AI Infrastructure Orders Inflect, Splunk Synergies Tracking Ahead, Tariff Overhang Caps Near-Term Multiple — Initiating Coverage at Hold, FV $58–68

Rating Action

We are initiating coverage of Cisco Systems, Inc. (CSCO) at Hold with a fair-value range of $58–68, anchored on ~17–19x our FY25 non-GAAP EPS framework of $3.55–3.65 (modeling toward the lower half of the company’s $3.68–3.74 guide given tariff risk).

The investment case has two genuine compounders — AI infrastructure order growth and Splunk-anchored Security — sitting on top of a stable Networking annuity that is now showing fourth consecutive quarter of accelerating order growth. Where we differ from the more bullish framing on the Street is on multiple expansion: at current levels, the stock has already absorbed much of the AI-orders-inflecting story, and the gating questions for an Outperform call (does enterprise AI orders ramp meaningfully in 2H, does Splunk cross-sell convert into broader-platform deals, does the tariff regime resolve favorably) are not yet answered. We want to see Q3 print to validate the campus refresh narrative and confirm AI orders break $1B with one quarter to spare.

We will revisit the rating at Q3 FY25 with explicit upgrade triggers: (1) AI orders YTD >$1B with at least one new top-six webscaler in the order book; (2) organic product orders ex-Splunk sustaining low-double-digit growth; (3) Security ex-Splunk accelerating from 4% toward high-single-digit; (4) FY25 gross margin guide stabilizing in the 68–69% range as tariff resolution clarifies.

Results vs. Consensus

Beat across every line that matters. The composition is the takeaway: order growth (29% reported / 11% organic) is the leading indicator, AI infrastructure orders are the durability indicator, and the FY25 guide raise (revenue $56.0–56.5B from $55.3–56.3B; EPS $3.68–3.74 from $3.60–3.66) is management putting some of the upside back into the model.

Segment Performance

Networking — Annuity Stabilizing, Data-Center the Engine

Reported Networking revenue was down 3%, but the optics matter: Q2 FY24 carried the last portion of elevated-backlog shipments, distorting the year-over-year. The forward-looking metric is order growth, where Networking grew double-digits driven by switching, enterprise routing, webscale infrastructure, and industrial-IoT.

• Campus switching: Orders up double-digits; WiFi 7 access points beginning to gain traction with return-to-office cycle. Two-to-three-quarter ramp expected in WiFi 7.
• Data-center switching (Nexus): Fourth consecutive quarter of double-digit order growth. The 800G Nexus on Silicon One 51.2T ships April — the inflection lever for enterprise AI infrastructure.
• Industrial IoT (ruggedized Catalyst): 1H FY25 orders >+40%; Q2 alone >+50%. Reads as a beneficiary of US re-shoring / industrial AI / sovereign infrastructure spend.
• Webscale: SP & Cloud orders +75%, with three of the top six webscalers up triple-digits and two more up >50%. AI infrastructure orders >$350M Q2.

Security — Splunk-Anchored, Organic Still Modest

• Reported revenue: +117% YoY (Splunk + SASE + network security).
• Organic ex-Splunk: +4%. The last four quarters showed sequential improvement in organic security demand — the question is when this converts to mid-to-high-single-digit organic growth.
• Hypershield: Two Fortune 100 platform deals booked; Splunk-led cross-sell drove a third platform deal involving multiple Cisco products and services.
• Cisco Secure Access & XDR: Each >1,000 customers and ~1M enterprise users in 12 months — a credible early ramp.
• AI Defense: Launched at AI Summit; ~20+ proof-of-concept candidates from the half of the Fortune 100 that attended; GA in March.

Collaboration — Stable, On-Prem Webex Decline Persists

Up 1%. Growth in Contact Center, CPaaS offerings, and Collaboration Devices, partially offset by ongoing decline in on-prem Webex Suite. Not a thesis driver in either direction; we model low-single-digit decline through the Webex transition completion.

Observability — Splunk-Anchored Like Security

• Reported +47% YoY.
• Organic ex-Splunk +3%.
• AppDynamics now integrated into Splunk’s on-prem log observer; Talos integrated into Splunk Enterprise Security 8.0.

Key Topics & Management Commentary

The AI-Orders Trajectory: $350M → $1B → What’s Next

Management committed to >$1B AI infrastructure orders for FY25 with $700M already in the YTD book through 1H. The composition Robbins disclosed is the analytical kicker:

“On the $700 million in AI orders, it’s a combination of systems, silicon, optics and optical systems. And I think if you break it down, it’s about half is in silicon and systems.”
— Chuck Robbins, CEO

The webscale customer concentration disclosure also matters: Robbins said combinations of Cisco 8000, Silicon One, optics and optical systems “are being deployed by five of the largest webscalers in their back-end training networks.” Asked about deployment cadence, Robbins offered the most thesis-relevant operating data point of the call:

“I’ve set with all these customers, and they’ve looked at me and said, if you can build more, we will buy more. So I think that a couple of them had initial targets for how many units they wanted in 2025, and that number has already gone up by 50%.”
— Chuck Robbins, CEO

On revenue conversion: Herren reaffirmed the prior framing that orders flow into revenue with a lag, with the ramp expected in 2H FY25. We model AI revenue contribution to FY25 revenue at the low end — the FY26 ramp is where the real numbers are.

Splunk Integration: Profitability Ahead, Revenue In Line, Leadership Transition

The integration narrative landed cleanly: profitability “well ahead” of expectations, revenue “in line.” A single sales-calendar quirk — Splunk’s historical fiscal-Q4-week-13 (the largest Splunk week of the year) fell into Cisco’s Q3 not Q2 this year — explained the surface-level Splunk math that Tal Liani flagged in Q&A.

The leadership transition: Gary Steele (President of Go-to-Market, former Splunk CEO) is leaving end of Q3 for an external CEO opportunity. Robbins acknowledged Steele’s integration role as “instrumental” and disclosed an internal/external search has launched. With integration largely complete and the Splunk synergy machinery operating, we treat this as a manageable transition rather than a thesis variable.

Tariffs: The Principal Near-Term Margin Variable

Herren’s framing on the Q3 gross margin guide was unusually direct:

“What I built into the cost of goods sold, that’s where the tariffs will hit, is the full effect of 25% in Mexico, Canada and additional 10% in China, with no mitigation at this point.”
— Scott Herren, CFO

The tariff math: prior China-tariff mitigation reduced exposure by ~80% over time. Cisco has “a number of steps” planned for various tariff scenarios. Pricing is “a lever” but not the first one Herren would pull. Read: the 67–68% Q3 GM guide is intentionally conservative; if tariffs land softer than the worst case, there’s upside to that number. If they land harder or expand, the downside is real. We model the midpoint and treat the dispersion as live risk.

Hypershield + DPU Switch: Network-Embedded Security as a Structural Vector

The Q2 announcement of a 9300-series Nexus with embedded programmable AMD DPUs got Q&A attention from Ben Reitzes (Melius). Robbins’ framing was strategically meaningful:

“You can run network services or security services on these DPUs in the network at the speed of ASICs as opposed to moving this traffic to a central place where you’re going to be limited by the processing power of a single box.”
— Chuck Robbins, CEO

The strategic implication: if Hypershield + DPU switches succeed at displacing security appliances by embedding security at the network layer, this is a multi-billion-dollar TAM expansion vector that doesn’t show up in any backward-looking Networking line. We flag it as upside optionality in the FY26–FY27 model rather than current quarter.

Federal & DOGE: De Minimis Direct Exposure

Asked by Simon Leopold (Raymond James) about DOGE-related federal-spend risk: US federal <10% of total business; 75% of US federal business is DoD. Robbins flagged a brief flurry of confusion at executive-order timing but said large deals “got back on track.” FY25 modeling assumes no material federal acceleration or deceleration. We do not treat DOGE as a thesis input.

Capital Return: $15B Buyback Authorization Signals Confidence

Q2 capital return $2.8B ($1.6B dividend + $1.2B buyback). The $15B incremental authorization brings total outstanding to ~$17B. Combined with the 14th consecutive annual dividend raise, the message is simple: management has high confidence in the cash-flow trajectory and expects to deploy capital aggressively against the share count. We model a buyback pace that materially exceeds the prior trailing-twelve-month run-rate.

Q3 FY25 & Full-Year Guidance

Read: The FY25 raise of $700M at the revenue midpoint and $0.08 at the EPS midpoint is meaningful but not parabolic; it largely reflects Q2 outperformance flowing through the model, not a structural reset of FY25 expectations. The Q3 gross margin guide is the single most important cautionary data point in the print — the 67–68% range is roughly 100bps below where we’d underwrite a clean FY25 trajectory absent tariffs. The fact that management did not raise the FY25 GM guide explicitly is the conservative posture; we read this as Herren preserving optionality on tariff outcomes rather than signaling an underlying margin problem.

Analyst Q&A — Notable Exchanges

Q&A was notably AI-orders-heavy this quarter, reflecting where the buyside thesis is converging. Notable threads:

• Simon Leopold (Raymond James) opened on DOGE federal exposure (de minimis, <10% with 75% DoD) and followed on co-packaged optics — Robbins kept the response architecture-agnostic, framing the go-to-market as systems-with-integrated-optics and standalone-optics flexibility.
• Meta Marshall (Morgan Stanley) probed customer demand propensity; Robbins gave the broadest possible framing of strength — geographies, portfolio, segments — with the “fear of slowing down too much and letting your competition get too much ahead of you” line as the most thesis-relevant demand framing.
• Ben Reitzes (Melius) got the Hypershield + DPU strategic framing — the most architecturally significant disclosure of the call. His follow-up confirmed enterprise AI prep is real but early.
• Matt Niknam (Deutsche Bank) drew the cleanest tariff disclosure: full proposed regime in COGS, no mitigation built in. The 67–68% Q3 GM is intentionally conservative.
• Samik Chatterjee (JPMorgan) probed for tariff-driven demand pull-forward (none seen) and webscale deployment cadence (Robbins disclosed several webscalers raised their 2025 unit targets by 50%).
• Tal Liani (BofA) raised the Splunk-down-11% optics; Herren’s response on the calendar quirk (Splunk’s historical Q4-week-13 falling into Cisco Q3 not Q2) was clean. Splunk underlying still growing double-digits.
• Amit Daryanani (Evercore ISI) drew the AI-orders composition data point (~half silicon/systems, ~half optics/optical systems) and the revenue-conversion timing (2H ramp, in line with prior framing).
• Jim Fish (Piper Sandler) probed Splunk + Cisco joint customer penetration (“still early, long sales cycles”) and silicon-vs-systems sales mix (predominantly systems today).
• David Vogt (UBS) challenged organic enterprise order deceleration (Robbins: “literally took over $1B more in orders in Q2 than Q1”) and probed white-box / DeepSeek competitive risk; Robbins’ framing was that AI democratization accelerates Cisco’s addressable market rather than threatens it.
• Antoine Chkaiban (New Street Research) got the cleanest framing on enterprise AI today: it’s mostly Nvidia-stack with bundled networking; the Cisco Hyperfabric / AI POD / DPU thesis is forward-looking optionality, not current revenue.
• Aaron Rakers (Wells Fargo) drew the Silicon One 51.2T inflection framing: April availability, designed for multiple use cases (deep-buffer, low-latency, enterprise / hyperscale).
• Karl Ackerman (BNP) tested whether the demand strength was lead-time-driven (Herren: no, lead times are normalized).
• Atif Malik (Citi, via Adrienne Colby) closed on telco; Herren framed it as episodic with structural AI-driven network-load tailwinds.

What They’re NOT Saying

• No specific FY26 framework. Management gave FY25 guidance updates but did not yet frame FY26 expectations or AI-orders trajectory beyond FY25. We expect this at Q4 FY25 alongside the formal FY26 guide.
• No customer-concentration data on AI orders. Robbins confirmed five of the largest webscalers are deploying Cisco 8000 / Silicon One in back-end training networks, but no dollar split between webscalers, no “customer A is X%” framing. We treat the AI orders book as concentrated until proven otherwise — this matters for the durability premise.
• No FY25 gross-margin guide. Q3 GM 67–68% is given; Herren said this “settles in” in that range for the full year — but no formal FY25 GM range. Reads as preserving tariff-resolution optionality.
• No Splunk standalone revenue disclosure. Reported Security +117% includes Splunk; organic ex-Splunk Security +4%. We have ARR composition but not a clean Splunk standalone revenue trajectory. The leadership transition (Steele departure) makes this a quarter where investors would have valued more disclosure granularity.
• No commentary on M&A pipeline beyond Deeper Insights. Herren noted continued strategic-M&A focus on AI complementary to internal R&D; the Deeper Insights deal closed in Q2. No further deals quantified or signaled.

Market Reaction

The print landed after-hours on February 12. Initial after-hours reaction was constructively positive, with shares up ~6.6% after-hours on the print. The trading session on February 13 confirmed the buyside’s positive reception — AI-orders trajectory, the FY25 guide raise, and the $15B buyback authorization were the three planks the buyside underwrote. The tariff-overhang on the Q3 GM guide attracted the bear-side commentary but did not break the constructive narrative on the day. Volume was elevated; CSCO outperformed enterprise-tech peers on the session.

The reaction is consistent with a market re-rating the franchise on the AI-orders narrative without yet pricing in a structural reset of long-term growth assumptions. We read this as appropriately calibrated — the print supports the cycle thesis but does not yet unlock the multiple expansion that would come from a clean second print.

Street Perspective

The bull case being made on the Street post-print converges on three planks: (1) AI infrastructure orders are inflecting structurally, with FY25 >$1B as the visible floor and webscalers raising unit targets ahead of plan; (2) Splunk integration is delivering profitability ahead of plan, with cross-sell into Hypershield / Secure Access / XDR creating a multi-product platform motion that justifies a higher security multiple; (3) the FY25 guide raise plus a $15B buyback authorization signals management’s confidence in the cash-flow durability, supporting capital-return-driven EPS accretion through the cycle.

The bear case being articulated centers on: (1) the tariff overhang is a real near-term GM risk that management could not fully size, with the 67–68% Q3 GM guide intentionally conservative but also genuinely lower than a clean trajectory would deliver; (2) organic Security ex-Splunk at +4% is well below where the Splunk integration thesis implies it should be at this stage; (3) Networking reported revenue down 3% (even with the backlog-comp explanation) makes the “Networking is the annuity” framing harder to underwrite without more order-to-revenue conversion data; (4) the Gary Steele departure adds a leadership-transition variable that the bull thesis doesn’t price.

Our read sides with the bear framing on (1) and (3) and treats the bull framing on (2) as a thesis upgrade waiting on Q3 confirmation. This is why we initiate at Hold rather than Outperform.

Model Implications

• FY25 revenue: We model $56.0B (low end of company guide $56.0–56.5B), reflecting tariff-driven demand uncertainty and conservative AI-revenue ramp assumptions.
• FY25 non-GAAP EPS: $3.55–3.65, modestly below the $3.68–3.74 company guide. We embed 50–75bps of FY25 GM erosion vs the implied trajectory absent tariffs.
• Q3 FY25: Revenue $14.0B (midpoint), non-GAAP EPS $0.91 (midpoint). The 67–68% GM guide is the single most important variable; if tariffs resolve favorably, this is a meaningful upside lever.
• AI infrastructure orders: We model FY25 orders at $1.05B (slightly above management’s >$1B floor) with FY26 trajectory at $1.6–2.0B contingent on enterprise AI POD ramp. Revenue conversion lag is ~3–4 quarters.
• Splunk: We hold the integration thesis intact pending Q3 confirmation post-leadership-transition. Organic Security ex-Splunk at high-single-digit growth by year-end FY25 is the upside trigger.
• Capital return: We model ~$11B in FY25 capital return ($6.5B dividend + $4.5B buyback), well within the $17B authorization runway and consistent with the YTD $6.4B pace.

Thesis Scorecard

Overall: Genuine cycle vectors layered on a stable annuity, with two of four bull pillars confirmed and two more tracking constructively. Bear case is appropriately scoped — tariff-overhang and organic-security-ramp are the two near-term variables that will determine whether the next print justifies an upgrade.

Action: Initiating at Hold; fair value $58–68 (~17–19x our FY25 non-GAAP EPS framework of $3.55–3.65). The investment case is genuinely interesting but the multiple has already absorbed much of the AI-and-Splunk narrative. We want one more clean print before paying for the upgrade. We will revisit at Q3 FY25 with explicit upgrade triggers around AI orders >$1B with one quarter to spare, organic Security accelerating, and tariff resolution.

Net: A solid initiation print. The AI-orders inflection is real, the Splunk integration is delivering, and the capital-return posture is supportive. We initiate at Hold not because the franchise is broken but because the price is no longer cheap enough for the unresolved variables.