data center networking and covered company earnings previews: AI data-center networking outlook rises as Goldman Sachs previews ANET, CLS and FFIV earnings

Overview

The report updates Goldman Sachs' networking-industry model using 2Q26 data from 650 Group and previews upcoming earnings for Arista Networks, Celestica and F5. Its central view is that AI data-center build complexity is increasing networking content, while campus-network modernization and Wi-Fi 7 also improve the spending outlook.

Core views

Goldman Sachs raises its networking outlook following updated 2Q26 data from 650 Group. It now expects AI Ethernet data-center switching spending to reach roughly $100 billion in 2030, up at a 65% CAGR from 2025, supported by demand from hyperscalers, tier-2 cloud providers and enterprises. The report attributes this to increasingly complex AI data-center builds, where high-performance, low-latency traffic requires greater networking attachment. Across AI data centers, the report forecasts 2030 AI server spending of about $1.3 trillion and AI Ethernet switching spending of $100.5 billion, versus $195.1 billion and $8.1 billion, respectively, in 2025. It also notes that annual 2026-30 AI-server forecasts were raised by about 4% on average. Campus forecasts were also increased. Goldman Sachs lifted its 2026-30 WLAN outlook by roughly 6% on average to approximately $14.3 billion in 2030, from $13.4 billion previously, principally because of higher Wi-Fi 7 expectations. Enterprise-switching estimates rose roughly 8% on average to about $30 billion by 2030, from $28 billion. The report links these revisions to network modernization for higher traffic flows and, in part, memory-price increases. Supporting current trends, enterprise indoor WLAN revenue grew 24% year on year to about $2.6 billion in 2Q26 and is expected to grow 18% in 2026; Wi-Fi 7 represented 66% of enterprise WLAN revenue, up from 62% in 1Q26. Enterprise switching grew 20% year on year in 2Q26. For Arista Networks, Goldman Sachs forecasts 3Q26 EPS of $1.11, above Visible Alpha consensus of $1.08 and guidance of $1.06-$1.08. Revenue is forecast at $3.48 billion versus $3.33 billion consensus and $3.3 billion guidance, while EBIT of $1.71 billion implies a 49.2% margin, above 48.8% consensus and within the 48-49% guidance range. The report sees sustained AI switching demand, improved fulfillment as Arista secures component supply, software differentiation through EOS and a single-stack architecture, and continued expense discipline. Arista had $9.7 billion of purchase commitments in 2Q26, up $0.8 billion sequentially, and $6.9 billion of deferred revenue. Goldman Sachs reiterates Buy and its $225 12-month target, based on 36x NTM+1-year EPS. For Celestica, Goldman Sachs expects 3Q26 revenue of $5.36 billion, broadly in line with $5.38 billion consensus, and adjusted EBIT of $450 million at an 8.4% margin. It reduces 3Q26 non-GAAP EPS to $2.85 from $3.02 because of the higher share count after the approximately $3 billion equity offering, whose proceeds are intended for working capital and capital expenditure supporting AI compute and data-center Ethernet demand. The report forecasts CCS communications revenue up 60% year on year and enterprise revenue up 189%, driven by hyperscaler demand, 800G switching and initial 1.6T program production. It expects ten awarded 1.6T programs to begin mass-production ramps in 3Q26, with a more meaningful ramp in 2027, and views the planned 2027 AMD Helios scale-up switch ramp as a multi-billion-dollar opportunity. Goldman Sachs expects operating leverage, a higher-value content mix and improving ATS profitability to expand margins; its 2027/28 EPS forecasts of $20.83/$32.90 exceed consensus of $19.28/$26.73. It remains Buy rated, but lowers its target to $480 from $505 while retaining an 18x NTM+1-year EPS multiple. For F5, Goldman Sachs forecasts F4Q26 revenue of $894 million, above $884 million consensus and the $870-$890 million guidance range. It expects non-GAAP operating income of $311 million, a 34.7% margin, and non-GAAP gross profit of $747 million, an 83.5% gross margin, aided by a favorable high-performance systems mix and lower-than-anticipated component-cost increases. The report expects Systems revenue to benefit from the iSeries refresh, capacity expansion and AI-driven traffic, while Software growth should accelerate in FY2027 as legacy SaaS and managed-service transitions conclude, renewals strengthen and Distributed Cloud Services expand. Direct AI use cases have exceeded $50 million of revenue. Goldman Sachs sees upside to FY2027 gross-margin guidance of 80-82%, with its estimate at 82.5%, if component availability is better than expected, but retains its Neutral rating and $400 target based on 20x NTM+1-year EPS. The market-share data show a competitive and evolving landscape. In front-end AI Ethernet networking, Celestica held 33%, Arista 25% and Cisco 24%; Cisco was also gaining share in back-end AI networking. Cisco led AI DCI routing with 48%, while Cisco and Arista led traditional data-center switching at 22% and 20%, respectively. These share trends frame both the opportunity from AI-networking growth and the competitive risks faced by covered companies.

Analysis framework

Goldman Sachs first updates sector demand forecasts using 650 Group market data, then evaluates market-share trends across AI, traditional data-center and campus networking. It applies those demand and competitive assumptions to company earnings forecasts, comparing projections with management guidance and Visible Alpha consensus, and values covered companies using forward NTM+1-year EPS multiples.

Methodology notes

• Industry AnalysisSupply-demand framework

  Data-center and campus-networking market forecasts by customer vertical, technology and spending category.

  The report uses projected spending and growth by hyperscalers, tier-2 cloud providers and enterprises to explain demand for networking equipment and the resulting company opportunities.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI infrastructure build complexity increases networking attachment and demand for switching, servers and related components.

  The report connects AI data-center investment and traffic requirements to demand for networking equipment, component availability and supplier revenue.

• Valuation methodsP/E and PEG Valuation

  Forward NTM+1-year EPS multiples for ANET, CLS, HPE and FFIV.

  Goldman Sachs derives stated 12-month targets by applying specified forward EPS multiples to each covered company.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Arista Networks (ANET)

  Covered AI data-center networking beneficiary

  Strengths

  AI switching exposure, secured memory and component supply, EOS software and single-stack architecture, and expense discipline.

  Comparison

  Arista held 25% front-end AI Ethernet networking market share, versus Celestica at 33% and Cisco at 24%.

  Risks

  Slower cloud capex, customer concentration, competition, supply-chain costs, commoditization and enterprise-strategy execution.

• Celestica (CLS)

  Covered manufacturing and infrastructure beneficiary

  Strengths

  800G and 1.6T program ramps, hyperscaler AI/ML compute exposure, improving demand visibility and expected operating leverage.

  Weaknesses

  Higher share count following the roughly $3 billion equity offering reduces near-term EPS.

  Comparison

  Celestica held 33% front-end AI Ethernet networking market share and is expected to benefit from high-speed networking share gains.

  Risks

  Competition, AI-server margin pressure, customer concentration, weaker data-center demand, fewer higher-margin design wins and ATS demand volatility.

• F5 (FFIV)

  Covered application delivery, systems and security provider

  Strengths

  iSeries refresh, AI-related traffic demand, Distributed Cloud Services and potential gross-margin upside from component availability.

  Weaknesses

  Software growth needs to reaccelerate and the stock remains Neutral rated.

  Risks

  Slower enterprise IT spending, supply-chain cost pressure and execution risk in cloud-based SaaS distribution.

• Hewlett Packard Enterprise (HPE)

  Covered networking and infrastructure company

  Strengths

  Exposure to enterprise and sovereign AI data-center infrastructure.

  Comparison

  HPE follows Cisco in AI DCI routing market share and is a leading vendor in non-AI DCI routing and enterprise WLAN.

  Risks

  Weaker corporate IT and data-center spending, white-box competition, integration-related churn, higher component costs and weaker AI infrastructure demand.

Key data

• AI Ethernet data-center switching market~$100 billion in 203065% CAGR from 2025; driven by broader AI-networking demand.
• WLAN market outlook~$14.3 billion in 2030Raised by ~6% on average from ~$13.4 billion previously.
• Enterprise switching market outlook~$30 billion in 2030Raised by ~8% on average from $28 billion previously.
• ANET 3Q26 EPS forecast$1.11Versus $1.08 Visible Alpha consensus and $1.06-$1.08 guidance.
• CLS 3Q26 revenue forecast$5.36 billionVersus $5.38 billion consensus.
• FFIV F4Q26 revenue forecast$894 millionVersus $884 million consensus and $870-$890 million guidance.

Impact & implications

The report argues that AI infrastructure spending should continue to favor networking vendors and manufacturing partners with high-speed switching exposure, component availability and execution capability. It identifies ANET and CLS as positioned to benefit from this trend, while FFIV's outlook depends more on the durability of its refresh cycle, software reacceleration and gross-margin trajectory.

Risks

• For ANET, slower cloud capital expenditure, customer concentration, competition, supply-chain costs, hardware commoditization and enterprise-strategy execution are stated downside risks.
• For CLS, competition, AI-server margin pressure, high customer concentration, lower data-center investment, fewer higher-margin design wins and ATS demand volatility are stated risks.
• For HPE, weaker IT and data-center demand, white-box competition, Juniper-integration churn, higher component costs and weaker enterprise or sovereign AI demand are stated risks.
• For FFIV, slower enterprise IT spending, supply-chain cost pressure and cloud/SaaS platform execution are stated risks.

What to watch

• AI Ethernet switching and AI-server market forecast revisions from 650 Group.
• ANET's component availability, AI revenue growth, scale-across contribution and margin execution.
• CLS's 800G and 1.6T program ramps, AMD Helios opportunity, component availability and post-offering margin expansion.
• FFIV's initial FY2027 outlook, Systems refresh durability, Software-growth reacceleration and gross-margin guidance.