There is upside, too: If IBM can make a go of it with these water systems, it could also become a major player in the management of the electrical grid, where there is probably even more to be done than is the case with water systems.

We’re not sure how big the market is for water-management systems. IBM’s estimate of $20 billion a year sounds too optimistic. But, whatever the ultimate size of the market, it should be a good one for IBM.

The strangest idea is the suggestion by at least one securities analyst that Dell should buy Accenture. The rationale is that many big computer companies have large service operations, and Accenture could help Dell make inroads in corporate data centers. But come on. The Dell and Accenture businesses are so different that there would be no cost synergies. Revenue synergies would be minimal because Accenture would have to stay reasonably agnostic about the hardware its customers buy, or customers would go elsewhere in search of unbiased advice. So, the only way to justify the acquisition premium that Dell would have to pay is if Dell, with no experience in professional services, could run Accenture better than its current managers are. Not likely. For good measure, combining the two would create all sorts of complexity that the individual businesses don’t currently face.

Maybe the Accenture idea is just an analyst’s pipe dream, but the IBM negotiations to buy longtime rival Sun Microsystems are real [see 1, 2, 3], and that idea appears to be a classic mistake in a consolidating industry. IBM seems to be overestimating the savings that it can wring out of Sun, while underestimating the complexity of the deal.

IBM is assuming it can quickly knit the Sun line of servers together with IBM’s, but such transitions are notoriously difficult. While the idea looks straightforward on a PowerPoint slide, computers don’t exist on slides—they consist of incredibly complex, specialized hardware and layers of software containing millions of lines of code, and it takes years to bring disparate systems into harmony. In the meantime, competitors will launch savage attacks at Sun products, arguing that any customer even remotely unhappy with Sun should leave now rather than suffer through years of transition that will eventually turn Sun’s equipment into IBM-like machines. To see how this assault might play out, you can look at any number of disastrous consolidations in the high-tech world, such as the Burroughs purchase of Sperry to form Unisys in the 1980s, the Compaq acquisition of Digital Equipment in the 1990s or the Alcatel-Lucent merger in the 2000s.

Analysts say that IBM seems to be assuming that it can take $1 billion a year of cost out of Sun, because that’s what would have to happen for Sun to achieve the same level of profitability as IBM’s workstation business. Certainly, some cost can come out. Sun isn’t known as the leanest company around. Still, it has been whacking away at costs on its own ever since the dot-com bubble burst, and acquirers often talk themselves into seeing more savings than are really there. The Sun folks will still need offices, support staff, etc. It’s hard to see how IBM could find enough efficiencies to justify the 100% premium it is reporting considering paying, over the price of Sun’s stock before the negotiations were disclosed.

IBM may be feeling confident because it has made a series of mostly successful acquisitions in recent years, but those were far smaller than a Sun deal would be. In addition, Sun poses a much more complicated cultural challenge. Sun’s culture says that engineers rule, while, at IBM, the sales force dominates. That clash will take some sorting out. In addition, Sun has made a living for decades by mocking IBM, which won’t make things any easier.

Our research into failures suggests that, if IBM does buy Sun, the real beneficiary will be Hewlett-Packard, which will have a competitor disappear without having to spend $13 billion and without having to go through all the hassle associated with a major acquisition. Dell, which is hoping to build on its more modest success in data centers, might gain, as well—if it doesn’t buy Accenture.

Cisco is off on its own odd tangent. Its agreement to purchase Flip Video for $590 million purports to be a move into an adjacent market but really isn’t. While it’s true that traffic from the Flip cameras can be carried on Cisco networks as people send videos around, that doesn’t mean Cisco needs to own the camera business, any more than it needs to make its own mainframes, which also supply a lot of network traffic. In fact, mainframes have much more in common with Cisco’s core business than a consumer business like video cameras does. While Cisco has a stellar record with acquisitions, the Flip Video deal sounds rather like the Sony decision to buy Columbia Pictures for $3.4 billion in 1989, on the theory that Sony should have its own movies for customers to play on their Sony VCRs and TV sets. Sony soon took writedowns that equaled the value of its investment.

It’s not that we hate everything going on in the computer industry these days. Cisco, for instance, seems to be making a well-timed move into what really is an adjacent market, by introducing a line of servers. Cisco already has good connections with the buyers, who manage corporate data centers. The servers, which differ from competitors’ because of their networking capabilities, draw on Cisco’s technology strengths.

As we implied at the top, we also like Oracle’s decision to declare a dividend. While it’s been acquisitive, and generally successful, just because you’re sitting on a mound of cash doesn’t mean you should rush off and spend it.

In the early 1980s, IBM’s then-CEO John Opel declared that IBM would hit $100 billion in revenue by 1990. Although it may be hard to remember back that far, IBM was the world’s most profitable company in the 1980s. Its market capitalization accounted for roughly three-quarters of the value of the entire computer industry. Opel wanted to keep IBM from getting too complacent, so he challenged the company to increase in size from $40 billion of annual revenue in 1983 to the magic $100 billion mark by the end of the decade.

So far, so good, according to the theory of BHAGs. The problem, as it turns out, is that BHAGs can distort people’s behaviors. When the CEO of IBM speaks, he isn’t asking the organization to do something; he’s demanding. So, people will do things to satisfy the CEO even if they aren’t in the best interests of the business.

IBM built factories that would churn out the mainframes that would be needed to meet the revenue goal–even as personal computers took off and as minicomputers diminished the demand for mainframes. To try to slow demand for personal computers, IBM delayed using new chip technology from Intel–and promptly lost the PC market to Compaq and then others, who jumped on the latest from Intel. To goose revenue, IBM shifted away from a leasing model, in which IBM would book each year some fraction of the value of the machines on customer premises. Instead, IBM sold machines outright–increasing revenue in the short run but making the company much more vulnerable to short-term changes in demand. When mainframe competitors began winning customers, IBM declared that revenue was sacrosanct. Salesmen were to protect market share at all costs. Well, customers aren’t stupid. One talked about his “million-dollar coffee mug”–all he had to do was put the competitor’s coffee mug on his desk when an IBM salesman came calling, and the price of the IBM equipment would get discounted by a million dollars. IBM had ignited a price war that crushed profits.

IBM not only didn’t reach $100 billion of revenue by 1990; the company had messed itself up so completely that the company had a near-death experience in the early 1990s, and the team that had succeeded Opel was kicked out.

If you’re going to establish a BHAG, first think about all the ways it could lead to problems. How could competitors stop you? How could your own people distort your directives? (Remember, as IBM learned about its sales force, people can be very creative when trying to earn compensation.) Could some technology issue push you astray? Etc. Etc.

Proceed with humility. Remember that even Collins, as widely respected as he is, got it wrong when he singled out Motorola’s Iridium project as a laudable BHAG in “Built to Last.” Iridium didn’t just get the technology wrong, producing a satellite phone-system in the late 1990s that was comparable to the cellphone technology of the mid-1980s. Motorola and other investors also got the structure of the business wrong, when they made Iridium a stand-alone operation. Once investors turned over their money, no one had any incentive to kill the project, even as cellphone technology left the satellite operation far behind. Not Iridium management, which had options in Iridium that would be worthless if the venture was scrapped. Not Motorola, which saw Iridium as a huge customer for networking equipment. With no incentive to halt things, Iridium went ahead and spent the full $5 billion of investors’ money and opened for business in 1998, only to file for bankruptcy protection in 1999 and auction off its assets for $25 million.

In other words, it’s easy to get a BHAG wrong.