transceiver provides direct connectivity inside the tank, while an extender can improve external access and module flexibility.
The best architecture depends on the tank, coolant, network layout, distance, and maintenance strategy.
6. Can Optech customize the immersion solution?
Yes. Optech can work with customers on cable length, mechanical design, materials, connector configurations, labeling, serial numbers, platform compatibility, and other project-specific requirements.
7. Does coolant type matter?
Yes. Different dielectric fluids may interact differently with plastics, labels, cable jackets, adhesives, and other materials.
Customers should provide the exact coolant manufacturer and model during project evaluation.
8. Can Optech customize the cable length?
Yes. Project-specific lengths can be evaluated according to tank dimensions, equipment location, cable routing, and electrical or optical requirements.
9. Can these products be used in AI server clusters?
Yes. The solutions are relevant to AI servers, GPU clusters, HPC systems, high-density data centers, and other applications where immersion cooling and high-speed networking are used together.
10. Can Optech support higher speeds than 100G?
Optech's broader high-speed connectivity portfolio includes products for 200G, 400G, 800G, and 1.6T-class networking. Specific immersion configurations should be evaluated according to project requirements.
11. Can Optech support Ethernet and InfiniBand environments?
Optech can evaluate connectivity for both Ethernet and InfiniBand applications. The exact host, speed, firmware, coding, FEC, and interface configuration should be confirmed for each project.
12. Can customers request engineering samples?
Yes. Optech welcomes engineering sample and customer qualification projects before volume production.
13. Is TAA available?
Optech can support Taiwan-origin and TAA-oriented options for eligible product configurations. Final TAA status should be confirmed for the specific product and procurement requirement.
14. Does Optech support OEM / ODM projects?
Yes. Optech can support customer part numbers, customized designs, labels, serial numbers, mechanical requirements, cable configurations, and other OEM/ODM needs.
15. Why choose Optech for immersion cooling connectivity?
Optech combines optical transceiver manufacturing, high-speed connectivity products, customization capability, immersion project support, Taiwan manufacturing, sample qualification, and TAA options, allowing customers to work with one supplier from initial development through volume deployment.
Conclusion
As AI servers become more powerful, thermal design and network design are increasingly interconnected.
Immersion cooling can help manage the growing heat density of AI and HPC systems, but reliable high-speed connectivity remains essential.
Optech provides two complementary solutions highlighted in the image:
Immersible 100G QSFP28
and
Immersion Liquid Cooling Extenders
Together, these solutions give customers flexibility to either place the optical connection directly within the immersion environment or move the optical module outside the liquid for easier maintenance and serviceability.
With custom design capability, platform support, Taiwan manufacturing, OEM/ODM services, and TAA options, Optech can support customers from engineering samples and qualification through large-scale deployment.
Optech welcomes AI server immersion cooling projects, customized development, samples, and volume orders worldwide.