01 Industry overview
The AI networking industry at a glance
AI networking is accelerating with worldwide datacenter Ethernet switching revenue reaching $12.3 billion in the second quarter of 2026, up 64.5% year over year as hyperscalers expand AI infrastructure.
- $12.3B[11]
- Datacenter Ethernet revenue, 2Q26
- 64.5%[11]
- Datacenter switching growth, 2Q26
- 20.4%[11]
- NVIDIA datacenter Ethernet share, 2Q26
- $57B[3]
- AI network fabric spending by 2029
- Market size
- The worldwide datacenter Ethernet switching market reached $12.3 billion in the second quarter of 2026, while the broader Ethernet switch market totaled $18.9 billion. Gartner separately forecasts worldwide AI network fabric spending of $57 billion by 2029.[11][3]
- Forecast & CAGR
- Gartner forecasts AI network fabric spending to reach $57 billion by 2029, representing a 38% five-year CAGR. Dell'Oro forecasts cumulative spending on switches used in AI back-end networks to approach $1 trillion during 2026–2030.[3][12]
- Market share & regions
- In the second quarter of 2026, the Americas generated $10.2 billion of Ethernet switch revenue, compared with $4.6 billion in Asia Pacific and $4.1 billion in EMEA. The Americas grew fastest, at 50.4% year over year, followed by EMEA at 36.9% and Asia Pacific at 35.3%.[11]
- Segmentation
- Datacenter Ethernet switching accounted for $12.3 billion, or 65.2%, of the worldwide Ethernet switch market in the second quarter of 2026; campus and branch switching contributed $6.6 billion. Within datacenter switching, 800GbE represented 41.2% of revenue and 200GbE/400GbE represented 33.6%.[11]
- Key players & concentration
- NVIDIA was the leading branded vendor in datacenter Ethernet switching in the second quarter of 2026, with a 20.4% revenue share and $2.5 billion in revenue. Cisco remained the largest vendor in the overall Ethernet switch market, with $5.4 billion and a 28.7% share.[11]
- Pricing
- IDC reported that higher component prices contributed to Ethernet switch revenue growth outside datacenters in the second quarter of 2026, with non-datacenter revenue increasing 15.7% year over year. The rapid shift to 800GbE, which represented 41.2% of datacenter switching revenue, is also moving spending toward higher-value equipment.[11]
- Trends, drivers & challenges
- AI training and inference expansion is driving demand for high-bandwidth, low-latency switching, with 800GbE accounting for 41.2% of datacenter Ethernet revenue in the second quarter of 2026. Key challenges include supply-chain and component-cost pressure, geopolitical restrictions, and the need to scale power-intensive AI clusters.[11][13]
- Consumer behavior & competition
- Demand is concentrated among hyperscalers, cloud providers, and large enterprises building GPU-dense AI clusters. Gartner expects more than 80% of hyperscalers to prefer Ethernet over proprietary technologies for AI networking by 2028 and more than 80% to use ODM or self-built equipment.[3]
- Regulation
- U.S. export controls remain a material regulatory factor for AI networking because advanced computing systems depend on controlled semiconductors and interconnected data-center infrastructure. In January 2026, the Bureau of Industry and Security moved to case-by-case licensing for certain Nvidia H200 and AMD MI325X exports to China, subject to security and compliance conditions.[13]
- Latest news
- In October 2026, the Ultra Ethernet Consortium released Specification 1.1 for AI and high-performance computing networking. Separately, IDC reported that datacenter Ethernet switching revenue reached $12.3 billion in the second quarter of 2026, with NVIDIA holding the leading branded-vendor position at 20.4%.[14][11]
02 Ranking
Top 10 Companies in the AI Networking Industry
The leading companies in the AI networking industry, ranked on market impact, performance, momentum, innovation and brand popularity.
Updated November 20, 2026 · Next review December 20, 2026
NVIDIA stands out for AI cluster connectivity because it sells Spectrum-X Ethernet, InfiniBand, BlueField DPUs and LinkX connectivity, while Broadcom focuses on switch silicon since it supplies high-density Tomahawk and Jericho switch silicon used in AI Ethernet networks.
Ranked by Overall score · How we score ↓
Editor's top 3 picks
Where to start, by need.
#1 Overall · Leader in DPU and AI NIC
NVIDIA
US
Ranked #1 · 9.7 · DPU and AI NIC
Why we picked it: It sells Spectrum-X Ethernet, InfiniBand, BlueField DPUs and LinkX connectivity for AI clusters, tied to $2.5 billion revenue in 2Q26.
Leader in switch silicon
Broadcom
US
Ranked #2 · 9.3 · Switch silicon
Why we picked it: Its merchant switch silicon supports high-density AI Ethernet switching through Tomahawk and Jericho platforms.
Leader in AI switch platforms
Arista Networks
US
Ranked #3 · 9.3 · AI switch platforms
Why we picked it: Its cloud and AI Ethernet switching shipped 150 million cumulative ports by 2025, with an operating margin of 42.8%.
Pick 1 is the #1 company overall. Picks 2 and 3 are the highest-ranked companies in the top 5 that add a new segment.
Full ranking
| # | Company | Overall ↓ | Impact | Performance | Momentum | Innovation | Company snapshot |
|---|---|---|---|---|---|---|---|
| 1 | NVIDIA DPU and AI NIC · US | 9.7 | 9.9 | 9.9 | 9.9 | 8.4 | |
Company snapshot · NVIDIA · Data as of November 20, 2026 NVIDIA sells Spectrum-X Ethernet, InfiniBand, BlueField DPUs and LinkX connectivity for AI clusters.[15]
| |||||||
| 2 | Broadcom Switch silicon · US | 9.3 | 9.0 | 9.8 | 9.2 | 9.4 | |
Company snapshot · Broadcom · Data as of November 20, 2026 Broadcom supplies high-density Tomahawk and Jericho switch silicon used in AI Ethernet networks.[16]
| |||||||
| 3 | Arista Networks AI switch platforms · US | 9.3 | 9.0 | 9.9 | 9.9 | 9.1 | |
Company snapshot · Arista Networks · Data as of November 20, 2026 Arista reported exceeding its AI-networking goals and shipped 150 million cumulative ports by 2025.[17]
| |||||||
| 4 | Cisco Systems AI switch platforms · US | 9.2 | 9.9 | 8.5 | 8.4 | 9.4 | |
Company snapshot · Cisco Systems · Data as of November 20, 2026 Cisco reported $4 billion of AI-infrastructure revenue in fiscal 2026 and sells AI data-center networking.[18]
| |||||||
| 5 | Marvell Technology Optical interconnect · US | 9.0 | 8.8 | 9.0 | 9.3 | 9.4 | |
Company snapshot · Marvell Technology · Data as of November 20, 2026 Marvell reported $6.1 billion of fiscal 2026 data-center revenue, spanning custom silicon and connectivity.[19]
| |||||||
| 6 | Coherent Optical transceivers and photonic components · US | 8.8 | 8.9 | 9.1 | 9.1 | 8.2 | |
Company snapshot · Coherent · Data as of November 20, 2026 Coherent supplies lasers, transceivers and photonics for AI data-center communications.[20]
| |||||||
| 7 | Advanced Micro Devices AI NICs, DPUs and rack-scale networking · US | 8.8 | 7.7 | 8.8 | 9.9 | 9.9 | |
Company snapshot · Advanced Micro Devices · Data as of November 20, 2026 AMD Pensando provides DPUs and AI NICs for front-end, scale-out and scale-across AI networking.[21]
| |||||||
| 8 | Huawei Technologies Data-center switches and AI Ethernet networks · CN | 8.0 | 7.7 | 7.0 | 7.0 | 9.9 | |
Company snapshot · Huawei Technologies · Data as of November 20, 2026 Huawei is identified among the major companies in the global AI data-center networking market.[22]
| |||||||
| 9 | Hewlett Packard Enterprise AI-native networking and data-center systems · US | 7.6 | 7.7 | 7.2 | 7.3 | 7.0 | |
Company snapshot · Hewlett Packard Enterprise · Data as of November 20, 2026 HPE completed its acquisition of Juniper Networks, doubling the size of its networking business.[23]
| |||||||
| 10 | Extreme Networks AI-native cloud networking and switches · US | 7.5 | 7.0 | 7.4 | 7.8 | 8.9 | |
Company snapshot · Extreme Networks · Data as of November 20, 2026 Extreme markets AI-native cloud networking solutions and reports more than $1.3 billion in annual revenue.[24]
| |||||||
How we score
- Impact (30%): scale and weight in the market (revenue, market share, employees, market position).
- Performance (20%): recent revenue growth and operating margin; an analyst assessment where a company does not publish them.
- Momentum (20%): growth trajectory and future potential.
- Innovation (15%): R&D intensity and how technologically advanced the company is.
Popularity (brand search demand, 15%) counts toward Overall but is not published as a column. Overall leaves out what is not known.
Scale Scores are shown on a 7–10 scale with one decimal; bars start at 7.0.
Cadence Recalculated when new figures are published; next review December 20, 2026.
Independence. Rankings are editorially independent. Scores, ranks, picks and the verdict follow the method above; no company can pay to change them.
Cite this ranking
APA
Samuel Norberg. (2026, November 20). Top 10 companies in the AI networking industry. In AI Networking Industry Statistics. Gitnux. https://gitnux.org/ai-networking-industry-statistics/#top-companies
Revenue of the largest AI networking companies
NVIDIA leads with $215.9B in annual revenue (group level, converted to USD).
03 Trends & forecasts
Top 10 AI Networking Industry Trends & Forecasts 2027
What is changing in the AI networking industry, with the evidence behind each trend and a forecast for 2027.
Forecasts were generated by an AI model from the cited evidence. They are estimates, not guarantees.
01AI fabric spending accelerates through 2027
High confidenceGartner forecasts worldwide AI network-fabric spending will reach $57 billion by 2029, implying a 38% five-year CAGR. Dell’Oro separately expects cumulative spending on switches for AI back-end networks to approach $1 trillion during 2026–2030.[3][12]
Forecast 2027: AI networking will remain one of the fastest-growing data-center infrastructure categories in 2027, with spending growth continuing well above conventional enterprise switching. The market will increasingly be budgeted as a core component of accelerated-computing systems rather than as a separate network refresh.
02Datacenter switching outpaces campus switching
High confidenceWorldwide datacenter Ethernet-switch revenue reached $12.3 billion in 2Q26, up 64.5% year over year, while campus and branch revenue reached $6.6 billion, up 15.7%. Datacenter switching therefore represented 65.2% of the total Ethernet-switch market in the quarter.[11]
Forecast 2027: In 2027, datacenter switching will continue to capture most incremental Ethernet spending as hyperscalers and AI-native cloud providers expand GPU clusters. Campus and branch demand should grow more slowly and remain more dependent on general enterprise refresh cycles.
03800GbE becomes the volume frontier
High confidence800GbE accounted for 41.2% of datacenter Ethernet-switch revenue in 2Q26, reflecting the rapid migration toward higher-bandwidth AI fabrics. Dell’Oro expects 1.6Tbps ports to become the majority of AI back-end ports by 2027.[11][12]
Forecast 2027: 800GbE will remain a major revenue category in 2027, but 1.6Tbps designs will move from early deployments toward mainstream AI-cluster procurement. Suppliers that can deliver high-density 800GbE and upgrade paths to 1.6Tbps should gain preference.
04Scale-up networking takes a larger share
Medium confidenceDell’Oro forecasts that scale-up networking will account for more than half of the AI back-end switch market by 2030. The Ultra Ethernet Consortium’s October 2026 Specification 1.1 also adds multiple improvements for scale-up networking.[12][14]
Forecast 2027: Scale-up architectures will gain share in 2027 as clusters connect larger groups of accelerators and require tighter synchronization. This will increase demand for specialized fabrics, congestion management and high-radix switches in addition to conventional scale-out networks.
05Ethernet gains ground against proprietary fabrics
High confidenceGartner expects more than 80% of hyperscalers to prefer Ethernet over proprietary technologies for AI networking by 2028, citing economies of scale and a broader ecosystem. The Ultra Ethernet Consortium released Specification 1.1 in October 2026 to improve Ethernet for AI and HPC workloads.[3][14]
Forecast 2027: Ethernet adoption in AI back-end networks will broaden in 2027, particularly among hyperscalers seeking interoperable supply and lower platform dependence. Proprietary interconnects will remain important in selected tightly integrated systems, but their share of new deployments should decline.
06Hyperscalers internalize network design
High confidenceGartner forecasts that more than 80% of hyperscalers will use ODM or self-built equipment for AI networking by 2028, up from only a handful currently. This reflects the scale and customization requirements of large AI clusters.[3]
Forecast 2027: In 2027, ODM and self-designed systems will expand their role in hyperscale AI deployments, especially for high-volume switch, optical and rack-level integration. Branded vendors will increasingly compete through silicon, software, optics and system integration rather than only through complete proprietary platforms.
07NVIDIA consolidates AI networking leadership
High confidenceNVIDIA held a 20.4% share of datacenter Ethernet-switch revenue in 2Q26, with $2.5 billion in revenue and 181.1% year-over-year growth. Its Spectrum-X platform combines Ethernet switches, DPUs and LinkX connectivity for GPU clusters.[11]
Forecast 2027: NVIDIA will remain a leading AI-networking platform supplier in 2027 as customers favor integrated compute-and-network designs. Competitive pressure will increase from Arista, Cisco, Broadcom-based systems and hyperscaler custom designs, limiting the likelihood of an uncontested market.
08Competition shifts toward integrated platforms
High confidenceGartner says providers that bridge networking and compute processing will take a significant share of AI-network-fabric spending. NVIDIA’s 2Q26 offering illustrates this shift by combining switches, DPUs and cables into an AI-networking platform.[3][11]
Forecast 2027: In 2027, product positioning will move further from standalone switches toward validated AI fabrics that include network operating systems, telemetry, optics, DPUs and cluster-management software. Vendors with tested end-to-end reference architectures should win a larger share of high-value deployments.
09Congestion control becomes a differentiator
High confidenceUltra Ethernet Specification 1.1 adds congestion signaling, programmable congestion management, improved transport functions and in-network collectives. These capabilities target the synchronization, telemetry and traffic-burst requirements of AI and HPC workloads.[14]
Forecast 2027: In 2027, buyers will evaluate AI switches increasingly on tail latency, congestion recovery, telemetry and collective-communication performance rather than on port speed alone. Software-defined congestion management will become a more visible source of vendor differentiation.
10Power availability constrains cluster expansion
High confidenceGartner forecasts worldwide data-center electricity consumption of 565 TWh in 2026, up 26% from 2025, and expects AI-optimized server power consumption to surpass conventional-server consumption in 2027. Gartner also forecasts that data-center power demand will reach 702 TWh in 2027.[3]
Forecast 2027: Power availability will increasingly influence where AI networking equipment is deployed in 2027. Network suppliers will face stronger demand for energy-efficient switch silicon, higher port density, optical integration and designs that reduce power per delivered bit.
04 Statistics
Market Intelligence
Market-intelligence figures track multiple AI networking scopes, including $12.3B datacenter Ethernet switching in 2Q26 and forecasted AI network fabric spending of $57B by 2029.
Market size and revenue in the AI networking industry
- USD 3.4895 billion was the U.S. 2024 AI-in-networks market value, covering the broader AI-in-networks category in which software is a named offering segment.[1]
- USD 10.93 billion was the global 2024 AI-in-networks market size, covering router and switches, AI networking platforms, management software and software-defined networking; the publisher projects USD 46.87 billion in 2029, but only the 2024 level is reported here.[1]
- USD 7.12 billion was the global 2025 AI-in-network market value, covering networking vendors, telecommunications providers and software specialists developing network intelligence and autonomous operations.[2]
Growth forecasts for the AI networking industry
- USD 28.10 billion is projected for the global AI-in-network market in 2034, representing a 16.49% CAGR from 2026 to 2034; the scope includes software specialists, networking vendors and telecommunications providers.[2]
- USD 46.87 billion is projected for the global AI-in-networks market in 2029, representing a 33.8% CAGR from 2024 to 2029; the scope includes AI networking platforms and management software alongside hardware.[1]
- USD 11.33 billion is projected for worldwide AI network fabric spending in 2028, implying a 27.1% CAGR through 2028; the scope is hyperscaler AI networking infrastructure rather than software alone.[3]
05 Statistics
Strategy & Growth
Strategy and growth data show a high volume of AI-infrastructure and networking deals, including 115 acquisitions tracked from January 2024 through September 27, 2026.
M&A and consolidation in the AI networking industry
- 23 horizontal acquisitions represented 0.3% of 11,372 AI-ecosystem majority acquisitions from January 2020 through August 2026; all involved compute, data-center or physical-infrastructure companies, a parent-category proxy containing AI networking.[4]
- $14.3 billion was the announced enterprise value of HPE’s January 2024 acquisition of Juniper Networks, a networking-products provider whose portfolio supports AI platforms; United States transaction.[5]
- 115 acquisitions were tracked, including 60 closed deals, across AI infrastructure, networking, data centers, semiconductors and related systems from January 2024 through September 27, 2026; this defined parent-category sample contains AI networking.[6]
- $274.5 billion of disclosed value covered 43 U.S.-dollar deals in the same 115-deal AI-infrastructure and networking sample through September 27, 2026; this is a parent-category aggregate containing AI networking.[6]
06 Statistics
Pricing & Packaging
Pricing and packaging data document subscription levels for AI agents and network-operations platforms, including $299 per month for Netverge’s starter tier.
Price levels in the AI networking industry
- $100 per month was the published U.S. price for each additional block of 10 Netverge AI agents, observed in November 2026.[7]
- $299 per month was the published U.S. starter price for Netverge’s AI-enabled network-operations platform, including 10 AI agents, on the pricing page observed in November 2026.[7]
- $2,708 per month was Plixer One’s published starting price when billed annually in the United States, observed in November 2026.[8]
- $1,083 per month was Plixer Scrutinizer’s published starting price when billed annually in the United States, observed in November 2026.[8]
07 Statistics
Finance
Finance figures quantify funding into broader data-center networking startups, including $410 million raised in North America in 2Q26.
Funding and investment in the AI networking industry
- $410 million was raised by data-center networking startups in North America in the second quarter of 2026; the dataset includes AI-networking software companies within its broader data-center-networking scope.[9]
- $125 million was raised by Aria Networks in its first disclosed funding round in April 2026 to develop AI-native networking infrastructure for data centers.[10]
- $233 million was raised by data-center networking startups in Europe in the third quarter of 2026; the dataset includes AI-networking software companies within its broader data-center-networking scope.[9]
Key questions about the AI networking industry
What is the 2026 datacenter Ethernet switching market size?
How fast is AI networking growing in switching and fabric spending?
Which companies lead the AI networking industry?
Which throughput tier is becoming the volume frontier?
How is the market structured across regions in 2Q26?
What regulatory factor affects AI networking hardware provisioning?
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
Samuel Norberg. (2026, November 20). AI Networking Industry Statistics. Gitnux. https://gitnux.org/ai-networking-industry-statistics
Samuel Norberg. "AI Networking Industry Statistics." Gitnux, 20 Nov 2026, https://gitnux.org/ai-networking-industry-statistics.
Samuel Norberg. 2026. "AI Networking Industry Statistics." Gitnux. https://gitnux.org/ai-networking-industry-statistics.
Sources & references
24 sources cited across this report
marketsandmarkets.com
polarismarketresearch.com
gartner.com
bcg.com
carlsquare.com
- convergedigest.com
- netverge.com
plixer.com
aperturetracker.com
streetinsider.com
idc.com
delloro.com
bis.gov
- ultraethernet.org
investor.nvidia.com
spglobal.com
investors.arista.com
investor.cisco.com
investor.marvell.com
ir.coherent.com
amd.com
thebusinessresearchcompany.com
hpe.com
extremenetworks.com
