Cisco Didn't Build a Growth Engine by Buying Companies. It Built Two — and Only One Works.

In September 1993, Cisco did something a hardware company almost never does: instead of building the next product itself, it bought the people who already had. The target was Crescendo Communications. The price was two million shares of Cisco stock — worth somewhere around $94 to $97 million at the September 1993 closing price.³ No cash changed hands; Cisco simply printed equity and absorbed a team. Thirty-one years later, almost to the season, Cisco closed its largest deal ever: $157 a share for Splunk, about $28 billion, all of it in cash.¹ The arc from $97 million in stock to $28 billion in cash is the whole story of Cisco — and the trap inside it.

The official story is that Cisco is a master of growth by acquisition — one disciplined machine that has bought hundreds of companies and folded them in. That telling is too clean. Cisco does not run one acquisition machine. It runs two, and only one of them reliably works.

The first machine: buy the road, keep the traffic

Crescendo set a template Cisco would run for decades. The genius was never the technology Cisco acquired — it was the channel it already owned. Cisco sold routers and switches to enterprise IT buyers through a relationship sales motion. When it bought an adjacent networking company, it wasn't really buying a product. It was buying something it could push through a distribution pipe that was already full of customers who trusted it. Cisco even classifies its own deals this way, sorting them into market acceleration, market expansion, and new market entry, and treating integration as a discipline that starts at diligence, not after the close.⁶ The tuck-in deal works because the hard part — getting the customer to pick up the phone — is already done. Cisco buys a feature and sells it as part of a relationship.

This is why the record looks heroic when you stay inside the networking core. The acquired team plugs into an installed base; the product becomes a line item in a deal the customer was going to sign anyway; the integration debt stays small because the buyer and the target speak the same language to the same people. Before Splunk, Cisco's biggest single deals — Cerent in 1999 and Scientific Atlanta in 2005, each around $6.9 billion⁵ — still lived close enough to the wire that the same logic held. The further a deal sits from that channel, though, the more the machine has to actually integrate, and integration is where the second machine breaks.

The second machine: the swing into a market the channel can't reach

In 2009 Cisco paid $590 million for the maker of the Flip — a pocket video camera sold to consumers in big-box stores.⁷ Nothing about it fit. Cisco's superpower was the enterprise relationship; the Flip lived on a retail shelf next to point-and-shoot cameras, bought by people who would never know Cisco's name. There was no installed base to push it through, no IT buyer to upsell, no relationship to extend. About two years later Cisco shut it down and took a charge of roughly $300 million.⁷ One analysis noted that there was not an analyst 'on the planet' who had thought it was a good acquisition.⁷ The Flip didn't fail because Cisco overpaid. It failed because the second machine has no channel — and without the channel, the acquisition is just a product Cisco now has to sell from scratch, in a market it doesn't understand, to customers it has never met.

Why Splunk is the question, not the answer

Splunk is the logical end of Cisco's strategy and its hardest test at the same time. At about $28 billion in cash, paid at a 31% premium to Splunk's unaffected price⁸, it dwarfs every deal Cisco had ever done — more than four times the size of Cerent or Scientific Atlanta.⁵ The thesis is that security and observability data belong next to the network, and that Splunk's analytics platform is the new layer Cisco's customers will buy. That is a market-expansion bet, the kind Cisco names on its own corporate-development page.⁶ The problem is that Splunk is not a feature you slot into a router quote. It is a large, independent software business with its own buyer, its own pricing model, and its own decade of product gravity. The integration debt is not small. And the timeline already hints at the difficulty: Cisco announced the deal in September 2023 but didn't close it until March 18, 2024¹² — a six-month wait before the real work even began.

The honest framing: a $28 billion all-cash software acquisition succeeds or fails almost entirely on integration, which is precisely the muscle the tuck-in machine never had to build. Crescendo worked because Cisco barely had to integrate it. Splunk will work only if Cisco can do the thing it has historically avoided.

The fair objection: maybe Cisco has earned the benefit of the doubt

The strongest counter is that Cisco has been acquiring for three decades and is clearly still standing — so calling the big swings broken ignores how many of them landed. That's fair. Scientific Atlanta took Cisco into set-top boxes; plenty of mid-size software and security deals quietly became core. And the Flip is a single, lurid failure, not a pattern. But notice what survives and what dies. The deals that endure tend to sit close to the network and the enterprise buyer — even Scientific Atlanta sold gear that rode Cisco's infrastructure story. The ones that died, like the Flip, abandoned the channel entirely.⁷ The pattern isn't 'big deals fail.' It's that distance from the existing channel, not price, predicts the outcome — and Splunk is the largest, most distant software business Cisco has ever tried to carry on its own legs.

Cisco's reputation as an acquisition machine is real, but it was built by the small deals, not the spectacular ones — the unglamorous tuck-ins that arrived pre-sold through a channel Cisco already controlled. The mega-bets are a different animal wearing the same coat. Splunk is the purest version of that animal: the most expensive thing Cisco has ever bought, the furthest from the wire, and the truest test of whether the company can finally do the one thing its winning machine never required. The genius was never that Cisco bought things. It was that, for thirty years, it mostly bought things it could sell to people it already knew.