Global Ethernet Switch Market Hits $18.9B in Q2 2026
If AI computing clusters are the factories of the AI era, Ethernet switches and routers are the highways, traffic hubs, and cross-city gateways that connect them.
According to the latest IDC data, the global network infrastructure market accelerated sharply in the second quarter of 2026. The global Ethernet switch market reached $18.9 billion, approximately RMB 126.6 billion, representing 43.4% year-over-year growth. The global router market reached $4.5 billion, approximately RMB 30.1 billion, up 22.8% year over year.
The strongest growth came from datacenter Ethernet switches, where revenue surged 64.5% year over year. The rapid construction of AI computing clusters is pushing datacenter networking toward higher bandwidth, lower latency, and increasingly dense 800GbE deployments.
At the same time, enterprise campus networks are entering a new hardware refresh cycle driven by Wi-Fi 7, security requirements, higher bandwidth, and AI-enabled operations.
🌐 Ethernet Switch Market Reaches $18.9 Billion #
In Q2 2026, the global Ethernet switch market reached $18.9 billion, up 43.4% year over year and maintaining the strong growth momentum seen in the previous quarter.
The market is increasingly divided into two major growth tracks:
- Datacenter networking, driven by AI training and inference infrastructure.
- Enterprise campus and branch networking, driven by hardware replacement and intelligent network upgrades.
Datacenter Networking: AI Pushes 800G Into the Mainstream #
The global datacenter Ethernet switch market reached $12.3 billion in Q2 2026, representing a 64.5% year-over-year increase.
The primary driver is the rapid expansion of AI training and inference clusters operated by hyperscalers, cloud providers, and large technology companies.
Large GPU clusters generate enormous amounts of east-west traffic. GPUs and servers must continuously exchange model parameters, activation data, gradients, KV-cache-related information, and other intermediate data. As the number of accelerators in a cluster increases, network bandwidth and latency increasingly become system-level performance constraints.
This is pushing datacenter networks toward much higher port speeds.
According to the reported IDC figures, 800GbE port equipment accounted for 41.2% of datacenter Ethernet switch revenue in Q2. Another 33.6% came from 200G and 400G equipment.
Together, these high-speed segments represented nearly 75% of datacenter switch revenue.
This shift is significant because networking is no longer simply an infrastructure layer surrounding the GPU cluster. For large distributed AI systems, the switching fabric is increasingly part of the computing architecture itself.
AI infrastructure therefore requires not only more GPUs, but also:
- Higher switch radix and port density
- 400G and 800G Ethernet connectivity
- Lower network latency
- Higher east-west bandwidth
- Advanced congestion control
- High-speed optical and copper interconnects
- More sophisticated telemetry and network management
As AI clusters become larger, network architecture becomes increasingly important to overall accelerator utilization.
Enterprise Networking: A New Hardware Refresh Cycle #
The non-datacenter Ethernet switch market, covering enterprise campus and branch environments, reached $6.6 billion in Q2 2026, up 15.7% year over year.
Unlike the datacenter segment, this growth is less dependent on AI cluster construction and more closely associated with enterprise infrastructure refresh cycles.
One driver is the transition toward Wi-Fi 7. Existing switching infrastructure in many organizations was not designed around the bandwidth and traffic characteristics of newer wireless networks. As wireless access points become faster, the wired network connecting them must also provide sufficient uplink capacity.
Security is another major driver.
Modern enterprise switches increasingly participate in broader networking and security architectures, including:
- AI-assisted threat detection
- Automated network operations
- Zero-trust architectures
- Higher-speed encrypted traffic
- Post-quantum cryptography preparation
- Integrated network telemetry
The result is a transition from the traditional model of a switch as a basic packet-forwarding device toward a more software-defined and security-aware network infrastructure platform.
Component costs also affect market revenue. Higher prices for memory and other key components can increase average system prices, allowing revenue growth to outpace unit shipments.
🌎 Regional Ethernet Switch Market Performance #
All three major geographic regions recorded positive year-over-year growth in Q2 2026.
| Region | Q2 2026 Revenue | YoY Growth |
|---|---|---|
| Americas | $10.2B | 50.4% |
| EMEA | $4.1B | 36.9% |
| Asia-Pacific | $4.6B | 35.3% |
| Global | $18.9B | 43.4% |
The Americas remained the largest market, accounting for more than half of global quarterly Ethernet switch revenue.
The region’s growth was strongly associated with North American hyperscale AI infrastructure investment, alongside continued enterprise digital transformation.
EMEA and Asia-Pacific also posted strong growth, supported by datacenter expansion, enterprise infrastructure upgrades, and networking demand in emerging markets.
🚦 Global Router Market Reaches $4.5 Billion #
The global router market generated $4.5 billion in Q2 2026, up 22.8% year over year.
The router market differs from the Ethernet switch market in one important respect: its growth is more evenly distributed across service providers and enterprises.
Key drivers include:
- 5G network expansion
- Backbone and core-network upgrades
- Cloud-network integration
- Datacenter interconnects
- Edge computing
- SD-WAN adoption
- Enterprise WAN modernization
Service Providers Remain the Core Router Market #
Telecom operators and cloud service providers generated approximately $3.3 billion in router revenue during the quarter, representing 74.2% of the global router market.
The segment grew 23.8% year over year.
Two major infrastructure trends are supporting this growth.
First, operators continue to expand 5G networks and modernize their backhaul and core infrastructure.
Second, cloud computing and edge computing are creating more demanding connectivity requirements between datacenters, metropolitan networks, and distributed edge nodes.
As traffic volumes increase, backbone routing systems must provide higher throughput, reliability, and scalability.
Enterprise Routers Move Toward SD-WAN and AI Edge Networking #
Enterprise router revenue reached $1.2 billion in Q2 2026, up 20.1% year over year.
Traditional enterprise WAN architectures are increasingly being replaced or supplemented by SD-WAN, which provides more centralized control, application-aware routing, and flexible use of multiple network links.
At the same time, enterprise branches are beginning to host more AI-assisted applications and edge workloads.
This creates new requirements for branch networking infrastructure, including higher throughput, better security, local processing, and more intelligent traffic management.
Regional Router Performance #
| Region | Q2 2026 Revenue | YoY Growth |
|---|---|---|
| Americas | $1.9B | 31.1% |
| EMEA | $1.1B | 20.9% |
| Asia-Pacific | $1.5B | 15.2% |
| Global | $4.5B | 22.8% |
The Americas again recorded the highest growth rate at 31.1%, followed by EMEA at 20.9% and Asia-Pacific at 15.2%.
Unlike the switch market, where AI datacenter spending is the dominant growth engine, router demand remains closely connected to broader telecommunications and enterprise networking investment.
🏢 Leading Ethernet Switch Vendors #
The competitive landscape continued to evolve in Q2 2026. Traditional networking vendors retained strong positions across broad enterprise and service-provider markets, while vendors focused heavily on AI datacenter networking recorded particularly rapid growth.
Cisco #
Cisco remained the largest Ethernet switch vendor in the reported Q2 figures.
Its Ethernet switch revenue reached $5.4 billion, up 36.2% year over year, corresponding to a 28.7% global market share.
Cisco generated approximately $2.2 billion from the datacenter switch segment, benefiting from continued AI infrastructure investment.
Its router business remained substantial as well, reaching $1.6 billion in quarterly revenue, up 31.6% year over year.
Cisco’s broad portfolio gives it exposure across datacenters, enterprise campuses, branches, and service-provider networks.
NVIDIA #
NVIDIA recorded the fastest growth among the major vendors discussed in the reported figures.
Its switch revenue reached $2.5 billion, representing 181.1% year-over-year growth.
Because NVIDIA’s reported switch business is concentrated in datacenter networking, the company’s revenue corresponds to approximately 20.4% of the datacenter Ethernet switch market.
NVIDIA’s strategy is closely tied to the architecture of large AI clusters. Rather than treating networking as an isolated product category, the company combines switching with technologies such as Spectrum-X, BlueField DPUs/SmartNICs, and high-speed interconnects.
This allows networking to be positioned as part of an integrated AI infrastructure stack.
Arista #
Arista Networks generated $2.5 billion in switch revenue during Q2, up 37.6% year over year.
Its reported global switch market share was 13.4%, while its share of the datacenter segment reached 18.7%.
Arista has a particularly strong position in high-performance datacenter networking, with substantial exposure to 400G and 800G Ethernet deployments.
Its concentration on cloud and datacenter customers makes the company’s growth closely connected to hyperscale infrastructure investment.
Huawei #
Huawei generated approximately $1.5 billion in Ethernet switch revenue in Q2, up 28.6% year over year, representing an 8.2% global market share.
Its router business reached $1.3 billion, up 21.6% year over year, corresponding to a 29.9% global router market share in the reported figures.
Huawei’s networking business remains strongly exposed to telecommunications operators, government and enterprise infrastructure, and markets where its broader networking portfolio is deployed.
HPE #
Following its acquisition of Juniper Networks, HPE continued to expand its networking portfolio across enterprise campus and branch environments.
Its switch revenue grew 1.6% year over year, with a reported 6.1% market share. More than 70% of its switch revenue came from enterprise-oriented business according to the supplied IDC figures.
HPE’s router revenue reached approximately $483 million, up 25.3% year over year, giving it a 10.7% global router market share.
The Juniper combination broadens HPE’s networking footprint across enterprise, datacenter, and service-provider markets.
⚡ Three Signals Shaping Network Infrastructure #
The Q2 2026 figures reveal three broader trends across the networking industry.
Signal 1: AI Is Reshaping Datacenter Network Spending #
AI cluster construction has become one of the most important drivers of datacenter networking expenditure.
Large-scale distributed training and inference create enormous east-west traffic between GPUs and servers. As cluster sizes increase, network bandwidth and latency increasingly influence accelerator utilization.
This creates a direct connection between AI compute investment and Ethernet infrastructure spending.
The networking fabric is becoming part of the AI compute system rather than simply an external connectivity layer.
Signal 2: 800G Is Accelerating #
The reported revenue share of 800G equipment in datacenter Ethernet switches increased from 35.9% in Q1 2026 to 41.2% in Q2.
That change indicates a rapid migration toward next-generation port speeds among large datacenter customers.
For networking vendors, this creates both an opportunity and a challenge.
The opportunity is a rapidly expanding high-end market. The challenge is that technology cycles become shorter as customers move more quickly from one Ethernet generation to the next.
The transition from 400G to 800G also affects the broader infrastructure stack, including:
- Optical transceivers
- DAC and AEC cables
- Network adapters
- Switch ASICs
- Network operating systems
- Congestion-control mechanisms
- Datacenter power and cooling
As a result, 800G adoption is not simply a switch-port upgrade. It represents a broader infrastructure transition.
Signal 3: Enterprise Networks Are Entering Another Upgrade Cycle #
Enterprise networks are also changing.
Campus and branch infrastructure increasingly needs to support Wi-Fi 7, higher traffic volumes, cloud applications, AI-assisted workloads, and integrated security.
The traditional separation between networking and security is becoming less distinct. Network infrastructure is increasingly expected to provide visibility, automated response, policy enforcement, and security functions alongside packet forwarding.
This creates another upgrade cycle independent of the AI datacenter boom.
🔭 From AI Factories to AI Networks #
The Q2 2026 networking market illustrates a broader change in how infrastructure should be viewed.
AI computing does not scale through GPUs alone.
A large AI cluster requires a complete infrastructure stack:
GPU → HBM → Server → NIC/DPU → Ethernet Switch → Optical Interconnect → Datacenter Network
Every layer can become a bottleneck.
As accelerator density increases, the network must scale accordingly. A cluster containing thousands or tens of thousands of GPUs cannot realize its full computing potential if the interconnect cannot move data quickly enough between them.
That is why the rapid growth of the Ethernet switch market is closely connected to the AI infrastructure cycle.
At the same time, enterprise networks are undergoing a separate transformation driven by Wi-Fi 7, security, SD-WAN, cloud connectivity, and edge computing.
The result is a networking market being pulled forward by three simultaneous forces:
- AI infrastructure expansion
- High-speed Ethernet migration toward 800G
- Enterprise networking and security modernization
The global Ethernet switch market reached $18.9 billion in Q2 2026, while the router market reached $4.5 billion. The numbers show that networking is becoming an increasingly important component of the AI infrastructure investment cycle.
The next phase will not simply be about connecting more servers. It will be about building networks capable of keeping increasingly large AI clusters synchronized, efficiently utilized, and scalable.