Hewlett Packard and Autonomy: Search and $4 Billion

February 12, 2024

green-dino_thumb_thumb_thumbThis essay is the work of a dumb dinobaby. No smart software required.

More than a decade ago, Hewlett Packard acquired Autonomy plc. Autonomy was one of the first companies to deploy what I call “smart software.” The system used Bayesian methods, still quite new to many in the information retrieval game in the 1990s. Autonomy kept its method in a black box assigned to a company from which Autonomy licensed the functions for information processing. Some experts in smart software overlook BAE Systems’ activity in the smart software game. That effort began in the late 1990s if my memory is working this morning. Few “experts” today care, but the dates are relevant.

Between the date Autonomy opened for business in 1996 and HP’s decision to purchase the company for about $8 billion in 2011, there was ample evidence that companies engaged in enterprise search and allied businesses like legal work processes or augmented magazine advertising were selling for much less. Most of the companies engaged in enterprise search simply went out of business after burning through their funds; for example, Delphes and Entopia. Others sold at what I thought we inflated or generous prices; for example, Vivisimo to IBM for about $28 million and Exalead to Dassault for 135 million euros.

Then along comes HP and its announcement that it purchased Autonomy for a staggering $8 billion. I attended a search-related event when one of the presenters showed this PowerPoint slide:


The idea was that Autonomy’s systems generated multiple lines of revenue, including a cloud service. The key fact on the presentation was that the search-and-retrieval unit was not the revenue rocket ship. Autonomy has shored up its search revenue by acquisition; for example, Soundsoft, Virage, and Zantaz. The company also experimented with bundling software, services, and hardware. But the Qatalyst slide depicted a rosy future because of Autonomy management’s vision and business strategy.

Did I believe the analysis prepared by Frank Quatrone’s team? I accepted some of the comments about the future, and I was skeptical about others. In the period from 2006 to 2012, it was becoming increasingly difficult to overcome some notable failures in enterprise search. The poster child from the problems was Fast Search & Transfer. In a nutshell, Fast Search retreated from Web search, shutting down its Google competitor AllTheWeb.com. The company’s engaging founder John Lervik told me that the future was enterprise search. But some Fast Search customers were slow in paying their bills because of the complexity of tailoring the Fast Search system to a client’s particular requirements. I recall being asked to comment about how to get the Fast Search system to work because my team used it for the FirstGov.gov site (now USA.gov) when the Inktomi solution was no longer viable due to procurement rule changes. Fast Search worked, but it required the same type of manual effort that the Vivisimo system required. Search-and-retrieval for an organization is not a one size fits all thing, a fact Google learned with its spectacular failure with its truly misguided Google Search Appliance product. Fast Search ended with an investigation related to financial missteps, and Microsoft stepped in in 2008 and bought the company for about $1.2 billion. I thought that was a wild and crazy number, but I was one of the lucky people who managed to get Fast Search to work and knew that most licensees would not have the resources or talent I had at my disposal. Working for the White House has some benefits, particularly when Fast Search for the US government was part of its tie up with AT&T. Thank goodness for my counterpart Ms. Coker. But $1.2 billion for Fast Search? That in my opinion was absolutely bonkers from my point of view. There were better and cheaper options, but Microsoft did not ask my opinion until after the deal was closed.


Everyone in the HP Autonomy matter keeps saying the same thing like an old-fashioned 78 RPM record stuck in a groove. Thanks, MSFT Copilot. You produced the image really “fast.” Plus, it is good enough like most search systems.

What is the Reuters’ news story adding to this background? Nothing. The reason is that the news story focuses on one factoid: “HP Claims $4 Billion Losses in London Lawsuit over Autonomy Deal.” Keep in mind that HP paid $11 billion for Autonomy plc. Keep in mind that was 10 times what Microsoft paid for Fast Search. Now HP wants $4 billion. Stripping away everything but enterprise search, I could accept that HP could reasonably pay $1.2 billion for Autonomy. But $11 billion made Microsoft’s purchase of Fast Search less nutso. Because, despite technical differences, Autonomy and Fast Search were two peas in a pod. The similarities were significant. The differences were technical. Neither company was poised to grow as rapidly as their stakeholders envisioned. 

When open source search options became available, these quickly became popular. Today if one wants serviceable search-and-retrieval for an enterprise application one can use a Lucene / Solr variant or pick one of a number of other viable open source systems.

But HP bought Autonomy and overpaid. Furthermore, Autonomy had potential, but the vision of Mike Lynch and the resources of HP were needed to convert the promise of Autonomy into a diversified information processing company. Autonomy could have provided high value solutions to the health and medical market; it could have become a key player in the policeware market; it could have leveraged its legal software into a knowledge pipeline for eDiscovery vendors to license and build upon; and it could have expanded its opportunities to license Autonomy stubs into broader OpenText enterprise integration solutions.

But what did HP do? It muffed the bunny. Mr. Lynch exited and set up a promising cyber security company and spent the rest of his time in courts. The Reuters’ article states:

Following one of the longest civil trials in English legal history, HP in 2022 substantially won its case, though a High Court judge said any damages would be significantly less than the $5 billion HP had claimed. HP’s lawyers argued on Monday that its losses resulting from the fraud entitle it to about $4 billion.

If I were younger and had not written three volumes of the Enterprise Search Report and a half dozen books about enterprise search, I would write about the wild and crazy years for enterprise search, its hits, its misses, and its spectacular failures (Yes, Google, I remember the Google Search Appliance quite well.) But I am a dinobaby.

The net net is HP made a poor decision and now years later it wants Mike Lynch to pay for HP’s lousy analysis of the company, its management missteps within its own Board of Directors, and its decision to pay $11 billion for a company in a sector in which at the time simply being profitable was a Herculean achievement. So this dinobaby says, “Caveat emptor.”

Stephen E Arnold, February 12, 2024


Got something to say?

  • Archives

  • Recent Posts

  • Meta