Sun Microsystems Inc, the darling of the dot-com boom, never really recovered from the bet it made on high-powered computers used to run the first generation of Internet datacentres.

But it's those same high-performance machines, favored by many telecom operators and financial services companies to handle high-transaction volumes in their switchboards or trading exchanges, that may prove to be its saving grace.

After decades of pursuing an idiosyncratic product strategy that resisted the growing standardization of the rest of the computer industry, Sun is in talks to be acquired by IBM for $6.5 billion in cash, according to the Wall Street Journal, and investors are betting a deal will happen.

The 100 percent premium on IBM's potential offer will be welcome news to Sun's long-suffering shareholders, who have endured a flatline stock price since early this decade and 70 percent plunge in its share price over the past year alone. IBM can take some solace in Sun's $3 billion in cash and short-term securities, or $1 billion or so of net cash.

The key attraction for IBM is likely to be Sun's high-end SPARC computer franchise, which, while competitive with IBM's own range of Power-class servers, mainframes and supercomputers, would bolster IBM against Hewlett-Packard's own transaction-heavy machines sold to banks, brokers and telecom operators.

Down the road, we also may look back at a IBM-Sun merger as partly due to this week's heavily marketed entry of network equipment maker Cisco Systems into the computer server market long dominated by IBM, H-P, Sun and Dell.

Cisco's land grab in the server market is likely to have more immediate impact on smaller data storage players. Cisco's moves are little threat to major server players IBM and HP. But its stated ambition to help define the next generation of Internet data centers is bound to prove a challenge to Sun's historic franchise in this market.

In many respects, Sun's strategy looks like a losing version of IBM's successful diversification over the past decade out of hardware and into computer services and software. Forty percent of Sun sales come from computers; 15 percent from storage, 40 percent from services to install and maintain products, with a tiny fraction from software. Most of these pieces will dovetail nicely with IBM's far larger businesses in these areas and may overcome IBM's often-stated preference for smaller software and services mergers over the past decade.

Sun's ambitions to move into software through acquisitions like the $1 billion deal to acquire database maker MySQL have been slow to produce results. Sun, which invented Java software in the early days of the Web, failed ever really to commercialize that franchise. Java has become a product brand and a ticker symbol rather than any sort of money-spinner for Sun. IBM and Sun both share a common commitment to open-source software that might hamstring other computer mergers.

Its revenue peaked in 2001 and remains 25 percent below its historic high level. Sun showed next to no revenue growth during 2008. It has reported losses in five of the past eight years, including last year. Its historic strength in financial services has become its latest bad bet.

But after years of gloom, that early bet on computers used to run the Internet may finally be coming good, with the support of IBM.