Here's what's strange about the AI networking narrative right now. Wall Street treats Arista like the pure-play AI networking winner and Cisco like the legacy dinosaur stumbling into AI by accident. That framing made sense two years ago, but I don’t think it does anymore.

Cisco just reported Q1 FY2026 earnings in November that beat on every metric, revenue at $14.9 billion (up 8% year-over-year), non-GAAP EPS of $1.00 (up 10%), and product orders accelerating at 13% growth. More importantly, AI infrastructure orders from hyperscalers totaled $1.3 billion in a single quarter, with management projecting $3 billion in recognized AI revenue for FY2026. Chuck Robbins compared the current AI infrastructure build to the late-1990s internet expansion. Bold comparison, but the order book backs it up.

Arista, meanwhile, hasn't reported yet. Q4 2025 results drop tomorrow (February 12), with consensus expecting $2.37 billion in revenue and $0.75 EPS. The company guided to $10.65 billion in total 2026 revenue, solid 20% growth, with $2.75 billion coming from AI. These are VERY good numbers.

My issue isn't with Arista's business. It's the price you're paying for it.

I think most investors are still mentally anchored to the Cisco of 2022-2023, when networking revenue was declining and the stock was stuck in the mid-$40s. That company doesn't exist anymore.

The AI pivot has been more substantive than the market gives credit for. Cisco's Silicon One chip family, which now powers both switching and routing for AI workloads, has shipped over one million units. The recently launched G300 chip delivers 102.4 terabits per second of switching capacity, which puts it in direct competition with Broadcom's Tomahawk and Arista's platforms for AI back-end networking. Every major hyperscaler is now a Silicon One customer, according to the Q1 earnings call.

And the numbers are backing this up. Networking revenue grew 15% in Q4 FY2026, with service provider routing posting high double-digit growth driven primarily by AI infrastructure. Data center switching and enterprise routing both contributed double-digit gains. This is a company shipping AI infrastructure at scale.

The campus refresh cycle adds another growth lever that Arista can't match. Cisco's enterprise installed base is massive with 86,200 employees across 90 countries serving the world's largest organizations. The company described a "multi-year, multi-billion-dollar campus refresh opportunity" as customers upgrade to WiFi 7, next-gen switching, and secure routers. Arista has a campus business, but it's a fraction of Cisco's reach.

Full-year FY2026 guidance of $60.2-$61.0 billion in revenue and $4.08-$4.14 non-GAAP EPS implies 6-7% top-line growth with meaningful operating leverage.

Not explosive, but growing. And at 21x forward earnings with a 1.9% dividend yield, you're not paying for perfection.

I want to be clear about something. Arista is a phenomenal business. It really is. The financials are almost obscenely good with 65.2% gross margins in Q3 2025, $10.1 billion in cash with essentially no debt, and a net income margin above 41%. These are software-like economics.

The AI story is compelling, too. Arista's partnerships with OpenAI, Anthropic, and Nvidia give it direct exposure to the most important AI infrastructure buyers on the planet. The Extensible Operating System (EOS) has become the de facto standard for many cloud-native architectures. And the company's $1.5 billion AI revenue target for 2025, growing to $2.75 billion in 2026, represents an 83% growth rate in its highest-value segment.

But here's where my enthusiasm hits a wall.

As of today Arista trades at roughly 55x trailing earnings, 45x forward, and carries a PEG ratio of 2.39. The stock is down 13% from its October high of $165, but that decline came after shares dropped 23% in just 21 trading days on weaker-than-expected margin guidance for Q4 2025. That kind of volatility is the tax you pay for premium valuations.

The question I keep coming back to is what has to go right for Arista to justify 45x forward earnings?

At $143 with roughly $3.00 in estimated 2026 EPS, the stock needs to grow earnings at 25-30% annually for the next three years just to get the forward P/E down to 25x by 2029.

That's not impossible. Arista has the business quality to do it, but it leaves zero room for execution misses, margin disappointments, or competitive encroachment from exactly the kinds of products Cisco just launched.

The insider activity doesn't inspire confidence either. CEO Jayshree Ullal sold 24,042 shares in November, and director Charles Giancarlo sold 7,900 shares at around $140 in early February. Insiders sell for plenty of legitimate reasons, but the timing right before earnings is definitely worth noting.

Morgan Stanley recently cut its Arista price target from $171 to $159 (still Overweight), while Rosenblatt raised its target from $140 to $165 but kept a Neutral rating. The analyst community is conflicted, and one could argue that this implies the easy money has been made.

Let me put the comparison in concrete terms, because I think the numbers tell a story the narrative obscures.

Revenue Scale: Cisco generates roughly $60 billion annually versus Arista's ~$10.65 billion (2026 target). Cisco is 5.6x larger.

AI Revenue: Cisco guides to $3 billion in AI infrastructure revenue from hyperscalers in FY2026. Arista targets $2.75 billion in AI revenue for calendar 2026. Cisco's AI number is actually bigger—and almost nobody talks about this.

Growth Rates: Arista is growing total revenue at ~20% versus Cisco's ~7%. That's a real differential. But Cisco's AI-specific revenue is growing faster than Arista's overall company.

Valuation: Cisco at 21x forward earnings. Arista at 45x forward. That's a 114% premium.

Capital Return: Cisco returned $3 billion to shareholders in Q1 FY2026 alone (125% of free cash flow), pays a 1.9% dividend, and has $14.2 billion remaining in buyback authorization. Arista returns nothing—no dividend, no meaningful buyback program.

Cash Position: Arista holds $10.1 billion. Cisco holds $15.7 billion with $28.8 billion in deferred revenue providing forward visibility.

Risk Profile: Cisco's beta is 0.87. Arista's is 1.42. In a market downturn, Arista falls harder.

I'm not arguing Cisco is a "better" business than Arista. On a pure-quality basis if you look at margins, growth rate, competitive position in cloud networking, Arista wins. But investing isn't about owning the best business. It's about owning the best risk-adjusted return, and that calculation increasingly favors Cisco at these prices.

Cisco reports Q4 FY2025 after the close today. Consensus expects around $15.1 billion in revenue and roughly $1.02 in non-GAAP EPS (the company guided $15.0-$15.2 billion revenue and $1.01-$1.03 EPS). The stock is at all-time highs, which means expectations are elevated. What I'm watching: any update on the $3 billion AI infrastructure revenue target. If management raises that number or provides more granular AI revenue disclosure, it could force a rerating.

Arista reports Q4 2025 tomorrow after the close. Consensus sits at $2.37 billion revenue and $0.75 EPS. The critical variable is 2026 guidance, specifically, whether the $10.65 billion revenue target and $2.75 billion AI target get reaffirmed, raised, or (the nightmare scenario for longs) tempered with cautious commentary about margin pressure.

The November selloff showed how brutally the market punishes Arista for any deviation from perfection. A 23% decline in three weeks because margin guidance came in slightly below expectations. At 45x forward earnings, you don't get benefit-of-the-doubt pricing. You get guillotine pricing.

The biggest pushback I hear on Cisco is that it's a "legacy" company trying to bolt AI onto a mature business. But that critique misses three things.

First, networking infrastructure is inherently essential to AI. You can't run a 100,000-GPU cluster without the switching fabric connecting those GPUs. Cisco isn't pivoting to AI, instead AI is coming to Cisco's core competency. There's a difference.

Second, the Splunk acquisition (closed March 2024 for $28 billion) gives Cisco an observability and security layer that Arista doesn't have. AI infrastructure needs monitoring and security. Having networking, observability, and security in one stack is a competitive advantage that's hard to replicate.

Third, Cisco's customer base is its moat. With 86,200 employees and relationships with essentially every enterprise and government agency globally, the campus refresh cycle alone represents billions in recurring upgrade revenue that has nothing to do with AI, it's pure installed-base monetization.

For Cisco: The biggest risk is execution disappointment on AI revenue conversion. Cisco has $3 billion in projected AI infrastructure revenue, but converting orders to recognized revenue in networking can lag. If tonight's report shows slower AI revenue recognition, the stock pulls back from all-time highs. Security revenue declining 2% and collaboration down 3% in Q1 are secondary concerns. That’s not fatal, but not great either.

For Arista: Concentration risk is the primary concern. Arista derives roughly 40% of revenue from its top two customers (widely believed to be Microsoft and Meta). If either hyperscaler pauses or redirects AI capex, the impact on Arista is disproportionate. The 45x valuation also means any guidance miss, even a slight one, triggers the kind of violent correction we saw in November.

For both: A broader macro slowdown that delays enterprise IT spending would hurt Cisco more (larger enterprise exposure), while a pullback in hyperscaler AI capex would hurt Arista more (higher AI revenue concentration).

I'm buying Cisco at $86 with a 12-month price target of $95 (roughly 11% upside plus 1.9% dividend = ~13% total return). The thesis is straightforward - AI infrastructure revenue accelerating, campus refresh cycle providing multi-year tailwinds, shareholder-friendly capital allocation, and a valuation that doesn't require heroic assumptions.

I'm holding off on Arista at $143. Not because the business is broken, far from it, but because the valuation demands flawless execution, and the risk-reward is asymmetric in the wrong direction. A 10% beat on earnings might push the stock up 5-8%. A 5% miss could send it down 20%. I'd revisit below $120, where you'd be paying closer to 35x forward earnings for genuine 20%+ growth, a much more reasonable starting point.

Sometimes the less exciting stock is the better investment. I think this is one of those times.

I have no stock, option or similar derivative position in any of the companies mentioned, but may initiate a beneficial long position through a purchase of the stock, or the purchase of call options or similar derivatives in CSCO over the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.