[SINGAPORE] After years of seeing great returns from investing in software companies, investors are now confronting the question of where returns will come from as the world embraces artificial intelligence (AI).
According to a report by consultancy Bain & Co, the "easy money" era of picking up a promising software-as-a-service company and watching revenues and multiples explode is coming to an end.
Even as tech deals as a percentage of all buyouts rose in the first half of 2025 to 22 per cent in North America, from 19 per cent at the end of 2024, the penetration curve in many areas has started to flatten. Software spending among companies is also easing off, and software revenue as a percentage of US gross domestic product has started to ease off.
Software companies are no longer competing in white space, or untapped market opportunity, but in spaces where companies have already tapped software.
"Relative digital laggard sectors like construction may have more room to run but, overall, growth derived by a company simply showing up in an underserved market will be harder to come by in the years ahead," added Bain in the report.
Returns will now depend on finding new sources of revenue growth and squeezing margins through operational efficiencies. This will come from displacing competitors, tapping AI, using modern pricing models such as outcome or value-based pricing, and expanding into other markets.
Software companies can also build new capabilities such as payments into their products or monetise data to find new revenue sources.
AI's need for compute power has now outstripped Moore's Law, which says the number of transistors on an integrated circuit doubles about every two years. In fact, AI's demand for compute power, the number of computations needed to support the model, has grown at more than twice that rate over the last decade.
Bain forecasts that total global compute requirements could reach 200 gigawatts by 2030, with the US taking up about half of that demand.
"Because AI compute demand is outpacing semiconductor efficiency, the trends call for dramatic increases in power supply on grids that have not added capacity for decades," said David Crawford, chairman of Bain's global technology practice.
To overcome this hurdle, the algorithms have to get better and more efficient to lower computational load. Technological breakthroughs such as the use of specific AI chips over general purpose graphics processing units will also need to occur to improve power efficiency.
"Working through the potential for innovation, infrastructure, supply shortages, and algorithmic gains is critical to navigate the next few years," Crawford added.