Arista's Cash Flow Momentum Builds: How Sustainable is it?

Arista Networks generated $2.78 billion in operating cash flow and approximately $2.69 billion in free cash flow in the first half of 2026, up about 50% year over year. Q2 revenue rose 37.7% to $3.04 billion and non-GAAP operating margin expanded 110bps to 49.9%, supported by AI infrastructure demand and high-Radix switching. Deferred revenue increased sequentially to $6.9 billion from $6.2 billion, while secured memory supply and supplier agreements support cash conversion; inventory and deferred-revenue volatility remain risks. Consensus EPS estimates have risen 11.3% for 2026 and 13.7% for 2027, though Arista's 16.97x forward sales multiple is well above the 4.07x industry average.

Analysis

ANET’s cash conversion is increasingly a financing signal rather than simply an earnings-quality signal: larger customer prepayments reduce working-capital needs and can fund capacity expansion without dilutive capital allocation. The key second-order beneficiary is supplier resilience—secured components and diversified manufacturing should allow ANET to capture urgent AI-networking deployments that slower competitors cannot fulfill. That said, deferred revenue is not equivalent to incremental end-demand; it can pull cash forward and make quarterly operating cash flow materially less informative than revenue, bookings, and remaining performance obligations.

The valuation leaves little room for a modest AI capex digestion. At roughly 17x forward sales, ANET needs sustained revenue growth well above mature-networking peers and continued near-50% operating margins to avoid multiple compression. Over the next 1-3 months, investor focus should shift from cash flow headlines to evidence that hyperscaler orders are broadening beyond a small customer set and that liquid-cooling/network upgrades are additive rather than displacing conventional switching spend. A bookings slowdown, deferred-revenue unwind, or gross-margin pressure from supply commitments would be the earliest falsifiers.

JBL and CLS offer a more cyclical, lower-multiple way to express AI infrastructure buildout, but their working-capital profiles differ sharply. JBL’s improved free-cash-flow outlook can support buybacks and selective capacity investments, while CLS remains more exposed to capex intensity and customer inventory normalization; therefore, not all EMS cash-flow growth should receive ANET-like valuation credit. The contrarian view is that ANET’s premium could persist if high-radix Ethernet becomes the preferred scale-out architecture, but that outcome is already substantially reflected in the share price versus contract manufacturers.

Key Decisions for Investors

• Maintain ANET as a core AI-networking long only on pullbacks or after verification of next-quarter bookings/RPO acceleration; target a 6-12 month holding period. Risk/reward is unattractive chasing strength at the current sales multiple—reduce if revenue growth falls below 25% or non-GAAP operating margin declines below 47%.
• Use a relative-value expression: long ANET / short CSCO over 3-6 months for continued AI Ethernet share gains, but size modestly because CSCO’s enterprise recovery and valuation support can narrow the spread. Exit if ANET order concentration worsens or CSCO demonstrates comparable hyperscaler switch-growth.
• Prefer JBL over CLS for a 6-12 month AI-manufacturing exposure if updated guidance confirms free-cash-flow conversion and capital-return capacity. Treat CLS as a watch item, not a fresh long, until capex intensity declines or backlog converts without further working-capital drag.
• Set an earnings alert for ANET deferred-revenue movement, customer prepayment terms, and inventory days. A sequential decline in deferred revenue combined with elevated inventory would indicate cash-flow normalization and increase downside risk to the premium multiple.