In the great AI spending spree, does ROI matter?Thats a question that every tech-related company should be asking themselves. The right answer could save businesses a lot of grief while setting them up for success.There will be a projected $1 trillion in spending on artificial intelligence in the coming years. But not all spending is created equal. To give you an idea of how much economic value needs to be created with AI, lets consider chips. If AI providers spend $150 billion just on chips, assuming they need a 50-60% gross margin, they will need to sell $500-$600 billion in services. Those buyers will then need all those AI services to create $1 trillion in economic value.It may seem like a long way from here to ROI, but that math is similar to what the U.S. was facing at the start of the internet boom. Today, the debate about whether the U.S. is overspending on AI plays differently across industries, depending on how much your companys existence depends on it.I group companies into three categories:1. Big TechFor the hyperscalerstechs big cloud companies like Microsoft, Google, Amazon, and Meta, there are two sides to consider. AI could open paths to even greater success and relevance in the future. Or AI could create a whole new level of innovation that disrupts their core business. Either way, those companies have every reason to invest heavily.Without knowing where the payoff will come from or what the killer app will be, Big Tech will have to explore nearly every avenue. There will be plenty of unnecessary investments, a few dead ends and even some fortuitous outcomes, but it will take years for this to play out.Fortunately for these companies, ROI wont be an immediate pressure, as long as their overall financial performance is solid and they are generating enough cash to make massive AI investments.2. Supply chainFor companies that make chips, servers, or operate data centers, there is a need to overinvest in capacity. Their priority is to meet demand, even if it means overestimating and having a surplus. In times of major AI investment its important to have ample AI infrastructure available, or your customers will go somewhere else.Its the game of supply and demand. Some companies may overextend, and others will fall short. If some collapse, there will be consolidation. Assuming theres ongoing demand, valuations will return to market levels.Of course, when you are an AI infrastructure supplier, your attention is on the volume of demand right in front of youthat makes your ROI time horizon more of a secondary concern.3. Applied AIFor companies already using AI, its easier to take a more measured approach to spending. At Verizon, we aim to be the best applied AI company in the U.S., so we make careful investments by setting ROI targets.This approach also helps to develop an AI spending strategy that starts at the implementation level. Earlier this year I cautioned against a centralized AI decision process and instead suggested empowered teams closer to the actual work to provide the insights.The agentic intertwining of AI with frontline agents is a pioneering area of customer care thats already paying off. Weve seen improvements in areas like call efficiency and effectiveness, personalization and satisfaction.With the right input you can focus investment on areas where youve shown progressareas where AI measurably improves your KPIs.History lessonSo are we overestimating the potential value of AI? Maybe not.Twenty-five years ago, the dotcom boom fueled massive investment in internet-based startups and fiber infrastructure. The spending may have been irrational but if you thought the internet was going to play a big part in our future, the logic was sound.When the bust came and the exuberance finally faded, there were a few survivors who went on to prosper. Others went broke, leaving some very attractive assets, like MCIs fiber network, to fall into our hands.For the most part, the internet worked out pretty well. Today, AI presents another enormous opportunity. As history has shown us, you have to take big risks. Right now, the U.S. is in the best position to do that.Sowmyanarayan Sampath is CEO of Verizon Consumer.