Arista Q2 Earnings Beat and Significantly Raised Guidance; Executives State "Demand Is Not the Issue, Supply Is the Main Bottleneck"

U.S. cloud networking and switch giant Arista Networks reported second-quarter results that comprehensively exceeded expectations, with quarterly revenue breaking through the $3 billion mark for the first time. Full-year revenue guidance was significantly raised to $12.6 billion, further confirming that the wave of AI infrastructure investment is accelerating its penetration into the network layer.

The financial report showed that the company's Q2 revenue increased by 38% year-over-year to $3.04 billion, surpassing the $3 billion threshold for a single quarter for the first time. Adjusted earnings per share (EPS) were $1.02, both significantly exceeding Wall Street expectations. Of greater concern to the market was Arista's raising of its fiscal 2026 revenue guidance to approximately $12.6 billion, which is $1.1 billion higher than previous expectations, corresponding to a year-over-year growth rate of about 40%.

During the conference call, management conveyed several key signals: there is no doubt on the demand side; the real constraint on revenue realization is the supply chain; the additional $1.1 billion in full-year guidance does not rely solely on AI backend networks, with front-end data centers, campus networks, and routing businesses all contributing; supply chain tightness may persist until 2028, and the company is enhancing delivery certainty through multi-year purchase commitments and expanding manufacturing capacity.

CEO Jayshree Ullal stated in a release: "Customers view the network as the central nervous system of infrastructure, spanning from clients to campuses, and then to data centers and AI hubs." Following the announcement, Arista's stock price surged 12% to $213 in after-hours trading, bringing its year-to-date cumulative gain to over 45%, with its market capitalization poised to break through $250 billion.

Quarterly Revenue Breaks $3 Billion for the First Time, Profitability Improves Simultaneously

Arista is an industry leader in high-end networking equipment (switches) for cloud computing and data centers, and a major competitor to Cisco. Its Q2 revenue was $3.04 billion, a 37.7% year-over-year increase, higher than the midpoint of the previous $2.8 billion guidance and exceeding Wall Street's expectation of $2.83 billion. This figure is symbolic—its full-year revenue in 2021 was approximately $2.9 billion, meaning a single quarter now exceeds the entire year's scale of that period.

Profitability remained robust. The non-GAAP gross margin in Q2 was 63.4%, lower than 65.6% in the same period last year, but rebounding from 62.4% in the previous quarter; the adjusted operating profit margin reached 49.9%, higher than 47.8% in Q1 and 48.8% in the same period last year; net income was $1.3 billion, with a net profit margin of 42.9%; diluted EPS was $1.02, a 39.7% year-over-year increase, higher than the analyst consensus of 89 cents.

Q3 guidance was also strong: revenue of approximately $3.3 billion, non-GAAP gross margin of about 63%, operating profit margin of 48% to 49%, and diluted EPS of $1.06 to $1.08, all exceeding analysts' previous expectations of $2.95 billion in revenue and 92 cents in EPS.

Full-Year Guidance Raised by $1.1 Billion, Growth Sources Broader Than Market Expectations

During the conference call, Evercore analyst Amit Daryanani directly pointed out the core proposition: AI networks are becoming more complex across multiple dimensions simultaneously, not just with larger cluster scales, but also requiring connections between campuses and multiple data centers. Management's response confirmed this judgment and outlined a broader growth path.

Arista raised its fiscal 2026 revenue guidance to approximately $12.6 billion, which is $1.1 billion higher than the $11.5 billion expectation given in May, and $2.1 billion higher than the previous Analyst Day target of $10.5 billion.

During the call, management noted that this $1.1 billion increase in guidance was not simply attributed to an explosion in AI orders: the annual target for AI fabric is at least $3.5 billion, the campus network target is at least $1.25 billion, and front-end data centers, enterprise customers, and adjacent routing markets will also contribute incremental growth.

Management explicitly stated that the front-end data center business will not "flatten," the enterprise business is still expanding, and the routing business, covering edge routing, cross-data center interconnection (DCI), and scale across scenarios, is seeing its market space reopen. This means Arista presents a model where "all product lines can grow if they can ship," rather than a logic of single-point explosions.

AI Network Divided into Three Battlefields, Arista Bets on Scale Out and Scale Across

During the conference call, management clearly divided AI fabric into three categories: scale up, scale out, and scale across.

Scale up mainly occurs within racks or for short-distance interconnections, where NVLink dominates in Nvidia's complete vertical solution, leaving limited room for Arista's participation. The latter two categories are truly suitable for Arista to leverage: scale out connects large-scale AI clusters, while scale across connects AI infrastructure between different campuses, different data centers, and different regions.

Management pointed out that the strategic importance of scale across is rising because computing power, physical space, and electricity face scarcity constraints, making it difficult for customers to permanently concentrate all computing resources in a single location. Arista's 7800 platform is positioned as the core spine for distributed scale across applications, emphasizing traffic isolation, contextual routing, and security capabilities.

The call also provided estimates for market size: the total addressable market (TAM) for scale across switching and routing is expected to be approximately $15 billion to $20 billion by 2030, with the market size in 2026 estimated at $3 billion to $4 billion. Arista expects to reach approximately $1.2 billion in this segment this year. Management stated that this business is still in its early stages, but the growth curve is steep.

Technical Differentiation Focuses on XPU Utilization, Not Just Bandwidth Competition

During the conference call, management elaborated on three technical capabilities to address external doubts about the competitiveness of Ethernet solutions.

SSU (Smart System Upgrade) aims to minimize service interruption during switch software upgrades. In AI training and inference clusters, XPU resource costs are extremely high, and downtime caused by network device restarts directly impacts computing utilization. The core value of SSU lies in resolving the conflict between secure upgrades and continuous computing operation.

MRC (Multipath Reliable Connection) addresses the issue in traditional AI networks where a single traffic flow can only take one path. It allows a single traffic flow to be dispersed across multiple paths within the fabric and reassembled at the receiving end, aiming to reduce performance losses caused by path conflicts.

SRv6 is responsible for path control, specifying transmission paths by tagging packets with segment IDs, and moving traffic away from hot links by combining real-time congestion signals. Management emphasized that Arista's advantage lies not in individual switches, but in the consistency of the EOS unified operating system extending from scale out fabric to long-distance scale across routing.

On the hardware side, the 7060XE7 is listed as the new generation of AI network products, featuring 100Tb capacity and 1.6Tb throughput, and introducing liquid cooling options. 1.6T related products will enter the trial phase in the second half of the year, with the number of customers expected to be in the single digits but large in scale; mass production is still planned for 2027.

Supply Chain Is the Core Hard Constraint, Tightness May Persist Until 2028

During the conference call, COO Todd Nightingale explicitly defined the supply chain as the main bottleneck for current growth and stated that the company has been continuously investing over the past six months to enhance delivery certainty.

Multi-year purchase commitments increased from $3.6 billion a year ago to approximately $9.7 billion at the end of Q2, further rising from $8.9 billion at the end of Q1, mainly corresponding to new products and chip purchases related to AI deployments. In terms of memory, supply for 2026 has been locked in, with visibility for DDR4, DDR5, and NAND extending to 2027; PCB and optical module capacity can be built within a 12-month window; the liquid cooling supply chain has been established, covering cold plate, quick-connect, and piping suppliers.

Manufacturing and distribution capabilities are expanding simultaneously: there are currently three contract manufacturers and three distribution facilities, distributed across the United States, Asia, and Mexico. Nightingale stated, "We have increased our capacity in manufacturing and distribution, and negotiated better component delivery terms."

Management also admitted that industry-level component shortages and cost increases are continuing, and extended their prediction for the duration of this cycle to 2028. Inventory at the end of Q2 was $2.5 billion, higher than $2.4 billion in Q1, with inventory turnover maintained at 1.7; deferred revenue was approximately $6.9 billion, higher than $6.2 billion in the previous quarter, with product deferred revenue increasing by approximately $600 million quarter-over-quarter, reflecting the time lag caused by new product ramp-ups and customer acceptance rhythms.

Cost Pressures Remain, But Margin Bottom Has Passed

The quarter-over-quarter improvement in gross margin partly came from tariff refunds, with management stating during the call that this item contributed approximately 20 to 30 basis points, while the rest largely came from improvements in customer mix. The impact of price increases has not yet significantly reflected in current results, and may only truly fall into financial figures by the end of the year or in 2026.

Full-year non-GAAP gross margin guidance remains in the 62% to 64% range, already including factors such as changes in customer mix, and rising memory and silicon costs; the operating profit margin target was raised from the previous range to 48% to 49%, showing that the company's expense control remains resilient despite increased investment in AI hardware, liquid cooling, and high-radix switch R&D expenditures.

Q2 operating expense structure showed R&D expenses of $278.1 million, accounting for 9.2% of revenue; sales and marketing expenses of $109.8 million, accounting for 3.6% of revenue; and G&A expenses of $23.1 million, accounting for 0.8% of revenue. The overall expense ratio remains low, constituting a relatively stable profit support in Arista's business model.

Customer Structure Anchored by Microsoft and Meta, Flexibility Exists for New Large Customers

During the conference call, management confirmed that Microsoft and Meta remain core large customers, with no change in commitments to these two clients. Meanwhile, management retained flexible wording regarding new customers contributing more than 10%: in 2026, there may be one, or possibly two, customers contributing more than 10% of revenue.

Management pointed out that there is no doubt about customer demand itself; the shipment structure depends on the supply chain, product priorities, and specific project rhythms. Among AI backend, front-end data centers, campus, and routing, whoever receives more supply contributes more revenue.

For Arista, the core proposition for 2026 has shifted from "proving whether there is demand for AI networks" to "stably delivering complex products in an industry cycle with tight supply." Quarterly revenue of $3 billion already demonstrates the intensity of demand, while the full-year guidance of $12.6 billion shifts the pressure to the execution side.